HR Wallingiord

HYDRAULIC DESIGN OF
STRAIGHT COMPOUND CHANNELS

VOLUME 1

by

P Ackers, Hydraulics Consultant

Content s o f Volume 1

Summary and design method

Deta i led development of design method ,
- Part 1

Report SR 281
OCTOBER 1991

Address: Hydraulics Research Ltd, Wallingford, Oxfordshire OXlO 8BA, United Kingdom.
Telephone: 0491 35381lntemational + 44 491 35381 Telex: 848552 HRSWAL G.
Facsimile: 0491 32233 International+ 44 491 32233 Registered in England No. 1622174
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by
any means, electron ic, mechanical, photocopying, recording or
otherwise without the prior permission of HR Wallingford.

C BR Wallingford, 1991
HYDRAULIC DESIGN OF TWO-STAGE CHANNELS HR REPORT SR 281,
DEC 1991

CORRIGENDA : to 22 Feb 1991, Volume 1

p9, line 2; should refer to plate 2

p20, last line levels should read levees

p4 7, para 2.4.8, 13; 0.43 should read 0.3

16; 0.85 0.5
0.94 0.9
18; 0.61 0.52

p53; the two sentences at the foot of this page should be at the head of p55.

p64, para 3.4.20, 16; 3.18 should read 3.11

p64, para 3.4.21, 14; 3.12 3.13

p11 0, para 5.5.4; The information in the last sentence is based on a

misunderstanding of earlier information, since amended by a personal
communication from Dr Myers.

p112, 113, paras 5.5.9 to 5.5.10; The actual geometry of the R Main cross­
section 14 differs from that used here, which was based on published information
corrected since the report was written. The reach is now known to be of irregular
gradient with non-uniform flow, so the hydraulic gradients used in the analysis are
not valid. The information on the R Maine in the text, figs 5, 9 and 5.10 and in
table 5.3 should be disregarded. This reach of river is no longer considered
suitable for this type of analysis.
This report describes the development of new and improved design procedures
for two-stage (compound) flood channels. This work was carried out by Peter
Ackers as consultant to HR Wallingford, with funding made available by the
Regional Water Authorites in 1988, prior to their demise when their
responsibilities in this context passed to the National Rivers Authority.
These funds were provided for the better dissemination of research results
on this subject into engineering practice.

The report is in two volumes. The first begins with a Summary and Design
Method which effectively provides a Manual for the hydraulic design of
two-stage channels. The detailed review supporting these new procedures
follows, continuing into volume 2, which also contains several Appendices.

The hydraulic engineer will find the essential information in the first
section, Summary and Design Method, but will probably wish to refer to some
of the details given in the main body of the report and in the Appendices to
extend his understanding of the complex behaviour of two-stage flood
channels.

Appendix 7 provides a design example of the computation procedures,

including tables indicating how observed stage-discharge data might be used
to extend the stage-discharge function. These tables will also provide a
cross-check for any computer programme developed to solve the recommended
hydraulic equations and logic procedures.

It is stressed that the equations given in this Manual are for the hydraulic
design of straight parallel two-stage conveyances, although information
will be found extending the application to small angles of skew (not
exceeding 10°), Information given on meandering channels in Chapter 8 of
the main text (see volume 2) shows that they behave quite differently.
Improvements in the hydraulic calculations for meandered and irregular
channels must await further work.
 HYDRAULIC DESIGN OF STRAIGHT COMPOUND CHANNELS

S UMMARY AND DESIGN METHOD

CONTENTS
Page

1. WHY COMPOUND CHANNELS NEED SPECIAL TREATMENT 1

2. INTERACTION EFFECTS 4

3. THE RESEARCH BACKGROUND FOR THIS PUBLICATION 6

4. HOW TO ASS E S S STAGE/DISCHARGE 10

5. TOLERANCES 15

6. ANCILLARY MATTERS : 16

- skew channel s
- s eparating the main channel and f lood p lain discharges
- converting river cross-sections to basic trapezoidal
g eometry
- extension o f existing stage discharge information to
greater depths
- incorporating these methods into 1 - D computational model s
- boundary shear stress
- sediment transport

7. ACKNOWLEDGEMENTS 21

8. FIGURES

9. PLATES
 DETAILED DEVELOPMENT OF DES IGN METHOD

CONTENTS

PART 1

Page

1. INTRODUCTION 22

1.1 The importance of compound channels and over-bank flows 22

1 .2 Scope of treatment: straight channels, the additional
problems posed by skew channels, meandering and curvature 23
1.3 Approach t o design: traditional calculation methods for
stage/discharge in compound channels 25

2. FLOW RESISTANCE IN CHANNELS OF COMPLEX CROSS-SECTION 28

2.1 Resume of resistance for simple open channels:

available formulae and their relevance and limitations 28
2.2 Compound cross sections: variations i n hydraulic
.parameters, field and laboratory cases: inappropriate
to treat as single cross-section (examples): the
summation method and variants in the literature;
choice of vertical division 32
2.3 Allowing for the effects of interaction: resume of
approaches in literature and their limitations: force
balance and interfacial shear: experimental studies
and typical results; more fundamental methods based on
turbulence theory and present limitations. 36
2.4 Features influencing the degree o f interaction: approach
from dimensional analysis; dependence on relative
roughnesses of flood plain and main channel: .on width
ratio; on depth ratio: on bank slope; cross-section
coherence. 44

3. HYDRAULIC DESIGN BASED ON EXPERIMENTAL ADJUSTMENT FACTORS 49

3.1 Research at Wallingford; scope of experiments and

measuremedts made. 49
3 .2 Other sources of experimental data 50
3.3 Recommended basic method: separate calculations for
channel and flood plain, th�n summed, and corrected for
interaction effects 52
3.4 y
Anal sis o f experimental results: flow regions; channel
coherence: influence of: flow depth flood plain width
channel side slope asymmetry with supporting plots and
empirical relations 56
3.5 Separation of main channel and flood plain effects 65
3.6 Influence o f flood plain roughness: form of roughness used 68
3.7 Hydraulic design formulae: formulae for interference
effects in different flow regions: choice of region;
goodness of fit to experimental data 74
 CONTENTS (CONT'd)

Page

4. SKEW CHANNE LS 82

4.1 The importance of momentum trans fer with non-aligned flow 82

4.2 Research on skew compound channel s ; increased e ffect of 83
interference due to skewnes s

4.3 Extension o f des ign method t o skew channe l s ; l imitations 86

s. OTHER SOURCES OF STAGE-DISCHARGE DATA FROM COMPOUND CHANNE L

EXPERIMENTS 88

5.1 A l lowance for width/depth r atio in general ised predictive

functions 88
5.2 Other sources of research data; scope and l imitations ;
methodology ; preliminary analysi s ; di fficulties ; data
sources and review 89
5.3 Summary of information from other l aboratory reseach 102
5.4 Skew channel s 106
5.5 Field information ; r ivers for which data were available ;
comparison of calculated s tage-dis charge above-bank with
f il ed obs ervations . 109
5.6 Conc lusions from other data sources 115

6. TURBULENCE METHODS 120

6.1 Resume of turbul ence theory as curr ently app l ied to

compound cross-sections . Internal fluid mechanics . 120
6.2 Turbulence methods and comparison with FCF data . Review
of two-dimensional methods , non-dimensiona l e ddy
viscos ity and its evaluation , comparison with experiment 121
6.3 App l ication , general ity and confirmation o f turbulence
methods . Approaches to ass e s s ing NEV , need for calibration 126
6.4 Comparison of turbulence method with emp i r ical method of
prediction . Present l imitations and potentia l 128

PART 2

7. ANCILLARY TOPICS 1

7.1 App l icat ion to more compl ex c ros s s ections; parameter 1

de finition
7.2 Shear stress ; experimenta l evidence of variation around 3
perimeter ; method for assessment of typical values
in channe l
7.3 C r i tical f low , energy and water leve l s ; Froude number 6
CONTENTS (CONT'd)

Pag e

7.4 Sources of basic information on roughne s s; 8

Lined channe ls
Natural rivers
Gravel bed channe ls
Sand bed channels
Vegetat ion

7.5 Need for and util isation of f i e l d data; extrapolation of 16

stage/discharge function
7.6 Incorporation into nume rical mod e l s: one-dimensional; 18
two-dimensiona l: separate zone p ro c edure vis-a-vis
lumped channel p rocedure

8. IRREGULAR P LAN FORM 21

8.1 Features o f meandering flows in-bank. Characteristic 21

geometry of natural channe ls , s inuo s i ty, p l anform losses,
reduct ion in conveyance therefrom, secondary currents
8.2 Above-bank flows in meander ing channel s . Review of 25
existing informat ion , reversa l o f secondary currents,
momentum and flow exchange, interference and effect
on conveyance
8.3 F low mode l s for s inuous, meandering and i rregular channe ls . 32
Hydromechanics approach, momentum approach, fine grid
mode l l ing

9. SEDIMENT TRANSPORT 37

9.1 Genera l aspects o f sediment t ransport. Modes of transport 37

9.2 Transport process and theory. D imensiona l ana lysis, 38
empirica l function s , including Ackers and White; suspended
load distribution
9.3 The influence of c ompound f low on bed material transport. 42
Typ ical river section, method o f c a l cul a tion for sand and
gravel, _lo s s o f transport capa c i ty above bank full,
performance o f equival ent s imp l e channe l, impl ications for
fluvial morphology.
9.4 Suspended sol ids· in compound channels 47

10. CONCLUDING R EMARKS 48

10.1 Summa ry of hydrau l i c design formu l a e for the conveyance of 48

straight compound channe ls; app l ic a t ion logic; tolerance on
ass essment; limitations
10.2 The advantages of compound chann e l s: environmental; SS
hydraul ic; maintenance
10.3 State of knowledge and need for further research 57
CONTENTS (CONT'd )

Page

11. ACKNOWLEDGEMENT S 60

12. REFERENCES 61

13. NOMENCLATURE 72

APPENDICES

1. Dimensiona l ana lysis applied to compound channe ls

2. Resistance functions for the SERC- F C F at Wal lingford
3. Coherence
4. Turbulence method: solution for general cross section
shapes
5. Data on channe l roughness
6. Example of channe l geometry conversion and stage discharge
computation
7. Ana lyses of other sources of laboratory data; tabular
sununaries
SUMMARY AND DES IGN METHOD

1. WHY COMPOUND CHANNELS NEED SPEC IAL TREATMENT

1.1 The term " compound channel " covers channe l cros s-sections having berms
or flood plains that come into action at high flows but which are normal ly
dry . The mechanics of flow in such two-stage channel s presents the drainage
engineer with a probl em. How is he to assess the stage discharge
relat ionship for a situation where the flow may have radically di ffering
depths and roughnesses over different parts of the cross-section? Is it
acceptable to treat the channel as if its overal l hydraul ic mean depth
(de fined as cros s-sectional area over wetted perimeter ) adequately des cr ibes
its cross-section? How should the e ffect of var iations of roughne � s over
the various f low zones be incorporated into a resistance equation? Are the
usual resistance equations such as Manning able to cover comp lex sections ,
bearing in mind that their derivations were based on simple cross section
shapes? Thes e questions have to be re solved i f the water leve l s to be
expected during floods are to be asses sed with reasonab l e accuracy and
assurance .

1.2 The usual approach to design found in hydraulic text books is out lined
in the fol lowing quotation : " . • it is necessary to split the section into
subsections . . . . Manning's formula may be applied to each in turn , and the
dis charges can be summed . The divis ion of the section into sub-sections is
a l i ttle arbitrary. S ince the shear stress across the arbitrary divisions
wil l be sma l l compared with the bed shear stres s , it may be ignored . "
(Chadwick and Mor fett , 1986). Note: a l l references appear in Vol ume 2,
sec tion 12 .

1.3 This " text book" procedure b egs several questions , not least of which
is . the assumption that the s imple addition of the calculated flows through
the separated flow zones wi l l give the correct answer . This is not so even
in the most bas ic case of a straight channel , and the discrepancy is too
great to ignore . The inter ference between the s lower moving berm flows and
the main channel flow increases head losses signi ficant ly , so that the
dis charge calcul ated by these methods wi l l s igni ficantly over-estimate the
true channe l capacity, in extreme cases by as much as the bank-ful l
dis charge . However, the basic method is attractive ly simple . What is

1
required. there fore. is an assessment o f the corrections needed to al low for
the inter-zone interference .

1.4 The di fferent velocit ies in the deep channe l and over the berms
generates strong shear and turbulence at the j unction between the zones , and
this influences the flow for a considerab l e distance either side of the bank
l ine . This turbul ence is the mechanism for extra head loss , and it must
depend on the transverse gradient o f velocity which characterises the shear
l ayer . Modern turbul ence theory is capab l e of handl ing such situations and
can provide very general solut ions . However , the present stage of
development does not yet encourage its use in normal engineering design ,
partly because of compl exity but mainly because of uncertainty accuracy as a
general prediction method , and so the method proposed here uses em� irical
adjustment factors . These are straight-forward to app ly and represent the
avail ab l e data wel l .

1.5 Figure 1 i l lustrates a compound channe l cross section with horizontal

berms and defines t erminology . Figure 2 shows the variation o f the usual
hydraulic properties of a real compound channel ; the cross-section area A ,
the wetted perimeter P and the hydraul i c mean depth R = AlP. treating the
section as one unit . (Note: nomerc l a ture wi l l be found in vol ume 2,
section 13) . The case i llustrated is a natural channe l , the River Severn
at Montford Bridge (Knight et a l . , 1989 ) . The Mont ford Bridge sect ion has
f lood p lains of unequal width with appreciab l e cross fa l l so that there is no
discontinuity in P , nor therefore in R. but even so the overal l hydraul i c
mean depth halves as the f low expands t o cover the flood p lains . The basic
form typically researched has horizontal berms and so shows discontinui ties
in P and hence in R at bank ful l stage .

1.6 There have b een many f low gaugings at Montford Bridge and , treating the
f low section as a unit , these observations may be interpreted within the
conventiona l frameworks of resistance functions . Figure 3 shows how the
calculated value of Manning ' s n varies with stage when based on ' whole
section ' analysis . As f low spreads to cover the flood p lain the n value
drops by a third , despite the knowledge that in reality the roughness of the
flood plains is not l ess than that of the main channel . This spurious
reduction in resistance arises because of the form ot the Manning equation :

n ...1
2
Thus as R calculated for the whole section falls by a factor o f 2 as the
flood plains are covered , this is reflected in a reduction in the calculated
n value , even though in no sense is there a reduction in the actual
roughness of the flow boundaries . I t i s a spurious e f fect of treating such
a complex section as a unit . Thi s not only demonstrates the inadvisability
o f t reating compound sections as a unit , it also demonstrates the confus ion
that arises if Manning's n i s used as an all embracing coefficient covering
not only the physical roughne ss of the boundary but all other influences on
head lo ss as well as corrections for i rrational methods of computation . It
is firmly recommended there for e , that for the purposes o f hydraulic design ,
Manning ' s n should be used only as a roughness coefficient related to the
physical r oughness o f the channel . Other influences on channel resi stance
should be expres sed s eparately , through appropriate adjustment factors .

3
2. INTERACTION EFFECTS

2.1 The influence of var ious flow and geometric features on the degree o f
interaction between the main channel and the fl ood pl ain i s exemp l i fied by
the ratio of the actual dis charge ( or conveyance ) to the nomina l discharge
( or conveyanc e ) , the latter being derived as the sum of the flows estimated
separate ly for the main channel and flood p l ain zones , from a knowledge o f
their geometry and roughness by using a s tandard resistance formula , e . g.
Manning . This ratio , the discharge adjustment factor D ISADF , thus a l l ows
for the interaction e f fe cts , and so to obtain a realistic es timate o f the
actua l discharge (or conveyance ) at any stage , the bas i c cal culation given
by the sum o f the nomina l zonal values has to be multipl ied by D ISADF . The
main features that a f fect this interaction and hence loss of disch�rge
capacity when f low is above bank are :

re lative depth o f flood p la in flow to main channe l flow

- roughnes s o f f lood p l ain compared with roughness of main channe l
- ratio o f flood p l ain width to main channel width
- number of f lood p l ains
- side s lope o f main channel
- aspect ratio o f main channel

In sma l l scale smooth compound channe l s , the Reyno lds Numbers on the flood
plains and in the main channel wou ld have to be added to this l i st , but in
a lmo s t a l l practical circums tances viscous e f fects are not important enough
to require consideration o f Reyno lds Numb ers .

2.2 The depth o f flow-on the f lood p lains relative to that in the main
channel is a major factor . As soon as the flood p lains become inundated ,
the flow in the main channel suffers from the interference o f the s lower
flood plain f low . With s imilar roughnes ses in the main channe l and on the
flood p lains , this influence increases to a maximum at a relative depth,
H* ' o f the order of perhaps 0.1 to 0.3. (H * =depth on flood p lains/depth
of f l ow in main channe l ) . The max imum reduction in flow ( re ferred to as the
di s charge de ficit) due to this interaction may be anywhere in the range 10 -
20% depending on other factor s ( D I SADF approx 0�8 to 0.9). As the re lative
depth increases further , the loss of capacity due to interaction diminishes

4
again because there i s l ikel y to be l e s s dif ference between main channel and
f l ood p lain velocities , but in practice the interference effect does not
become neg l i gibl e unl e s s the berms are relatively narrow or the relative
flow depth becomes considerab l e . The ratio of flood p lain width to main
channel width is an important factor , wide f l ood p lains tending to show
wor s e interaction e f fects than narrow one s .

2.3 The difference between the average vel o cities on the flood p l ain and in
the main channel inf luences the degree of interference . Thus one woul d
expect extra f l ood p lain roughness t o exacerbate the e ffect and the unlikely
combination of relatively smooth berms with a somewhat rougher deep section
to diminish it . Bank s l ope is a l s o a s ignif icant factor as with vertical
banks i . e . a rectangul a r main channel , there i s c lo s es t proximity .of the
fast channel f l ow to the slower f l ood p l ain flow . Gentle side s l opes to the
main channel on the other hand provide a transition zone that might l imit
the interaction effe cts .

2.4 The l ikely influence o f the interaction between main channel and f lood
p lain flows c learly depends on how comparabl e the hydraul i c conditions in
these zones might be : if vel ocities and depths are very s imi l ar , then we
can expect interaction e f fects to be sma l l ; i f they are very dissimil ar ,
then maj or e ffects are t o b e expected . The degree t o which the different
zones exhibit f l ow s imil arity is a rather new concept , the section ' s
" coherenc e " , and this may be defined theoretical ly . The c loser the
coherence i s to unity , the more l ikely i s the hydraul i c s o f the section to
app roach that of a non-compound channel . C oherence i s defined within the
main text , see section 2.4 therein , and expl ained more ful ly in Appendix 3 .
In e ffect , i t i s the ratio o f the flow c a l cul ated for the whol e section
(with zonal friction factors weighted according to the respective wetted
perimeters ) to the sum of the s eparate ly c a l cu lated zonal f lows (before any
allowance for inter ference ) .

2.5 This parameter varies with flow depth in a given channel , and the
function for the Mont ford Bridge natural river section is shown in fig 4.
Channel coherence b rings together in one parameter mos t o f the factors
expected to influence the hydraul i c s of compound channe l s , and thus provides
an important c lue to the interaction e f fect .

5
3. THE RESEARCH BACKGROUND FOR THI S PUBLICATION

3 .1 The Flood Channe l Fac i li ty ( FC F ) at Wal l ingford was jo int ly funded by

the Science and Engineering Research Council and Hydraulics Research Ltd , at
whose premi ses the equipment was bui l t . The program o f investigation was
organised as a series of individual but c losely co -ordinated proj ects by
University groups , and has been instrumental in providing an unique set of
large scale data , which are both accurate and comprehensive . The faci lities
themse lves have been ful ly described e lsewhere ( see Knight and Sel l in ,
198 7 ) . This SERC - FCF i s i l lustrated in P lates l to 4 . Other informat ion on
the performance of compound channel s was a lso assembled and utilised where
pos sib l e to extend the coverage of the prediction procedures to a wider
range of geometries . However , s everal of the other sources o f labqratory
research data contained uncertainties or inaccuracies , and many deal t with
geometries far removed from those of practical interes t . These other data
sets were general ly of less value than had been hoped , though some proved
very useful in confirming the methods to be described and further developing
them to cover a wider range of circums t ances . Particularly important in
that respect are the col lections of field ob servations that were used to
validate the recommended procedures .

3.2 The basis of analysis of a l l the experimental data was through a

comparison of the measured discharge with the nominal tota l discharge , as
calculated from zones s eparated by vertical divis ions . Alternative
parameters were cons idered to estab l ish their relevance and s ignificance ,
and those recommended here were found to best represent the interference
e ffect . The mos t r e levant parameters to represent changing depths on the
flood p lain are :

- r e lative depth , ( H-h ) /H = H*

- channel coherence , COH

(H is the total flow depth ; h is the depth of the main channel ) . The
discrepancy between the bas ic calculation ( i . e . the sum o f the separately
computed flows in main channel and on flood p l ain b e fore a l lowance for
interaction ) and the measured f low was treated in several dif ferent ways , of
which the fo l lowing proved mos t usefu l :

6
 - adj ustment factor , measured discharge/basic calculation , DISADF
- discharge de ficit as proportion of bank ful l flow , DIS DEFBF
- discharge de ficit normalised by the calculated velocity difference
and the product of tota l flow depth and main channe l depth ,

Q* 2 = ( QCALC - QMEAS / (VC -VF ) H h

3.3 The degree of interference between the channel flow and main channel
flow shows different trends as the f low depth varies . Figure 5 shows the
observed s tage- discharge results for a particular geometry , with B/b = 4 . 20 .
This figure shows the discharge adjus tment factor , DI SADF , i . e . the factor
by which the sum of the calculated zona l flows has to be multipl ied to agree
with the observed discharge , p lotted against relative depth . The flow
passes through three distinct regions of behaviour , each of which �equires a
different function to represent the trends :

- Region 1 is at relatively shallow depths where the inter ference

e f fects progres sively increase with depth , up to relative depth 0 . 2 , when
the " lo s s " o f capacity is over 1 0 percent .

- Region 2 covers depth ratios from 0 . 2 to 0 . 4 for this particular

geometry , with the inter ference e f fect diminishing towards a discharge
loss of about 4 percent .

Region 3 occurs with further increa se in depth , which causes an

increase in the inter ference e f fect again .

- Also shown on Figure 5 is the theoretical coherence for that sample

geometry , COH , and it wil l be seen that DI SADF always l ie s between COH and
unity . The imp lication of thi s is that the channe l conveyance always
exceeds the " s ingl e channe l " computation but is less than the sum of the
zonal computations . Had these part i cular experiments been continued to
greater depths , they woul d mo s t probably have followed the COH funct ion ,
ie a sing le channe l computation becomes appropr iate at considerab le depths
of flood berm inundation . This forms Region 4. Note that
Region 4 does not imp ly that there i s no interaction between the main
channel and the flood pl ains: the main channel dis charge continues to be
a ffected by the presence of flood p l ains .

7
3.4 The depth l imitat ions between the regions shown in the samp l e p lot o f
Figure 5 a r e not general ; they depend o n var ious parameters , and dif fer
considerably with rougher flood plains: nor can it be as sumed that i f
(H-h ) /h > 0 . 5 , the interference e ffects are entirely negl igib l e . What i s
c l ear , however , i s that different f l ow regions exist , and consequent ly
different design formulae are required for each zone , as wel l as the means

for estab l ish ing which region a particular design case wil l l ie within .
Predicting the stage discharge curve i s therefore a rather compl icated
procedure , though ea s ily handled by means of a modest computer program .

3.5 A detailed exposition o f the analysi s o f the different groups o f test

results is given in the main report , see Chapter 3 . Thi s summ ary provides
the overall picture and provides the design method deduced from th�se
analyses . Empirical formulae were obtained for each region of flow , and
progress ively deve loped into general functions covering all the geometries
tested . These were then compared with other data in the research
l iterature , to further deve l op the method to cover main channe l width/depth
ratios other than the s ingl e value of 10 covered in the research in the FCF .
This neces s itated the introduction o f an a l l owance in the predictive
equations for width/depth rati o , the aspect ratio factor ARF . Th is re fers
particul arly to r eg ion 1 , there being no evidence that aspect ratio
influences the predicted f l ows in other regions. When the width to depth
ratio for the main channel is ten , ARF = 1 , but more typical ly ARF = 2b/ 1 0h ,
where b i s the s emi bed width . When the main channel width/depth ratio
exceeds 20 , it may be considered to be wide , with ARF = 2 for all greater
aspect ratios .

3.6 Detail ed information was also obtained on ve locity distributions , and

th i s provided a basis for assessing the dis charges within the main channel
and over f lood p la in zones s epar ately . Figure 6 shows the discharge
de � icits , ie the di f ferences between nominal calcul ated disch arges and those
actua l ly occurr ing , norma l ised by bank ful l f l ow , and th is is typical of the
information about the separate influences of flow interaction on the two
zone s . Thi s shows that the bulk o f the discharge de ficit compared with the
basi c calcul ated va lue arises because of inter ference e f fects in the main
channel: the flood p l a ins contribute a relatively sma l l discharge addition .

8
3.7 In c ertain o f the tests at Wal lingford the f l ood p l a ins were roughened
by surface piercing rods ( see Plate 3 ) . P r e l iminary tests were made to
determine the basic friction formula for this form of roughness under
non-compound conditions . The basi c friction law developed adds the drag o f
the rods t o the friction arising from the smooth cement mortar finish o f the
solid channe l sur face . al lowing also for the b lockage e ffect of the rods .
The rod r oughness provided much higher friction factors on the flood p l a in
than in the unroughened main channel . Thus the series o f tests with
added flood pain roughnes s provided a radi ca l ly different case from those
with main channel and flood p lains o f equal roughnes s . The methods proposed
here cover the extremes of roughnes s ratio satisfactorily . so are expected
to cover any intermediate roughnes s condition .

3.8 To illustrate the e f fect o f extr a rough flood p lains . r esults for
c omparabl e tests are p l otted in Figure 7 as discharge adjustment factor
against r e lative depth . The four regions for test 02 are indicated but ,
bearing in mind that region 1 i s the zone o f increasing inter ference with
depth , there is no evidence that the tests with very rough flood plains ever
entered region 2 . The results show progre s sively increasing inter ference
e ffects up to the maximum depths covered , reaching the very severe condition
approaching 40% loss o f channel capacity ( in excess o f bank ful l discharge
in fact ) when the depth on the f lood plains equal s the depth of the main
channel . Incidental l y , i f the f lood p lains are much rougher than the main
channe l , the section coherence does not app roach unity as the depth
increases . For'the FCF tests with rough flood p lains it remained around
0.4.

9
4. HOW TO AS SESS STAGE/DISCHARGE

4. 1 The calculat ion for any depth in the range o f stages of interest begins
with the "standard" hydraulic computations for the main channel and flood
plains separately . using the preferred resistance equat ion with appropriate
roughness coefficients based on the known sur face conditions . No pre ference
for any particular basic resistance l aw is imp lied in what fo llows . In many
fie l d situations the Manning equation woul d be considered most appropriate .
though in artificial channels the Co l ebrook-White equation may be preferred .
The first step is to divide the channel section into its component parts .
with vertical divisions between main channe l and f lood plains . and to work
out their separate parameters : area . wetted perime ter . and hydraulic mean
depth . Only the so l id perimeter is inc luded : the vertical interf�ce is not
inc luded in the wet ted perimeter . The hydraulic gradient has also to be
known . o f course . These "standard" calculations provide the basic discharge
for the given depth . the sum of main channe l and f lood plain flows . This
has to be adjusted for interference effe cts . to obtain the true predicted
flow . uti l ising the equations summarised below .

4.2 Because there are four possible regions o f flow. in e ffect four sets of
computatations are required . to assess the discharge as i f it were in each
flow region in turn . There is a logical procedure then for selecting which
region is in fact app l icable at each of the depths considered . The
methodo logy is thus rather comp l ex . though readi ly programmed for computer
solution . Because of this unavoidable compl exity . and to avoid imposing on
a l l the users of the new methods the need to develop software , in due course
it is hoped that a PC disc wil l be issued which enabl es the user to go
directly to a so lution . Unti l such general user-friendly software is
developed . the hydraulic engineer is provided with the appropriate
e quations . for whi ch he/she may prepare an appropriate program for solution.
Something l ike 5 0 0 - 600 program instructions are required for a
comprehensive appl ications package . The various parameters used in the
solution are de fined on first appearance . but they are a lso listed in
Chapter 13 of the main text .

Region 1.

4.3 This is the region of relatively sha l low depths where inter ference
effe cts increase progressively with depth . This is best represented by Q*2'
10
the discharge deficit norma lised by the ve locity di fferential and the
product of flow depth and main channel depth . The relevant equations are :

Q* C = 1 240 + 0 . 395 B/w + G H * 3

2 C
- .

where :

Flood p l ain discharge de ficit Q* (V -V ) Hh ( ARF) 4

2F C F
=

and :

Main channel discharge de ficit = Q* (VC -V ) Hh (ARF ) 5

2 F

For s e � 1 . 0 : 6

For s e < 1.0: 7

There is a narrow range o f conditions for which Q* C as calculated above

2
might be negative , implying that interaction e ffects woul d increase
discharge . This is not ever l ikely in practice of course , and so to retain
some minimum interaction effect , with shal l ow flood p lain depths or with
partial inundation of s l oping flood p lains , it is suggested that a minimum
value o f Q* C might be , say 0 . 5 , and Q* shoul d then be set to zero . This
2 F
wil l have the e f fect o f generating a step , not exceeding 5% o f bank- ful l
f low in the stage discharge function at bank- ful l e levation .

In the above :

H* ( H-h) /H
fc friction factor c a l culated for main channel , 8gR S/V 2
C C
f friction factor c a l culated for f lood p lains , 8gR S /V 2
F F F
B semi-width of section including flood p lain ( s ) a t elevation o f
flood p lain ( s ) ; or water sur face width i f partal ly inundated
we semi top width o f main channe l at elevation o f flood plain ( s)
se side s lope o f main channel

Then :
11
 D ISDEF 8

Q . - D I S DEF 9
b aS1C

where :

N = number of flood plains ( 1 or 2 )

F
V = calcul ated basic velocity in main channel
C
V calculated basic veloc ity on flood plains
F
Q . = the sum o f the zonal basic discharge calcu lations
b as 1c
Q = required flow prediction for region 1
R1
ARF = aspect ratio factor , typica l ly 2b/ 1 0h

Region 2.

4.4 This i s the zone o f greater depth where the interference effect
diminishes again . The most g eneral function in this region expresses the
requisite discharge adjustment in terms of the channel coherence , COH , and
the relat ive depth H* . It expresses the observation, in Figure 5 for
example , that the graph of D I S ADF in Region 2 runs paral lel to but below the
graph o f COH . It fol lo ws that the adjustment to discharge is given by the
coherence calcu lated for a greater relative depth that the actual value .
This i s an empirical ob servation, not a theoretical deduction . Hence :

D I SADF (H* , channel geometry and roughness )

2
= COH ( [ H*+shi ft ] , channel geometry and roughness) 10

where for s e �1 . 0 , shift = 0 . 0 5 + 0 . 0 5 N 11

F

for �c < 1 . 0 , shif t = -0 . 0 1 + 0 . 0 5 N + 0 . 6 se 12

F

The basic de finit ion o f COH may b e expres sed i n terms o f the g eometric
ratios o f the compound channel : let A* =N F AF /A ; P*=N P F/P c ; f*= f F/ f c .
C F

Then :

( 1 + A* ) / [ ( 1 + A* ) / ( 1 + f*P *) ]
COH = 13
1 + A* / (A* / f* P * )

12
So to work out D ISADF in region 2 , the values o f A* , P * and f* inserted in
the above relate not to the actual relative depth , H * , but to the " shifted"
value , H* + shi ft . Note that the corresponding " shiftedu depth , H ' , used to
calculate COH i s given by:

H ' = Hh/ ( h - shift H ) 14

Then :

15

Region 3.

4.5 This is a relatively narrow region o f flow , best represented by D ISADF

as a function of COH , calculated in this case for the actual relative
depth .

D ISADF = 1 . 56 7 - 0 . 667 COH 16

3

17

Region 4.

4.6 This is the region where the coherence o f the cross-section is such
that i t may be treated a� a s ingle section , with perimeter weighting o f
friction factors , when calculating overal l flow . This does not , however ,
mean that the separate zonal f l ows so calculated provide accurate
assessments o f the flows in those zones . For total flow computation
however :

D ISADF COH 18
4

19

Choice o f region.

4.7 The logic behind the selection o f the appropriate predictive equation
is dependent upon the calculation of discharge for all regions in turn ,

13
referred to above as Q and Q respectively. The choice of the
R1• Q RZ' Q R3 R4
appropriate region and hence appropriate total discharge proceeds as
follows:

21

If Q and Q then Q QR3 22

R1 < QR2 R3 < QR2
=

unless Q > Q when Q = Q

R4 R3 R4

14
5. TOLERANCES

5.1 The performance o f thi s set o f predictive equations was first checked
by reference back to the experimental data obtained from the FCF by the
various research teams working on different aspects o f the research . The
percentage discrepancies between the individual results and the predicted
discharges for the observed depths , geometries etc were assessed , and
analysed statistical ly , to obtain mean errors and the s tandard error o f
estimate . Table 3 . 2 o f the detailed text summarises these results , with
statistics for groups of experiments as wel l as for the total set , including
those with roughened flood p lains . Broadly speaking , over the whol e data
set there was no residual mean error , and the standard error of the estimate
( the variabil ity) was under 1%. As a set o f predictive equations �hey
represent the actual flows to high accuracy , but o f course the tolerance in
any app lication in practice involves other tol erances as wel l as any errors
in the predictive functions themselve s . Thes e include :

discrepancies ar �s �ng because o f interpolations between , and

extrapolations beyond , the conditions tested
- knowl edge o f , and vari ability in , the geometry of the section
- the simpli fication of the actual geometry to suit the method
- the basic friction law used in the calculations
- the accuracy o f the friction coefficients used
the hydraulic gradient and the as sumption of steady uni form conditions

The computational accuracy may wel l be within 2 or 3 percent , as

demons trated by some of the better qual ity field data examined in Chap ter 5 ,
but there are probably larger tolerances in discharge from other
uncertainties .

15
6. ANCI LLARY MATTERS

Skew channels

6.1 The stage/discharge functions obtained for flood plains aligned with

the deep channel outlined above were compared with results for channels

skewed by up to go from the valley floor (see Plate 3). There was a limited

set of information for this condition in the SERC-FCF (Elliott and Sellin,

199 0), but as might be expected they indicated rather greater interference

effects. See Section 4 of the main report. The predictive functions for

straight channels may be applied to angles of skew (�) up to 10° provided

this extra interference is allowed for. The necessary adjustment is

obtained through:

. DI SDEFSKEW D ISD EF AL GNE x (1.03 + 0.07 4�) 23

I D
=

where � is in degrees. This will be conservative, i.e. an under-estimation

of conveyance, if the flood plains are much rougher than the main channel,

when it appears that a modest skewness has little additional effect.

Separating the main channel and flood plain discharges

6.2 Within Region 1, the predictive method provides these discharges, in

that

24

and

25

where Q is the discharge for each flood plain. Knowing these separate
RlF
discharges, the mean velocities in these zones can be calculated of course.

6.3. The computations of total discharge for the other flow regions is

obtained by an overall adjustment for interference effects to the total

basic flow: the separate adjustments for main channel and flood plain are

not assessed in this procedure. However, in terms of the accuracy required

for engineering purposes, the addition to the flood plain flow is much
16
smaller than the reduction in the main channel flow , and so in Regions 2 , 3
and 4 i t would be reasonable to ignore the adjustment to the flood plain
discharge , allocating it all to the main channel . Hence :

Q Q - DI SDEF 26
R2 , 3 , 4C = Cbasic

where D ISDEF i s the predicted overall discharge deficit , and

Q Q 27
R2 , 3 , 4F = Fbasic

An alternative procedure is suggested in Chapter 3 where calculations are

proceeding through increasing depths to establish t he stage discharge
function . In that case , there is evidence that the discharge adjustment for
the main channel , D ISADF , does not change much beyond the limit o f
C
Reg�on 1, a t least up to relative depth , H * , = 0 . 50 . Hence for higher flows
in Reg ions 2 , 3 and 4 , the value of D I SADF at the limit o f Region 1 could
C
be retained .

Converting river cross-sections to basic trapezoidal compound

geometry

6.4 When available , the data handling routines on the PC software disc will
include an option for doing this on screen , either for a single s ection or
for the average of several sections defining a reach o f river . The
essential element s o f the proces s are as follows : they are illustrated in
Figure 7 . 1 in Volume 2 and an example is g iven in Appendix 6 .
- average the river bank elevations for the two sides ( unles s there is
a flood plain on only one side)
- define the bank line s , which form the vertical divisions between main
channel and flood plains
choose a realistic bank slope , say by averaging the surveyed slope
over the upper two thirds o f the bank height : this gives s e
- work out an average bed level that g ives the same cross-sectional
area below bank level: this de fines h
- identi fy the back of each flood plain and so obtain the total width
across the valley floor ( 2B )
the various predictive functions can b e applied even i f the flood
plains are not horizontal , as the calculat ion of Q . can be
Fb aS1C
carried out for any g iven flood plain profile and flow depth .
17
Extension of existing stage discharge information to greater depths

6.5 The methods of computation given above can be applied in the reverse
direction , in that for any kno wn point on the stage discharge curve . the
predictive procedures can be applied for a range o f assumed flood plain
roughness coefficients , in order to select the one which mos t closely
represents the observed dis charge . To do this sati s factorily , it wil l first
be necessary to have a reliable value for the main channel roughness
coefficient . This is b e s t obta ined by analysing the upper range o f within
bank flows , checking whether there is any apparent trend in coefficient
values with depth that might continue to the above bank condition . The
functions are too complex to apply directly in the reverse direction with
f lood p lain roughness unkno wn , though c learly a computer program c �n be
written to do just that by iteration . Information on typical roughness
coef ficients wi l l be found in Appendix s. with discuss ion in Section 7 . 4 of
Vo lume 2 .

6.6. Thus the existing s tage dis charge information can be used to obtain
realistic roughness coef ficients , separated from any extra head loss due to
interference and so providing a more rel iable basis of extending the stage
discharge curve than has existed hereto fore . It is explained in Chapter 7
ho w gross errors may have occurred in the traditional approach to assessing
field data for t wo stage flow: frequently the flood plain roughness has been
adjusted to match observed discharges , whereas in fact it is the main
channel flow that suffers from reduction due to interference effe ct s . An
exampl e o f extending the stage discharge function is given in Appendix 6 ,
Volume 2 .

Incorporating these new methods into 1-D computational models

6.7 1-D computationa l mode ls require geometric information at the many

cross-sections used to define the hydraulic system , as wel l as a method of
assess ing hydraulic resistance . Some mode ls may use the cross-section data
to define a unitary channe l : this is not recommended be cause by so doing the
roughness coe fficient is a lso required to take account o f spurious changes
due to the geometric anomal ies introduced by flow over the flood p lains , as
wel l as real changes in roughness with stage as the flood plains are
inundated , and the ext ra resis tance due to interference e ffects. However ,

18
i f the model requires the sections to be treated as units , not divided into
main channel and flood plain zones , the predictive method given above cou l d
b e used a s a roughness/cross-section pre-processor , t o deduce overal l
equivalent resistance coeffic ients and/or conveyances as functions o f flow
depth .

6.8 Other model s wil l use cros s-section information in its more rational
form , with separate data for flood p lain and main channel . In this case
also it woul d seem appropriate to use the predictive methods given here in
the form of a pre-processor to provide the conveyance/depth function at each
section in the model . Conveyance , K, is usually defined by :

K = QltS 28

where S is the hydraulic gradient , and can calculated using the predictive
equations over the required range of depths .

Boundary shear stress

6.9 Some information on boundary shear stress is given in Section 7 . 2 o f

Volume 2 . I n ef fect , the higher velocities in the main channel spil l over
on to the flood p lain and so give increased shear stress close to the
channel bank . On the other hand , the interfacial shear stress tends to
l imit the main channel flow and so reduces the boundary stress compared to
that which woul d arise in the absence of interference e f fects . Local
increases on the flood p lain can be as much as by a factor of 5 relative to
the value that would be calculated from the local flow depth and channel
gradient .

6 . 10 The average shear stre s s on the main channe l bed is approximated by

re�ucing the basic value , c pgHS or pgR S by the factor D I SADF 2 ' on the
0 C C
basi s that with a s quare l aw resis tance function such as the Manning
equation for rough surfaces , the boundary shear wil l be proportional to
vel ocity squared . Experimental data suggests that the mean bed shear stress
wil l be somewhat above that based on hydrau l ic mean depth R ' but closer to
C
that than to the depth based value .

19
Sediment transport

6. 11 Later phases o f the research programme in the FCF at Wal l ingford

envisage studies of the transport of both bed material and suspended
sediment . In theory , the main channel 's transport capacity for bed material
must be reduced by the interference from flood p l ain f lows compared with the
transport capacity i f there were no interference . This fol lows from the
fact that the channel discharge , vel ocity and b oundary shear stress are all
reduced when f low is overbank , compared with the values that woul d otherwise
occur at that depth . S ome samp l e calculat ions o f total bed
material transport given in Chapter 8 of Volume 2 show that the reduction
might typica l ly be by a factor of 2 or 3 . Another signi ficant feature i s
that when the total transport o f bed material is expressed as a transport
concentration , this is appreciably less with overbank flow than when the
channel was running within banks . Clear ly , any consideration of channel
regime ( including the s imulation o f overbank flows in morphological mode l s )
mus t take account o f the interference e ffect . Thes e are tentative findings
as they are based on the trans fer of sediment transport functions for s imp l e
channe ls t o compound channels : further research is needed to confirm o r
amend this assumption .

6. 12. Suspended sol id transport may spread through the ful l flow depth , and
certainly the finest material in suspension , the wash load , wil l be found in
the near-surface layers . The interfacial shear not only transports momentum
across the bank l ine , from main channel to flood p lain , it a l so transports
sediment due to lateral turbul ent diffusion . So sediment kept in suspension
within the deep water channe l can d i f fuse s ideways into the s lower moving
flows on the flood p lains , wher e it might settle out . Al though the lateral
diffusion process has been studied in the SERC-FCF using dye tracers ,
comparabl e research using sediment is in a future programme , so no
qu�ntitative information is availab l e for use now . However , the process o f
lateral diffusion and settl ement wil l b e recognis ed as that which generates
the leve ls of maj or lowland rivers .

20
7. ACKNOWLEDGEMENTS

This analysis o f compound channels would not have been possible without the
support o f funding from s everal Water Authorities . whose river duties have
now been taken over by the National Rivers Authority . I t has depended on
support from staff o f HR Wal lingford . as wel l as detai led l iaison with the
SERC Proj ect Co-ordinator , Dr D W Knight , o f the University o f Birmingham .
The careful and detailed work o f the UK research groups invo lved deserves
special ment ion . Without their research and the availabi lity o f their high
calibre results , the work described here could not have proceeded to a
successful conc lus ion . The co-operation o f many other research groups and
·
individua ls i s also ackno wledged ( see Chapter 5 and the ackno wledgements
( Chapter 1 1) of the main report ) .

21
Figures
 I· :!:
B
W
C

}� :
' /

� A1
I �: 1 H SF

I. .I r r
HF
b
'I
bw bwc
F

Fig 1 Compou nd channel cross-section: geometry and termi nology

1 2 4 5 6
1� 1 I
I I
1m

-4.0
s H
5
5
5
1
2
-63.0
=

-1.33
=
=
h Gm
3
s =
4
s = +1.50
+23.0
00
� 5
x vertical

+7.1
I
s
5 = exaggeration
6 B
63m •1•8•1• �·� •l•g •I• 23m ::
5 =
7

I·

62 0 100 200 300 400 500 600

2
Area m

60 20 40 60 801001201401�0 P�'
Perimeter (m)

A----
-
-
R
-

58
I

57.282
-
, .....
_,- 'Q:.
z
56.282
- "
...
0 /...... ...
56 'Q:.
.....
0 ... - '1'
.s , /
Gi : I
> tl
CD 54
_.
'
I,'
52 I,
I

50 0
II
'Q:. 50.282
2 3 4 5
Hydraulic radius (m)

JBW/1/10-91130

Fig 2 Chan nel at Montford Bridge on River Severn ; (i) cross-section

(ii) hydraulic parameters ; cross-sectional area, A, wetted
peri meter, P, hydraul ics mean depth, R. (Knight et al, 1 989)
 61 H (m)
(H-h)IH
0.4 10
60
9
59 0.3
8
58
0
0.2
6 .JL57.282
E
7
57 0.1
.JL56.282
-

"ii Bankfull 0 6
56
j 5
55
4
54

3
53
0 0.01 0.02 0.03 0.04
Manning's n

Fig 3 River Severn at Montford Bridge: variation of Manning's n with

stage when considered as single section ; based on field
measu rements (Knight et al, 1 989)

.£:
:c
:t 0.5

0
0.3 0.5
y
Conve ance;,113t;ingle channel
1.0
Sum o component conveyances
COH =

JBW/2f1 0-91130

Fig 4 River Severn at Montford Bridge: variation of channel coherence

-COH , with relative depth
 /
0.6 1-

/ Region 4
1- / ··· -'
�
•
/ •

G� •

•
Region 3

0.4 1-
� •
•
_.....,... •

� •
•
Region 2
1- •
•
•
•
•
0.2 1- ••
•
•

•
• Region 1
•
- •
•
•
•
•

0 I I I I I I I I I I I I I I

0.85 0.90 0.95 1 .0

DISADF

Fig 5 Sample test results from SERC-FCF: plot of DISADF (ratio of

measured discharge to sum of zonal calculated discharges);.
also coherence - COH, to same scales; test 02, averages of 3

0.50

0.45

0.40 X
A • 0 6 +
�(,
0.35
.<§ (\,
q,·
\ (\,
0.30 X \6.b-•+�
� (/;)•
:X:
;::::,. 0.25
:X:
' .
£ 0.20

0.15

0.10

0.05

0.00 .___....___.___.._._-L----'---'--'

-0.20-0.15-0.10-0.05 0.0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55
DISDEFBF
Individual +me Test- 01 t:.. me
Test- 02
o me Test - 03
fp f f
data x A p • p
JBWr.3/10-9113D

Fig 6 SERC-FCF results for three widths of flood plain: discharge deficits
for flood plains (-ve) and main channel (+ve) as proportions of bank
full flow. (Lines drawn apply only to region 1)
 + A
CO 0
0 .50
+ .f
0
0.45 - + A
....
0
0.40
"'* §
A 0
0
0.35 -

0
0.30 - iJ.r CD
+
I +
-
+
f.
0.25
++ cP
e. 0.20 -
CQ:>
+
0
0
1- ++
+
0.1 5

+. A Rough ea
1- +0+
0 Smooth 0
0.10

cP
0.05 1-

I I I I I I I I I t
0.00
-0.50 -0.55 -0.60

DISADF
0.65 0 .70 0.75 0 .80 0.85 0 .90 0.95 1 .00

Individual
data + Test � 7a A Test � 7b 0 Test · 02

JSWf4/10-01/30

Fig 7 SERC-FCF resu lts : comparison of roughened flood plains and

smooth flood plains in terms of d ischarge adjustment factor,
DISADF
Plates
Plate 1 General view of SERC-FCF at H R Wal li ngfo rd. Straight channel
with flood-plain width restricted by movable wall , seen from
downstream

Plate 2 Flood-p lains roughened by pattern of vertical rods supported

from t1mber frame
Plate 3 layout of flume for s kew channel experi ments. Note that the
flood-p lain limits form the angle of skew

Plate 4 Experiment i n p rog ress with meandered channel :

60° cross-over chan nel
PART I
 DETAILED DEVELOPMENT OF DES IGN METHOD

1. INTRODUCTION

1.1 The importance o f compound channel s and over-bank f l ows .

1.1.1 The term "compound channel " covers channel cross-sections having
berms or f lood p lains that come into action at high flows but which are
normal ly dry . The basic form of compound cross-section is a central deep
channel with symme trical side berms ( or f lood p l ains ) which themselves have
a horizontal bed . Man-made drainage channe l s may come close to that
idealised form , but in hydraul i c engineering practice compound channel s wi l l
not in general have such a simp l e cross section : they may be asyrnrn�tric ,
have a less regular deep channel section , unequal berm widths with
cross-fa l l s etc . C learly natural rivers di ffer considerably from the
ideal ised cross-section , and also have the added complexity of p l an
irregu l arity : a lthough relatively straight reaches may occur , natural rivers
contain many changes of direction o ften with irregular meanders , with flood
p lains of variable width . Compound channe l s thus take many forms , some
basica l ly s imp l e but many being o f considerable complexity .

1. 1.2 The more complex forms o f compound channel are a l s o favoured in terms
of environmenta l management . Schemes including such man-made channe ls are
closer to natural ly occuring systems , and they are increasingly favoured
over s imple artificial drainage channel s . Their berms wil l provide suitable
habitats for water-s ide vegetation and the wi ld- l i fe it wil l support and
shel ter ( Hydrau l ics Research , 1 9 88 ) . Also , the deep channe l within a
compound section is more l ikely to be sel f-maintaining from the sediment
point o f view than a sing l e wider channel with the same f lood capacity .

1. 1.3 The mechanics o f f l ow in such two-stage channel s presents the

drainage engineer with a problem . How is he to assess the stage discharge
re lationship for a s ituation where the f l ow may have radical ly differing
depths and roughnesses over different parts of the cross-section? Is it
acceptab l e to treat the channel as if its overal l hydraul ic mean depth

22
( defined as cros s-sectional area over wetted perimeter ) adequately describes
its cross-section? How should the e f fect of variations of roughness over
the various flow zones be incorporated into a resistance equation? Are the
usua l resistance equations such as Manning abl e to cover comp lex sections .
bear ing in mind that their derivations were based on s imp l e cross s ection
shap e s ? These questions have to be resolved i f the water leve l s t o b e
expected during floods are to b e assessed with reasonab le accuracy and
assurance .

1.1.4 The prob l em of representing the flow resistance o f complex , yet

commonp l ac e , channel s applies equal ly to computational river mode l s . In the
interests of economy , one dimensional , lumped cross-section mode ls have
t raditionally been used , with their inherent s imp li fications o f the f l ow ,
for conditions which may include not only compound channel cross-sections
but . a l so exchanges of f low between the deep main r iver channel and the
f lood-plain . I f even the basic case o f a straight prismatic compound
channel is not wel l understood , it is unlikely that numerical model s with
the ir many other s imp l ifications of the geometry and flow wil l be abl e to
s imu late accurately the hydraulics of real river systems .

1.1.5 Natural r ivers overflow their banks during periods of high discharge ,
causing potential damage to life and property . Those responsib le for f l ood
protection expend a considerab le proportion o f their budgets on schemes to
l imit the frequency . extent and impact o f floods . by the provis ion o f flood
embankments , channe l improvements and warning systems . They therefore
require reliab l e methods for predicting river l eve l s , and an essential
element of that i s a reliab le method for assessing the capacity of the
drainage system . In al l probabi l i ty the system include s compound channels
for which conventiona l methods o f hydraulic assessment are inadequate . The
main obj ect o f this publ ication i s to up-date those conventional methods to
incorporate the results of recent research .

1.2 Scope o f treatment

Straight channels .

1.2. 1 Although most natural channe ls are curvilinear over the bulk o f their ·

lengths , reasonably straight sections do occur . and are the preferred

23
reaches for hydrologic measurement . In those situations , a reliable method
for extrapo lating beyond the observed range of discharge , based on s ound
physical principles o f hydraul ic performance , is required to cover extreme
events with appreciable f lood plain flow . S traight drainage channe ls o f
compound sect ion are also used a s river improvements , and in urban
s ituations where berms may have cons iderable amenity value . As we l l as
providing the basic configuration on which much research has been conducted ,
straight channels are thus o f very real importance . An understanding of
the ir hydraulics is a necessary foundation for understanding the more
complex cases .

Skew channels .

1.2.2 The term " skew channel " refers to the s ituation where the deep
channel and the val l ey floor are not aligned with each other .- This means
that as one flood plain contracts and the other expands , flow is forced
across the deep channe l , a process which one woul d anticipate introduces
radical ly different f l ow patterns in the main channel . This is a common
situation for natural rivers during floods : there is interchange o f f l ow
between river and flood p lain , and the skew channel provides a basic case
for des cribing the effe cts on overall resistance and river stage of this
f l ow exchange proces s .

Meandering and curva t ure .

1. 2.3 A channe l cros s ing the val l ey floor at an angle must do so over a
l imi ted l ength , related to the angle o f skew and combined width o f the flood
plains . It fol l ows that in natural rivers skewnes s s i s c l os e ly associated
with curvature and meandering . The hydraulics o f the skewed channe l is thus
a pointer to the hydraulics o f meandered channels , wher e there i s a sequence
o f f l ow exchanges from right flood plain to left and vice versa . This
.
process , together with the influence on resis tance of the intervening bends ,
provides a logical progression to the true complexity o f many r iver
systems .

1.2.4 This report follows this progress ion from straight compound channels ,
through skew channels to meandering and irregular rivers , though
concentrating on stra ight channel s . Supplementary information on rivers o f

24
complex plan form will follow as a result of later work . Particular use is
made of research carried out with support from SERC ( the Science and
Engineering Research Council) , HR , Walling ford , DOE (Department of the
Environment) , MAFF ( Ministry of Agriculture , Fisheries and Food) and several
o f the water authorites ( now replaced in terms of responsibility for rivers
by the National River Authority) , It s ets out to explain why and in what
way the des ign procedures applicable to s imple channels require modification
for these other , yet commonplace , s ituations . This leads to recommended
des i gn methods that may be used for the range o f situations facing the
drainage engineer : how to calculate the stage-discharge curve for a g iven
compound cros s-section , roughnes s and channel gradient ; how to design a
channel with berms for a specified duty ; the modification to those
procedures where the channel is gently skewed ; methods for natural rivers
of greater cross-s ection complexity ; suggestions for the incorporation o f
s imilar hydraulics into one dimensional models ; the broad effects on the
flow and boundary s tresses in the main channel and on the flood plain ; and
s ome preliminary views on sediment transport under compound channel
conditions .

1.3 Appr oach to design

1�3.1 The usual method found in hydraulic text books is outlined in the
following quotations : "The cross-section of a channel may be composed o f
several distinct subsections with each subsection different in roughnes s
from the others . For example , an alluvial channel subj ect to seasonal
floods generally consists of a main channel and two s ide channels . The s ide
channels are usually found to be rougher than the main channel ; so the mean
velocity in the main channels is greater than the mean velocities in the
s ide channels . In such a case , the Manning formula may be applied
s eparately to each sub section in determining the mean velocity of the
subsection . Then , the discharges in the subsections can be computed . The
total discharge is , therefore , equal to the sum o f these discharges . " ( Ven
Te Chow , 1 9 59 ) . " • • it i s neces s ary to split the section into subsections • • •

Manning ' s formula may be applied to each in turn , and the discharges can be
summed . The divis ion of the section into sub-sect ions is a little
arbitrary . S ince the shear stress across the arbitrary divisions will be
small compared with the bed shear s tress , it may be ignored . " ( Chadwick and
Morfett , 1 9 8 6 ) .

25
1.3.2 The above seemingly simp le procedure begs several que stions , not
leas t o f which is the unsupported assumption that the simp le addition o f the
calculated flows through the arbitrarily s eparated flow zone s wil l give the
correct answer . It wil l be apparent in what fol lows that this i s not so
even in the basic case o f a straight channe l , and that the discrepancy is
too great to ignore . The interference between the s l ower moving berm flows
and the main channe l flow increases head losses signi ficantly , so that the
dis charge calculated by these " text book" methods wi l l be an over-es timate
of the true channel capacity , in extreme cases by as much as the bank- ful l
dis charge . However , once a decis ion i s made about the division l ines
between the zone s , the basic method is attractive ly s imp l e . What i s
required , therefore . a t least a s a first step , is a rt assessment o f the
corrections needed to a l low for the inter-zone interference . TQe
e s tab lishment o f empirical adj ustment factors forms the basis o f chapter 3 ,
and this concept is extended t o skewed channe l s i n chapter 4 .

1. 3. 3 The potential ly di f ferent ve locities in the deep channe l and over the
berms generate s trong shear and turbulence at the j unction be tween the
zones , and thi s inf luences the f l ow for a considerable distance either s ide
of the bank l ine . Thi s extra turbulence i s the mechanism for extra head
l o s s and it must depend on the transverse gradient of vel ocity which
characterises the shear l ayer . Modern turbulence theory is capab le o f
handl ing such situations and has the cons iderable advantage o f having
complete generality . Empirical adj ustment factors are restricted to the
range of cross sections teste d , and these tend to be " c lassic" compound
sections with a trapezoidal deep channel and horizontal berms . Methods
based on turbulence theory can dea l with any shape of cross section , so
deserve careful asses sment . These methods are reviewed in chapter 6.

1.3. 4 Given an improved method o f handling the hydraulics o f compound

channels , the conventional algor ithms of one-dimensional mode ls may be
updated . The extra computational e f fort to do thi s wi l l be minima l provided
simp l e directly solvab le expres sions can be found for the various factors
influencing the correction required to the bas ic compound sect ion
calculation . As far as pos sib l e , therefore , algebraic formulations wi l l be
p rovided g iving a direct solution to the p rob lem of computing the
stage/di scharge function in compound channe l s .

26
1.3.5 This report thus provides a rea sonably comprehensive treatment of
compound channels , including basic theory and a review o f research a s well
as the recommended design methods that have resulted from that work . The
hydraulic engineer need not follow through the whole publication each t ime
he wishes to design a compound channel . The Summa ry Report contains the
basic methodology for assessing the stage-discharge function for a
" standard" compound channel consisting o f a trapezoidal channel with berms ,
and also explains how a typical river section with flood plains can be dealt
with , even if its section is not the ideal compound trapezium. The detailed
support for the recommended method will be found in Chapter 3 . An example
manual s olution o f the design equations will be found in Appendix 6 , though
in due c ourse i t i s anti cipated that computer soft-ware will be developed to
s implify applic at ion .

27
2. FLOW RESISTANCE IN CHANNE LS OF COMPLEX CROSS SECTION

2.1 Resume of resistance for s impl e open channel s

Avai lable formulae :

2 . 1. 1 The most commonl y app l ied formula for open channel s i s the Manning
e quation :

• • • 2.1

where

V = average vel ocity o f flow through the cross-section

n = Manning ' s roughne s s coe fficient
R hydraulic radius ( hydraulic mean depth) given by cross section area , A ,
I wetted perimeter , P .

(Nomenclature i s defined on first appearance and i s l isted in ful l in

Section 1 3 )

Although genera lly ascribed t o Manning , in fact thi s e quation was not one o f
those recommended i n the usua l ly quoted paper (Manning , 1 89 1 ) . I t is ,
nevertheless , of almost universal popularity for typical open channel s . It
should be used with some care , however , because it is by no means a
universal resistance function : it is unsuitab l e for extremely rough
conduits , such as corrugated culverts and unlined rock tunne l s or for the
smoother range of man-made structures , such as good qual ity concrete
spil lways and drainage channel s .

2.1.2 The limitations o f the Manning equation for simp l e (non-compound)

channe l s are best explained by re ference to the comprehensive framework for
flow resistance provided by turbulence theory . It would not be appropriate
to go into great detai l here , but in essence turbul ence theory provides a
description of the vel ocity distribution and its dependence on the roughnes s
o f the boundary and on fluid properties , inc luding viscosity . The velocity
distribution functions for smooth and rough boundaries usually quoted are

28
tho se derived by Prandtl ( 19 3 3 ) , although more recent theories have also
been propounded giving somewhat more complex express ions :

• • • 2.2

where

u = the local mean stream velocity a distance z from the boundary

V* the shear vel ocity defined as (�/ p )
� the shear stress at the boundary
p = the density o f the f luid
K . = a turbul ence constant ( the von Karman constant )
z 0 = a constant o f integration representing a boundary displacement

For .smooth boundaries :

2.3

where

u = f luid viscos ity

� = a constant

For rough boundaries :

z 0 = ak5 .. . 2.4

where ak5 = a l inear measure o f the textural roughness of the boundary

Thus for . smooth and rough boundaries respectively the velocity distribution
is. given by s imi lar functions :

Smooth :

u/v* A ln (v* z/u ) + B • . • 2.5

Rough :

29
u/v* = A ln ( z I k s ) + B' . . • 2. 6

with A 1 /K 2.5 , B = 5 . 5 and B ' = 8 . 5 ( according to Nikurads e ' s results ,

193 3 ) .

2. 1.3. These quite fundamental functions for the l oca l ve locity

distribution may be integrated over the cross-section of flow to give
resistance equations . Although that procedure might in theory cover a range
o f cross-section shapes , only two are relevant here : a circular s ection and
an open channel wide enough to ignore the influence of its banks . I t is
usua l too to abandon Naperian logs ( ln) in favour of common logs (base 1 0 ,
log) , and also t o modify the coefficient values on the basis o f c lassical
experiments on pipe friction by Nikuradse , thus obtaining :

Circ_ular pipes :

Smooth :

1/i f = 2 . 0 log ( Rei f/ 2 . 5 1 ) • • • 2. 7

Rough :

1/if = 2 . 0 log ( 3 . 7 1 D/k 5 ) 2 . 0 log ( 14 . 8 R/k5 ) • • • 2. 8

where

f = friction facto r , g iven by 2gDS/Va

Re = Reynolds number de fined as VD/�
D = p ipe diameter
S = friction gradient ( the s lope of the energy gradient )

Note that R for a circular cross-section = D/4 .

Wide open channels :

Smooth :

30
1 / .J f 2 . 0 log ( Re l f/3 . 02 ) • . • 2. 9

Rough :

1/l f = 2 . 0 log ( 12 . 3 R/k ) . . . 2. 10

8

Comparing the p ipe functions with the wide open channel functions , both the
smooth and rough express ions are affected in exactly the same way by the
cross section shape .

2 . 1 .4 Many surfaces o f practical interest have a degree o f roughness that

render s them transitional between smooth and rough : this is so for many
l ined channe l s but i s much less l ikely to be the case for natural channels
and f lood p lains which wil l be hydraul ically rough . The most frequently
used resistance equation for sur faces that may l ie anywhere in the range
smooth to rough is that due to Co lebrook and White , ( Colebrook 193 9 ) often
shown in graphical form and referred to as " the Moody diagram" ( see for
examp l e Chadwick and Morfett , 1985 ) This very general equation is avai l ab l e
as design chart s ( Hydraul ics Research , 199 0) and tables (Hydraulics
Re s earch , 1990) and takes the form for pipe s :

V/ 1 ( 2gDS ) = -2 l og [ (k / 3 . 7D ) + 2 . 5 1v/D I ( 2gDS ) ) . . . 2. 11

8

For sections not too far removed from circular , this may be generalised by
rep lacing the pipe diameter , D , by the hydraulic mean depth , R , (D = 4R) , to
give :

V/ I ( SgRS ) = -2 log [ (k / 14 . 8R ) + 0 . 628v/RI ( 2gRS ) ] • • . 2 . 12

8

However , for wide open channel s the " correct" version is :

V/ I ( SgRS ) = -2 log [ (k / 1 2 . 3R) + 0 . 7 5 5v/R / ( 8gRS ) ] • • • 2 . 13

S

2 . 1 .5 The first term o f the above i s the rough turbulent equation for wide
open channe ls :

V/ / ( SgRS ) = 2 log [ 1 2 . 3R/k ] . . . 2 . 14

8

31
Note that flow depends on re lative roughness , k s /R , but not on the vi scosity
o f the fluid in this reg ion o f flow . Also that resistance fol lows a square
law : S is proportional to (velocity) for const ant re lative roughness . Also ,
it may be shown that power law resistance formulae are in good agreement
with the more academic logarithmic relation over certain range s o f relat ive
roughness , for example the Manning equation with R2 ' 3 is a good fit in the
range 7 < R /k s < 1 3 0 with n = k s 1 ' 6 /2 6 (ks in mm . ) .

2 . 1 .6 At this point , no preference need be expressed for one equation over

a l l others in terms of representing the basic resistance to flow in a simple
channe l . The straight conversion of the Colebrook-White transition equation
by r ep lacing D by 4R is not accurate for wide channe l s , but then there are
such uncertainties in the roughness value k s that they probably swamp the
difference between the factors 1 4 . 8 and 1 2 . 3 . In what follows , however ,
reference to the Colebrook-White equation wil l be t o the wide channel
version . This very general formula is preferable to the Manning equation
for re lative ly smooth lined channe ls , for which there wil l a l so be rel iable
information on the values o f k 5 to use ( see appendix 6 ) . Provided a good
estimate of k 5 i s available , it is also applicab le to rough channe ls .
However , the hydraulic engineer has traditiona l ly used the Manning equat ion
for such channels ( not inappropriately , of cours e ) so that because of its
popularity and the availabil ity of va lues o f the roughnes s coef ficient n for
many cases , the Manning equation could hardly be discarded . Although
inappropriate for the ana lysi s of laboratory experiments on smooth channe ls ,
i t i s far from being supers eded in engineering design .

2.2 Compound cross-sections

2.2. 1 Figure 2 . 1 il lustrates a conventiona l compound channel cross section

and serves also to define some of the terminology used in the manual :

depth o f main channel below the berms ( f lood p lains )

depth o f flow in main channel
= =
H - h depth of flow over berms ( f lood p lains )
= ha lf bed width o f main channe l
B = ha l f total channe l width at f lood p lain leve l , i . e . 2B = top width of
channel plus width of berm ( s ) or f lood plain ( s )
se = main channe l side s lope , horizontal/vertical

32
s f lood p lain s ide s lope
F
=

bw bed width o f main channel

c
bw base width o f each flood p lain
F
=

2.2.2 Figures 2 . 2 and 2 . 3 show the variation of the conventional hydraulic

properties o f compound channel s , the cross-section area A, the wetted
perimeter P and the hydraul ic mean depth R = A/P , treating the section as
one unit . Two cases are i l lustrated , a natural channel ( R Severn , at
Montford Bridge : Knight et a l . , 1989 ) and an artificial section ( the
exper imenta l channe l at Wal l ingford , with B/b = 4 .2) . The latter shown in
Figure 2 . 3 has hor izontal f lood plains and so shows radical discontinuities
at the bank ful l level in P and R, arising from the sudden increase in water
sur face width when f l ow exceeds bank ful l . The Montford Bridge se�tion
( Fig 2 . 2) has fl ood p lains o f unequal width with apprec iable c ro s s fa l l so
that there is no discontinuity in P o r therefore in R , but even so the
over a l l hydraulic mean depth halves as the flow expands to cover the f lood
p lains .

2.2.3 Ther e have been many f low gaugings at Montford Bridge and , treating
the f l ow s ection as a unit , these observations may be interpreted within .the
conventional frameworks of resistance funct ions . Figure 2 . 4 ( i ) shows how
the calcul ated value o f Manning ' s n varies with stage (Knight et a l . 1 9 89 ) .
As f low spreads to cover the flood plain the n value drops by a third ,
despite the know ledge that in reality the roughness of the flood p lains i s
not l e s s than that o f the main channe l . Thi s spurious reduction in
r e si stance arises because of the form of the Manning equation :

. • • 2 . 15

Thus a s R c a l culated for the whole se�tion falls by a factor of 2 as the

f�ood p la ins are covered , thi s i s reflected in a reduction in the calculated
n value , even though in no sense is there a reduct ion in the actual
roughnes s o f the flow boundaries . It is a spurious e ffect of treating such
a comp lex s ection as a unit .

2.2.4 The vari at ion o f the friction facto r , f = 8gRS / V , with Reynolds
Number , Re = 4VR/v i s shown in figure 2 . 4 ( i i ) (Knight et a l , 1 9 89 ) and a

33
looped function emerges . It i s most unl ikely that any des ign method could
result from such a presentation and there are two reasons for this :

- as the flood pl ains become inundated , the reduction o f R with increase o f

stage provides a somewhat arti f icial reduction in the calculated friction
factor , not related to the f low resistance , which actua l ly increases

- expressing the Reynolds number as 4Q/Pv , the rapid increase of P as the

f lood p lains become covered over-r ides any increase in discharge

These combine to produce a looped funct ion that has no physical meaning , so
that quite apart from rejecting Reynolds number as a suitable parameter for
l arge s cale rough turbulent f l ows , i t is obvious that treating the .flow

section as a uni t creates prob l ems because R is not monotonic with stage .
I t would seem , ther e fore , that any rational basis o f design must treat the
f lood p la ins (berms) separat e ly , in order to avoid the problems posed by
treating the cross-section as a uni t .

2 . 2. 5 The above findings are confirmed by laboratory test s . Figure 2 . 5

shows samp l e results from the Wal l ingford channe l expressed as the variat�on
of Manning ' s n with flow depth (Myers and Brennan , 1990) . Here the channel
and flood p l ains have equal roughness , being moulded in smooth cement
mortar . Again the radical reduction in apparent n value causing a
discontinuity at bank ful l flow is a spurious result of the sudden change in
the hydraul i c mean depth . This not only demonstrates the inadvisabi lity o f

treating compound s ections as a unit , i t a l so demonstrates the confusion

that would arise i f an overal l value o f Manning ' s n were to be used as an
a l l embracing coe f f ic ient covering not only the physical r oughness o f the
boundary but a l l other influences on resistance and corrections for
irrational methods of computation . It is firmly recommended therefore , that
fo� the purposes of design , Manning ' s n should be used only as a r oughnes s
coefficient rel ated to the phys ical roughnes s o f the channe l . Other
inf luence s on channel resistance wil l be expressed separately , through
appropriate adjustment factors .

2.2.6 Figure 2 . 6 s hows these results in the form o f friction factor p lotted
against Reyno lds Number , Re , treating the Wa l l ingford compound channel as a
uni t . Rather than a looped function as for the River Severn observations ,

34
the horizontal f lood p l ains produce a discontinuity and overlap in Re ,
coupled with the art ificial reduction in frict ion factor when the banks are
inundated . Not only would such a dis continuous function be a problematic
basis of de sign , it also has little physical significance : any real
dependence o f resistance on Reynol ds number is masked by the discontinuity
in R when used in thi s way to describe the flow cross- section .

2.2.7 When conducting field measurements o f rivers using the velocity area
method , in essence the mean velocity in each vertical is estab l ished and
then the resulting dis charge intensities , q = average velocity x depth , are
integrated across the channel width . Thi s same procedure is app lied when
flow is above bank , of course , when the discharge intensity in the main
channe l depends largely on its depth , gradient and boundary roughnes s and
that over the flood p l ains depends on their flow depths , hydraulic gradients
and roughnesses . I t i s not surprising , therefore , that a similar approach
_
to calculation should be expected to provide at least first order accuracy
of estimation when assessing the upper range of the stage-discharge curve ,
knowing the cross section geometry , gradient and the roughnes ses of the
different e l ements of the cross-sect ion . The section would typica l ly be
divided vertical ly into zones with different depths and/or roughnesses , a.
resistance equation such as Manning would be app l ied to each zone separately
and the component discharges woul d be summed to g ive the total discharge for
the g iven stage . If that procedure were accurate , there would be l ittle
need for a new des ign method : the purpose here is to demonstrate its
shortcomings and hence to improve upon it .

2.2.8 When app l ied to a basic compound channel with a central deep channel
and side berms , the basic method reduces to three component calculations :
the main channe l and the two f l ood plains . By analogy with the ve locity
area method o f flow measurement , vertical divi sions between main channel and
f lood p lains woul d be the natural choice , and these division lines would be
left out o f the computation of wetted perimeter for the component part s , o f
course . There has been considerab le discussion , however , over the choice o f
the zone boundaries , o n the premi s e that i f i t were pos s ib l e t o define a
surface o f zero shear stress , this would be more logical than the assumption
o f a vertical p l ane o f separat ion . On this basis , s loping planes of
separation at the bank l ine have been considered . However , the basic
premise that an adequate knowledge of the location of a zero shear surface

35
would so lve the prob lem is faulty . This would only be s o if the extra
turbulence generated by the interference between the shal low s l ower flood
plain flows and the deeper faster main channel flow did not also influence
the flow patterns and hence the boundary shears in those zones . Even i f
surfaces of zero shear could b e defined for all cases o n the basi s of
research , a des ign method wou ld also require ful l information on how the
interference a f fected the estimation of f l ow in the separate f low zones .
Because o f their under lying logic and simp l icity , only vertical divis ions
between the deep channel and f lood plains are considered . Methods o f
correcting the basic flow resistance calculations either s ide o f these
divisions - or after combination - for the effect o f interference between
them wi l l be derived and exp lained .

2.3. Allowing for the effects o f interaction

2.3 . 1 Quantification of the effects of the inter ference between main .

channe l and f lood p lain has b een attempted by several authors using a range
o f methods . The main methods in the l i terature are :

1. Adjusting the division l ine between the deep and sha llow zones o f flow ,
perhaps coup led with including those sections in the wetted perimeter
of one or more zones

2. Us ing Manning • s n as a lumped resis tance coe fficient and seeking

empirical functions with the ratio of flood plain depth to main channe l
depth , and any other relevant flow parameters

3. As sessing the apparent shear force on the as sumed interface by

re ference to empirical information and so a l lowing for thi s in the
computations for each zone

4. Us ing experimenta l research to assess adj ustment factors for the

separate f low zones

5. S imi larly, but applying the correction factors to the total f l ow

6. Turbulence models predict ing the lateral spread of the interacting

shear layer and hence the lateral velocity profile .

36
2.3.2 None o f the methods based on al ternative divi sion l ines between the
main channe l and the flood pl ains removes the need for further adj us tment
factors and so they do not provide the sought after solution . This was
demonstrated by Nal luri and Judy ( 1985 ) , see Figure 2 . 7 ( a ) , and
independent ly by Prinos and Townsend ( 1984) see Figures 2 . 7 (b ) and ( c ) .
Consequent l y , vertical divis ions are used in a l l that fol lows , as it i s more
basic than the al ternatives that have been tested with little success .
Us ing the Manning equation as a lumped resistance parameter rather than
s o l e ly as a measure o f true roughnes s is theoretically unsound .
Figure 2 . 4 ( i ) showed how unreal such a procedure is for a real river , and
Figure 2 . 5 shows a typical set o f data from the Wa l l ingford facil ity
analysed in terms of Manning ' s n . The considerable reduction o f n when the
flow inundates the flood plain is spurious , as there is no actual reduction
in the boundary roughness . Thus methods 1 and 2 l i sted above are considered
inappropriate for des ign .

2.3.3 Method 3 above is based on the consi deration of the force equi l ibrium
on the compound channe l , for example as sketched in Figure 2 . 8 . For steady
uni form flow , the weight component per unit l ength down s lope is balanced by
the the boundary shear s tress integrated over the entire wetted perimeter �

2 . 16

wher e subs cripts AV , T , C and F denote average , the total section , the deep
channel and the flood plain respectively. In effect this de fines the
average shear stress , which must vary cons iderab ly around the perimeter .
Inserting the vertical divis ion between the flow on flood plains and main
channel , the equi l ibrium for the separate zones can be expressed as :

Main channel :

. . . 2 . 17

Flood p l a ins :

• . . 2 . 18

where

37
SF the shear force at the interface , assumed to assist flood p lain flow
=

I
and resist main channel flow

and

• • . 2 . 19

2.3.4 Correlations o f depth-averaged apparent shear stresses on the

interface have been carried out by Myers ( 19 7 8 ) , Wormleaton et al ( 1982) ,
Knight et a l ( 1983 ) , Baird and Ervine ( 19 8 4 ) and others for a range o f
channel and flood plain geometr ies , bed s lopes and boundary roughnesses .
The method i s to measure a l l the component s in the force balance equation
with the exception o f the interface shear , which can then be deduced from
equs 2 . 17 and 2 . 18 . Wormleaton et al o f fered the empirical relationship :

T = 1 3 8 4 ( OV) 0 • 8 8 2 ( bw /bw )
•
o . 7 z 7
• • • 2 . 20
AVI C F

wher e OV is the difference in mean velocity between the channe l and flood
p lain . Baird and Ervine ' s function was based on experiments on s ixteen
different cross-sections , with smooth boundaries :

T AVI = (H/ ( H-h ) -�) 1 • 5 / ( bw /h) [0 . 5 + 0 . 3 ln ( bw /h]

C F . . . 2 . 21
pg ( H-h) S

whe re � is the relative flow depth at which velocities either side o f the
division between main channel and flood p lain are c lose enough for T� to
AVI
be negl ig ible .

2.3.5 The ratio o f shear forces at the solid boundary to the streamwise
weight component was suggested by Radoj kovic ( 19 7 6 ) as an index of the
degree of interaction between the main channe l and flood plain sub-sections .
The coefficients are given by :

• • . 2 . 22

• • • 2 . 23

38
This formulation can be developed to give the total dis charge , if one knew
what the discharges in the separate zones would be in the absence o f
interface shear , Q ' and Q ' :
c F

• • . 2 . 24

Unfortunately the matter is not so s traight forward , as Q ' and Q ' cannot
c F
be calculated from an unmodified resistance equation : interaction affects
the shear at the solid surfaces as we l l as providing shear at the division
l ine .

2.3�6 The problem with approach number 3 i s thus that it is only a partial
solution to the des ign probl em . Firstly , a knowl edge of the inter facial
shear stress is by i t s e l f insuf fic ient : we a l s o need to have a method o f
as sessing how the shear stresses at the sol id boundaries are influenced by
the flow interfe rence . Figure 2 . 9 sketches how these b oundary shears might
be a ffected , and it is known from measurements made that they depart
s igni ficantly over an appreciable width from the va lue s that would be
calculated from a basic friction l aw for the s eparate sections . It seems
that the interaction changes the pattern o f secondary currents in such a �ay
that these external boundary stress change s are no l e s s s ignificant than the
internal interfacial shear s tress . Knowledge of the latter without the
former is of no avail in terms of design .

2.3.7 I f empirical information i s to provide a des i gn method , then i t makes

sense to include adjustment factors that cover both the e ffects mentioned
above : the change in s o l i d boundary shear stress from a normally calculated
mean value due to interaction and the shear at the assumed vertical
interface . Thus , with Qc " and Q " the calculated components of dis charge
F
using a �tandard resistance formula for the two zone s , with their individual
gepmetries , roughnes s e s and gradient s , then the total flow is given by :

Q = F Q " + F Q " . . . 2 . 25
T C C F F

where F e and F are the respective adjustment factors , to be determined on

F
the basis o f experiment . This i s the bas i s o f method 4 l isted above .

39
2.3.8 Appendix 1 lists the independent variables that control the flow in a
straight compound channe l : a l l the geometric properties of the section ,
density and viscosity o f fluid , gravity and s lope , and the roughnesse s o f
the main channel and flood p lain . I f the flow i s rough turbul ent then the
fluid viscosity does not influence the prob l em , and the dimensiona l analysis
o f Appendix 1 shows that :

where b e ' and b ' are mean widths rather than bed widths . For smooth
F
channe l s , the Reyno lds numbers of the zones would rep l ace the relative
roughnesses in the above statement ; transitional flow conditions WQuld
require both for comp l etenes s .

2.3.9 C learly an imposs ib ly wide range o f experiments would be needed t o

cover all ranges of the many r e levant parameters in any one research
programme . However , there are some 40 or so references in the l iterature to
experiments on compound channe ls . I f a l l were availab l e for further
analysis there may wel l be a wide enough body of information , though most�y
on smal l laboratory channel s ( Ho l l inrake , 19 8 7 , 1988 , 1989 ) . In attempting
to develop empirical methods , sirnpli fications were sought by individual
author s , for example by assuming that the primary effect of some of these
variab les is adequate ly taken care o f within the computations o f Qc " and
Q " . Different authors have there fore used a range of sub-sets and
F
sirnpli fications o f the above , although , as Appendix 1 makes clear , any
departure from the complete form of dimensionless statement makes the
remainder questionable : it is a matter for care ful experimenta l and fiel d
calibration to ensure that what remains is i n practice sufficient . One such
sub-set would be :

. • . 2 . 27

The s imp l i fications in this are :

- the ratio of friction factor s , f , suffices as a measure of the

interaction , the inf luence of Reynolds Number ( i f any) and relative
roughness being taken care of by the sub-sect ion flow cal culations

40
- the inf luence o f the s lope o f the bank forming the edge of the f lood plain
is too remot e to affect the overal l hydraulics
- the primary geometric effects are the relative depths and widths

2 . 3 . 10 To evaluate Fe and F from experiments requires that the flows in

F
the flood p lains and main channe l are measured separately , through careful
velocity traverses for example . Although this has been done in some test
series , notably those carried out on the Wal l ing ford flood channel facility ,
mos t experimenters have knowledge o f the total flow only. This leads on to
method 5 l isted in 2 . 3 . 1 , the introduction and assessment of an overall
correction factor :

F ( Q " + Q " ) . . . 2 . 28
.
T C F

where

2 . 29

A typical s imp l ified sub-set o f the above is simil arly:

. . • 2 . 30

Again , i f there is a vis cous influence on flow , as with smooth and

transitional channels , then the two Reyno lds Numbers would be required in
addit ion . Whether such restricted functions suffice is a matter to be
tested , o f course .

2 . 3 . 11 Reasoning from the case where both the main channel and the f lood
p lain are. very wide , Appendix 1 postulated that the basic form of overal l
di�charge adj ustment might preferably be considered as a deduction from
rather than as a multiplying factor for the sum o f the f lows ca lculated for
the zone s separately:

• . . 2.31

This method appears to have been first proposed in 197 1 ( Zhel eznyakov ,
1985 ) . Zheleznyakov a l so sugges ted that Q was much greater than Q ' the
C F

41
former being subtractive and the l atter additive . He a l so anticipated one
o f the suggestions in Appendix 1 that the correction term shoul d be
norma lised , i . e . expressed in non-dimensional form , by the bank ful l
discharge QBF . We might examine the fol lowing functional relationship for
straight compound channel s , a l ongside the other methods suggested in
Appendix 1 for normal i s ing the dis charge defect :

• . • 2 . 32

with the last term , the s lope a t the outside o f the flood p lains , being
s igni ficant only when the f lood plains are narrow.

2 . 3 . 12 Method 5 ·thus provides a feasible empirical approach to co�pound

channel design . What remains is to estab l ish the requis ite correction terms
for _ a wide range of conditions . The Wal l ingford experiments provide the
initial data s et to be used , but it wi l l be es sential to bring in other sets
of experimenta l data , both to cover a wider range o f geometries and
roughnes s combinations and to provide independent confirmation o f any
empirical deductions .

2 . 3 . 13 Method 6 , turbulence mode l l ing , has been described by E l s awy et a l

0 9 B 3 ) , Kel le r and Rodi ( 19 8 4 ) , Radoj kovic and Dj ordj evic ( 19 85 ) , Samuel s
( 19 89 ) . Knight , Shiono and P irt ( 1989 ) and others . Whereas other methods
are e ssent ia l ly one-dimensional , ie they seek an overa l l representation o f
the mean flow in the s ection , turbulence theory i s e ssentially three
dimens ional : it seeks a solution o f the Navier-Stokes equation for steady
uni form f low in an open channel o f general shape where there i s both bed
generated shear and l ateral shear . The so lution o f the ful l Navier-Stokes
equat ion is feasib l e , and becoming more acce s s ib l e as the ava ilab i lity of
power ful . computers increases , but it cannot yet be r egarded as an economical
to.o l normal ly availab l e to hydraul ic engineers . I t is for this reason that
the more practical developments have in essence reduced the method to a
two-dimensional approach , by us ing the depth integrated form o f the bas i c
turbulence equation . This a l lows for the e f fect o f lateral var iation in
discharge intensity though not for secondary current e f fects , a s igni ficant
l imitation , a l though more recently methods have appeared in the l i terature
that seek to account too for the e ffect of secondary current s , in e ffect by
lump ing these two influences together in an overa l l parameter .

42
2 . 3 . 14 The s olution proposed by Knight et al ( 1989 ) uses the depth
integrated equation together with the depth averaged eddy vis cosity , to
obtain analytical s olutions for constant transverse depth and for l inear
depth variation . Thes e so lutions are functions o f the assumed eddy
viscosity , the frict ion factor for the f l ow zone , its depth and cross s lope ,
as we l l as the channel gradient . This method was tested against the stage­
dis charge data for the River Severn at Montford Bridge ( see Fig 2 . 2 ( i ) for
its cross-sectional parameters ) , in e ffect being calibrated for that
s i tuation unti l it achieved a c lose degree o f agreement in terms o f
s tage-discharge . One o f the inputs to the mode l i s a set o f values for the
fr i ct ion factors in the seven flow zone s : so also are values o f the eddy
viscosity . Calibration o f the model consisted o f assessing and adjusting
these individual values , the friction factor s being functions o f the local
depth but the eddy viscositie s , though varying from zone to zone , being
constant with fl ow depth .

2. 3 . 15 Figure 2 . 1 0 il lustrates the l eve l o f agreement final ly achieved in

terms o f depth mean velocity distribution across the width , for three
dis charges that were observed in detail in the field (Knight et a l , 1 9 89 )
The mode l represents the main features o f the flow distribution wel l :
maximum ve locity in the main channe l and the discharge distribution across
it s width and typical flood p lain ve locities . I t s deficiency is mainly
c lo se to the upper edge of the bank between main channe l and flood plain ,
where the model indicates a more pronounced dip in the discharge intensity
than was observed . Although promi s ing , this approach mus t sti l l be regarded
as a research area . Indeed , until there is good information on the values
of the turbulent eddy viscosity to use in dif ferent zones of f l ow , and their
dependence on a l l the geometric and flow parameters , unfortunate ly it does
not yet provide a general design metho d .

2 . 3 . 16 Ke l ler and Rodi ( 19 8 8 ) use a k - e turbulence model , again i n a

ver tically integrated form, and used a series o f experiments on compound
channel s o f four di fferent shapes , including cases where channel and floo d
p lain roughnesses differed , to calibrate and validate the mode l . Thes e
procedures give a great deal o f detailed information , such a s the
distributions of shear stress and velocity across the width , and so testing
them against experimental results is not just a comparison of the s tage
di s charge curves . Figure 2 . 1 1 shows samp le s o f their results for velocity

43
and shear distributions across the channel width , with experimental
velocities for comparison . The Ke ller and Rodi mode l has some problems with
vertical and steeply inc l ined banks , but is nevertheless promis ing .
Undoubtedly an improved understanding o f the f luid mechanics is the way
forward , and turbulence mode l l ing i s the natural l ine o f development .
Whether it i s yet abl e to provide a general method o f des i gn for compound
channel s is doubtfu l , the criterion being the achievement of an accuracy o f
prediction within one o r two percent without individual calibration in
respect of the values of eddy viscosity to use .

2 . 3 . 17 The scope o f turbulence mode l l ing as a design procedure is

considered further in Chapter 6 . The vertical ly-integrated two-dimens ional
theory is given in Appendix 4 .

2.4 Features influencing the degree of interaction

2.4. 1 Equation 2 . 28 provides a framework for cons idering what might

already be known about the influence o f various flow and geometric feature s
o n the degree o f interaction between the main channel and the flood p lain ,
exemp l i fied by their inf luence on the factor F
by which the total flow
T
calculated without regard to interaction has to be mul t ip lied to provide a
rea i istic estimate . The rel evant features are contained in the functional
s tatement of equation 29 , general ised to include the vis cous e ffects for
smooth and transitional surfaces , which then become s :

F = func tion [ H/ (H-h) ; k /H ; k / (H-h) ; Re ; Re ; b c ' /H ; b ' / ( H-h) ;

T SC SF C F F
s ; s 1 • • • 2 . 32
e F

Thes e features are thus :

relative depth of flood plain flow to main channel flow

- relative roughnes s o f main channel
- relative roughnes s o f f l ood p lain
- main channel Reynolds Number
- flood p lain Reyno lds Number
- aspect ratio o f main channe l
- aspect ratio o f f lood plain

44
 - s ide s lope o f main channe l
- s ide s lope o f f lood pla in

2.4.2 Before considering in detail the large scale experiments conducted on

the Flood Channel Fac i l ity at Wa l l ingford , some broad indications from
previous ly pub l i shed work wil l provide a measure o f the problem faced by the
des igners o f even straight simply-bermed channel s . No researcher has
covered a l l the above possible influences : each worker has been restricted
to various degrees by the scale o f apparatus avai lable , and time scale and
funding for his activities . All have therefore had to compromi se on the
imp l i cations o f the dimensionle s s statement o f e quation 2 . 32 by covering a
l imited number of variables . The problem here i s that it can not be assumed
that the results o f the empirical analysi s o f , s ay , varying the widths o f
the f lood plains i n smooth channel s wi l l apply wi thout modif ication t o rough
channel s . We might hope it would , but this needs to be demonstrated . Thus
there are many sources o f information but in terms o f the empirical
functions derived therein , they are unlikely to apply to compound channel s
i n general .

2.4.3 I t is genera l ly agreed that relative depth is a maj or factor . As

s oon as the f l ood p lains become inundated , the flow in the main channe l
suf fers from the inter ference o f the s lower flood plain flow . Most
researcher s found this influence to be at a maximum at a re lative depth , H* ,
of the order of perhaps 0 . 1 to 0 . 3 and then to diminish progressive ly as the
depth of flow on the f lood p lain increased . The maximum reduct ion in flow
due to this interaction may be anywhere in the range 1 0 - 50% depending on
the source o f information and cross-se ction geometry . A s the relative depth
increases , then interaction diminishes because there is likely to be less
difference between main channel and f lood plain ve loci tie s , but in practi ce
it doe s �ot become negl igib l e unless the berms are relatively narrow . This
is . another maj or factor : the ratio of flood p lain width to main channel
width , wide f l ood p l ains tending to show wor s e interaction e ffects than
narrow one s .

2.4.4 The difference between the average velocities on the flood plain and
in the main channel is o ften us ed as a parameter for the degree o f
interference . Thus one would expect extra f lood p lain roughnes s t o
exacerbate the effect and the unlike ly combination of re latively smooth

45
berms with a somewhat rougher deep section to diminish it . How the overa l l
roughness o f the system affects the issue is not c lear , partly because few
experimenters have covered both smooth and rough sur faces in otherwise twin
sets of tests . Reynolds number is not expected to have any influence in
r ivers and streams because they wil l operate in the rough turbulent range of
f low . However , Reyno lds Number is a factor in laboratory scale tests
because the gener a l ly smooth laboratory channe ls do suffer from vis cous
e f fects . Bank s l ope must be a s ignificant factor as with vertical banks
i.e. a rectangular deep channel , there i s c l osest proximity of the fast
channel flow to the s l ower flood p lain flow . Gentle side slopes to the main
channel on the other hand would provide a transition zone that might l imit
the interaction e ffects . Any skewness between the axis of the main channel
and val ley f loor would be expected t o have a maj or e ffect on the interaction
proces s , because the f low exchange forced by tapering flood plain on one
side and expanding flood p lain on the other would radically influence the
exchange o f momentum . That more complex condition is covered l ater in
Chapter 4 : here only s traight channel s with a ligned f lood plains are
considere d .

2.4.5 The likely influence o f the interaction between main channel and
f lood p l ain flows depends on how comparable the hydraulic conditions in
these zones might be : if ve locities and depths are very s imi lar , then we can
expect interaction e f fects to be sma l l ; if they are very dissimilar , then
maj or e f fects are t o be expected . The degree to which the different zones
exhibit flow similarity wil l be referred to as their "coherence " : the
greater their coherence the more l ike ly is the hydraulics of the section to
approach s imple channe l (neg l igible interaction) conditions .

2.4.6 Channel conveyance is a useful parameter in considering how the

concept of coherence might be defined . Conveyance , KV • was defined by Ven
Te Chow ( 19 5 9 ) as :

KV = Q! I S • • • 2 . 33

but it i s pre ferab l e to rede fine i t to be cons istent with dimensional

analysis , as :

Q/1 ( 8gS ) A I (A/fP ) . . • 2 . 34

46
Thus the conveyance of a simp le channe l can be represented by the cross­
section area , wetted perimeter and friction factor . For a compound sect ion ,
the theoretical conveyance before a l l owing for any in terac tion effec ts
is g iven by the sum of the conveyance s of the main channe l and flood
plains :

• • . 2 . 35

for the s i tuation o f two symme trical f lood plains .

2.4.7 This l eads to a parameter for the coherence o f the channel section ,
namely the ratio o f the theoretical conveyance calculated by treating it as
a s ing l e unit to that calculated by summing the conveyances o f the separate
zone s . This concept is developed in Appendix 3 and in its most general
form , the s ection coherence is defined a s :

i=n i=n i=n

L
A t [L A . IL ( f . P . ) ]
� � �
i= l i i = l i = l
COH = . . . 2 . 36
i=n
L [ A�. / ( A�. / f �. P �. ) ]
i=1

Thi s parameter varies with flow depth in a given channel , of course , and
three cases are i l lustrated in Figure 2 . 12 : the Wal l ingford channel
i l lus tr ated in Figure 2 . 3 ; the same but with flood p lains reduced to 0 . 2Sm
wide ; and the Montford Bridge natura l r iver section shown in Figure 2 . 2a .
For the smooth Wal l ingford channel , the appropriate friction factor was used
(varying with depth) and for the Montford Bridge s ection a constant value of
Manning ' s n was app l ied for this i l lustration , with depths related to the
lower edge of the f lood p lains .

2.4.8 The artificial channel with horizontal f l ood plain and flood
pl ain/main channel width ratio 3 shows a particularly low COH value of about
0 . 43 when the f lood p l ains are first inundated , increasing to 0 . 9 5 when the
flood p lain flow depth e qual s the depth o f main channel . With narrow flood
plains , width ratio 0 . 3 , COH is l e s s s ensitive to depth and c loser to unity ,
lying between 0 . 85 and 0 . 9 4 . The natural river section has wide flood
plains with some cro ss fall (note that Figure 2 . 2a has considerab l e vertical
exaggeration) with minimum COH value ( 0 . 6 1 ) , not j us t above bank ful l but
when the ful l width o f f l ood p la in i s inundated . Above this the trend is
very s imi lar to the wide l aboratory channel , whil s t below the trend is

47
towards unity because its sloping flood plains avoid the disconuity in COH
at bank ful l . ( Fo r these calculat ions the main channe l zones 3 , 4 and 5 o f
Figure 2 . 2a were taken together , as were the remaining f lood plain areas)

2.4.9 Whether this de finition o f channe l coherence provides a co-ordinating

parameter in the analysis o f experimental results remains to be seen . Its
potential benefit i s that it brings together i n one parameter most o f the
factors expected to influence the hydraulics o f compound channel s . A
cor o llary is that the closer to unity COH approaches , the more l ikely it is
that the channe l can be treated as a singl e unit , using the overal l
geometry .

48
 �
B

W
C

,. .,

:
' I /
/L1
"'
i I <:tJ ,
1
SF
h

I. .I l r
HF
b
•I
bwF bwc

Fig 2.1 Compo u nd channel cross-section: definitions and

termi nology

1m
s H
s = -4.0
1 h 6m
s = -83.0
2
s
3 = -1 .33
s = 00
s : = +1 .50 5 x vertical
s6 = +23 .0 exaggeration
s = +7.1
7
63m

0 1 00 200 300 400 500 600 Area m 2

62 �-
I I I I I I 1 1
20 40 60 80 1 00 1 20 1 40 1 60 Perimeter (m) -r---�--�----�
,
60 -·
P "' ' A -- - R
I .,...
-

-
-
I -

- _,_
58
, 2. 57.282
z ..... ..
0
_

/.. ... 2. 56282

0
56 .. - .,
I /
5
G) : /
> t/
fD 54
....I
I
I'I
52 I I
I
I I
2. 50.282
00 �------�--J---�--�
0 1 2 3 4 5

Hydraulic radius (m)

JBW/S/1 0-91130

Fig 2.2 C h a n nel at Mo ntford Bridge, River Severn ; (i) cross-section

(ii) hyd raulic parameters; cross-sectional area, A, wetted
perimeter, P, hydraul i cs mean depth, R. (Kni g h t et al, 1 989)
 €
I
'

I Max flow depth

SF = 1
0.30 _l !
0 1 5m
. s -1
,
I
.I
I i

1 2.25m 0.1 5m
- ------------------------�
�- 1�

0 0.5 1 .0 1 .5 Area m2
300
---�--���-�-

�-

250

200

E
-

Bank full
·

1 50
discontinuity
s
..s:::
0..
0
(D I
1 00 l�p
I
I
I
· 50 I

0
0 50 1 00 1 50 200

Hydraulic radius (mm)

JBW/6/1 0-91130

Fig 2.3 Flood Channel Facility at Wallingford : (i) cross-section

(ii) hydraulic parameters; cross-sectional area - A, wetted
perimeter - P, hydraulic mean depth - R
 61 H (m)
(H-h)JH
10
60
9
59
8
58
6 JL 57282
0 57
7
E
-
JL 56282
1 56
6
j
5
55
4
54
3
53
0 0.01 0.02 0.03 0.04 0
Manning's n

o.� r-------�

0.1 0 Bank full

.;:::-
0.08
!
- 0.06
...

�
- 0.04
1 47
--
- -
r::
0 --
+: -
() -
- ...
·;:: 01 19
u.
21 5
""-
0.02 1 97 Discharge (m3s.1)

1 07
Reynolds number (4UR/v)

JBW1711 0.91130

Fig 2.4 River Severn at Montford Bridge: (i) variation of apparent value
of Manning's n with stage (ii) variation of friction factor - f, with
Reynolds Number .. Re. (Knight et al, 1 989)
 0.32
Legend
0.30
Bib
Trapezoidal
0.28
X
A 2.2
4.2
0.26
D
0 6.67
0.24

0.22

0.20

:[ 0.18
•

0.1.6 [Link]
0)
.1!1
r.n

0.14

0.1 2

0.1 0

0.08

0.06

o:04

0.02

0
5 6 7 8 9 10 11 12 13 14

Manning's n (x1 o"")

JBW/8/10-91130

Fig 2.5 Flood Channel Facil ity at Wallingford : variation of apparent

value of Manning's n with stage. (Myers and Brennan, 1 990)
 0.0225 r--�-----,

s - 1 .021 x 1 o·3
s ... 1 .0

0.0200

0.0175

Bank full
discontinuity
0.01 50

0.01 25

Overbank
flows
0.01 00

0.0075 ��-L����--�---�-----L--���
5x1 04
Reynolds Number - Re

JBW/9/1 0·01130

Fig 2.6 Flood Channel Facility at Wallingford : variation of friction factor - f

with Reynolds Number - Re.
 Single channel method
Q)

- ..... Vertical interface, excluded

a

o-
�

Vertical interface, included

<IS 1 .3
..c ,._
- -
Horizontal interface, excluded
0
Ill
- - .o
- -

Horizontal interface, included

'6
- - c o
- -
�
... Diagonal (bank top to WSL at CL), excl.
:::J , v v

Diagonal (bank top to WSL at CL), incl.

1 .2 -
!3
+-
'\,
Q)
'\, • •
'
E
.2
Q) , _ _.o
�
1 .1
'
<IS
..c
0
�
.!/!
"

� 1 .0

�
Relative depth H., (H-h)IH
"5 0.9
0
lil
0
0
0
� 0.8
a:

Q)

'Single-channels' method
�
b
'Separate-channels' method, vertical excl.
<IS
.c
�
W<;>rmleaton et al (nr = 1 .9)
1 .8

.. .. Pnnos et al (nr = 2.0, Wc = 20.3cm l

'6
..
.. .. . . . ......
� . .. .;.: .
Pr!nos et al (nr = 2.0, Wc = 30.5cm
6
:::J

Pnnos et al (nr = 1 .3, Wc = 20.3cm

�-;- . .. ..
1 .6
!3
· ·. ·��
.
.... . . �
0

Myers (nr = 1 .0)

Q)

-·--
X

.a........"".
E D
0
1 .4 -- ·

··
.2
-- ·
---

"""' · · - ·- .. .. - -a_ ·- . ..
Q) -o. -
,. _ _ _ _ _ _ _ _
--
·
•
�
•

· ·x.. .. .. · ·0"" · · ··:�� - - - -. ..

_
•
• •
<IS
.c
0
.!I!
1 .2
"
..A.J? �JO o.;5 ... o.3o.,. o.35 ·?'4o- � o.45
-
]i
0.05 &·J.P
· - - ..
•• • •• • 0.50
.2 1 .0

�
"5
�0 0.8

0
.Q
1ii 0.6
a:

Q)

'Separate-channels' method vertical incl .

� c
<IS

'Separate-channels' method, horizontal excl

..c
�
1 .8

Wormleaton et al (nr = 1 .9)

Pr!nos et al (nr = 2.0, We = 20.3c1Tl)
'6
�
�
Pr!nos et al (nr = 2.0, Wc = 30.5cm)
:::J 1 .6 A• A

Pnnos et al (nr = 1 .3, Wc = 20.3cm)

!3
••
•
0

Myers (nr = 1 .0)

Q) X
E
c
.2 1 .4
Q)
�
<IS
.c
0 1 .2
Ill
'6
]i
.2 1 .0

�
"5
0 0.8
iii
0
0
0
�
0.6
JBW/10110·91130
a:

Fig 2. 7 Various treatments of division between zones i n calculation of total flow from
basic zonal flows, in co mparison with measured d ischarge.
(a) Nall uri and Judy ( 1 985) data for assymetric channel, b = 0.1 95 m, B = 0.695m,
h = 0.1 5m, seven division methods.
(b) Results reported by Prinos and Townsend ( 1 984), full l i ne single channel
with common avera g e velocity, broken line vertical interface not i nclud ed in
wetted peri meter.
(c) Ditto, broken l i n e vertical interface i ncluded in wetted peri meter, full l i ne
horizonta l i nterface n ot i ncl uded.
Fig 2.8 Balance of forces in elements of a vertical sided compound
channel . (Wormleaton and Merret, 1 990)

Reduced shear Increased shear

Note:
Shear stresses are plotted on alternate
sides of the flow section for clarity.
They do not change sign.

JBW/11110-91130

Fig 2.9 Sketch of boundary shear stress distri bution in compound channel
 -- Analytical solution X data Stage (m)
2.00 H = 7.81 0 = 330.80 A 6.09

a= 330.80 m3s·1

..c:
a..
� 0.50

0.00
-1 00
Lateral distance Y (m)

-- Analytical solution X data Stage (m) . Q( m3/s)

2.00 H = 6.92 0 = 220.60 A 5.20 229.8

E
�
� 1 .50
� a= 220.60 m3s·1
�
Q)
> 1 .00
lii
E
Q)

..c:
a..
� 0.50

0.00
-1 00 -80 -60 -40 -20 0 40 60
Lateral distance Y (m)
20

-- Analytical solution X data Stage (m)

2.00 H = 6.45 0 = 1 88.80 A 4.73

E
�
::J"'
1 .50
� a= 1 88.80 m3s·1
]CD
>
c: 1 .00

E
«<
CD

..c:
a..
� 0.50

0.00
- 1 00 -80 -60 -40 -20 0 40 60
Lateral distance Y {m)
20

JBW/1211 0-01/30

Fig 2.1 0 Depth mean velocities deduced from Knight et al model compared
with field data from Montford Bridge
 1 .6

1 .4

1 .2 X Experimental data (velocity)

Prediction (velocity)

Prediction (bed shear stress)

1 .0 2.0

E
< 0.8 1 .6
I::J
0
X
0.6 X
1 .2 .ti:
�

,
0.4 , 0.8
..
�
- - - - - - - - - - - - - - - - - - - - - - - - - - - -
,

0.2

y/8
0 0.1 0.2 0.3 0.4 0.5

1 50 750mm 200

t 'I'
2so ·I
I«
508mm i 'I'
1 02
•!
1� t .
1 ft
I
60 I 20.4 :
I
l
03 .

I
I
(C) (B) q_
<l

2.2 2.2
2.0 o--o Experimental curve (velocity) l
2.0

Prediction (velocity).
11
I I
1 .8 1 .8
I I
Prediction (bed shear stress)
1 .6 I I 1 .6
I
1 .4 I 1 .4
E I 0
< 1 .2 I 1 .2
I::J
1 .0 1 .0 s
... -
0.8 .. 0.8
,
.. ..
_ _ _ _ _ _ _ _ _ _ _ _ ,

0.6 0.6

Rood lain
0.4 0.4
0.2 0.2
0

y/8
0 0.1 0.2 0.3 0.4 0.5
JBW/13110-111130

Fig 2.11 Velocity and s hear stress p redictions com pared with laboratory
measurements, by Kel ler and Rodi K-e model
 �
'i; 0.5

0
0.3 0.5 1 .0
as single channel
COH a = Conve�ance
Sum ofcomponent conveyances

JBW/14110-9100

Fig 2.1 2 Chan ne l coherence - COH3, as function of ratio of flood plain flow depth to
main channe l depth : (i) w•He h o rizontal flood plains; (ii) narrow horizontal
flood plains; (iii) natural river c hannel with sloping flood plains.
3. HYDRAULIC DES IGN BASED ON EXPERIMENTAL ADJUSTMENT FACTORS

3.1 Research at Wa l l ingford

3.1.1 With the exception of some large-scale tests several years ago in an
outdoor fac ility in Amer ica ( US WES , 1 9 5 6 ) most o f the work to date on
compound channe l s had taken place in various laboratory channel s in
univer s ities , between 6 and 1 5 m in length and between 0 . 6 and 1 . 2 m wide .
These were often too sma l l for a real istic flood p lain width to be
s imulated , and at a scale which gave shallow flood plain flows and low
Reynolds number s so that ful ly turbulent flow may not always have been
achieved . For this reason , it was decided to concentrate newly available
special UK resources in a large central fac ility . The SERC F lood Cpannel
Fac i l ity ( SERC-FCF) was j ointly funded by the S cience and Engineering
Research Counci l and Hydraulics Resear ch Ltd , (HR , Wal l ing ford) at whose
premises the equipment was built . The research program was organised as a
series of individual but closely co-ordinated proj ects , largely but not
exclusively funded by SERC . The research carried out at this national
centre for flood channel research has been instrumental in providing an
uni que set o f large scale data , which are both accurate and comprehensive �

3.1.2 The fac i lity at Wal l ingford provided a channel 5 6m long with total
width 1 0m , with discharge up to 1 . 0 8m3 / s . The layout of the flume is shown
in Figure 3 . 1 and P late 1 . I t was fully instrumented for measurement of
water leve l s (hence hydraul ic gradient ) , discharge (by standard ori fice
meters ) , velocity traverses using miniature prop e l l er meters , boundary shear
stress (by Pres ton tube ) , three dimensional flow patterns and turbulence
measurements , using laser anemometry . Many of the measurements involved
computer control of the instruments themse lves , as we l l as sophisticated
data logging .

3.1.3 I t would not have been practicab l e within the funds availab l e to make
the s lope of the fac il ity adj ustable . Hence a l l experiments were with fixed
channel gradient of nominal ly 1 in 1 000 . The channel depth in the first
series of exper iments on straight channel s was kept at 0 . 1 5m and maximimum
flow depth was 0 . 3 0m , expected to cover the main region o f adverse
interaction between the flow zone s . The bed width of the deep channel was
1 . 5m in a l l these tests , but f lood plain widths , and channel and flood p lain
bank s l opes , were varied . Tab le 3 . 1 shows the var ious geometries used in
49
the Series A tests on st raight channels . Most o f the research was with
channel and flood p lains formed o f smooth cement mortar , but some tests were
also made with the f lood plains roughened by rods through the ful l depth o f
flow ( see Plate 2 ) . Later series o f tests concerned skew and meandering
channe l s ( see P l ates 3 and 4 ) , and also dispersion .

3.1.4 Preliminary tests were conducted to estab l ish the appropriate

resistance function for the cement mortar surfaces , and also for the
rod-roughened f l ood p lains . The former included tests at depths above the
normal bank ful l condition of 0 . 1 Sm depth , having extended the banks upwards
to give a simp l e trapezoidal cross-sect ion 0 . 3m deep . The results o f these
calibrations are reported in Appendix 2 .

3.2 Other sources of experimental data

3 .2 . 1 Other sources of information were also uti lised in extending the data
base o f stage/discharge information for compound channel s . This additional
data came from the fol l owing publ ications in the research literature :

Asano , Hashimoto and Fuj it a , 1985

Ervine and Jasem , 199 1
Kie1 y , 199 1
Knight , Demetriou and Hamed , 1984
Knight and Deme triou , 1 9 8 3
Myers , 1978 , 1984 , 1985
Prinos and Townsend , 1983 , 1984
US WES , U S Waterways Experiment Station , 1956
Wormleaton , Al l en and Hadj ipanos , 1982

Re ferences are given in Chapter 1 2 , Volume 2 .

3.2.2 These tests in hydraulics laboratories covered a wide range o f

conditions , both in terms of geometry and roughnes s ratio between main
channe l and flood p l ain . The range of main channe l width/depth ratios was
from 1 . 3 to 3 0 ; overal l width ratio B/b up t o 3 0 ; gradients from 0 . 2 2 to
1 . 8/ 1000 ; and flood p lain roughnes ses up to three times the main channel
value , in terms of Manning ' s n. However , much of this research was carried

50
out in Univers ity hydraul ics laboratories , and so was o f relatively smal l
scale . In order to avoid severe Reyno ld ' s number e ffects e . g . possib ly
laminar flow , as wel l as measurement difficultie s , depths had to be kept
relatively high compared with widths . There is thus a dominance o f work at
channel aspect ratios of the order o f two , far removed from the usua l range
o f practical compound channel s , and certainly far different from rivers with
f lood p lains . The exception to this is the research by Asano and
co l leagues , who used a range o f channe l aspect r atios based on an analys is
o f r ivers in Japan . This is the only work at aspect ratios greater than
than ten , as used in the FCF at Wal l ing ford . The Japanese r ivers show a
dominant aspect ratio between 20 and 3 0 ; and a dominant B/b ratio of 3 to 5 .
Thes e values may be representative o f many al luvial rivers o f modest s ize in
the UK too .

3.2.3 The largest scale studies other than those in the FCF were the
l imited series o f tests a t Vicksburg , at the US Waterways Experiment
Station . These were conducted in an out-of-doors 9m wide flume , and
included tests with f l ood plains roughened by the addition o f sheets of
expanded metal .

3 . 2 . 4 . The criterion for choos ing data for analysis was largely the
avai l ab i lity of a set of stage-discharge results , though the information
given was not a lways as comprehensive as required for r e l iab le analysis .
Typical gaps in pub l i shed information are accurate resistance functions for
both the main channe l and flood plains , without which interpretation of the
stage-discharge data in the context of determining corrections to basic
zonal computation procedure is very problematica l . Thi s is especial ly so
where arti ficial roughness is used on the f lood p lains . Another typical gap
in information is the water temperature . Much o f the research was with
smooth surfaced cross-sections , so water t emperature is required to estimate
the fluid viscosity , which is relevant in smooth - and fairly smooth -
turbulent flow conditions . The pub l i shed papers are dominated by the use o f
the Manning formula , whether for smooth o r rough conditions . This
resistance formula is most appropriate for rough turbul ent conditions ,
though it can r epresent smooth condit ions in an open channel system at a
given gradient reasonably wel l , with the coe fficient n then taking on some
o f the role of Reynolds number and hydraul ic gradient ( see Appendix 2 ) .
However , with the forms of artificial roughness typica l ly used , it can not

51
be assumed that quoted values o f Manning ' s n are accurate over a range o f
depths .

3.2.5 There have been many research proj ect s on compound channel s , so not
a l l data from past research could be included in the analyses . In one or
two cases , the research was at too sma l l scale or at geometries too far
removed from practical conditions to be considered worth fol lowing up
anyway . In other cases there was a lack of basic information that could not
be resolved through correspondence with the research group . The aim of the
ana lysis of stage/ discharge data from other sources was essent ia l ly to
validate the predictive formulae obtained from the FCF research , by applying
those results to quite different geometries and roughness conditions . This
work is ful ly described in Chapter 5 , where it is used to val idate . and in
certain respects further calibrate , the methods of predict ing stage
discharge deve loped in this Chapter .

3.3 Recommended basic method

3.3.1 Hydraul i c design methods previously described in the extensive

l iterature on compound channel s were reviewed earl ie r , and it was conc luded
that for engineering purposes the most appropriate method would be to make
separate calculations for the main channel and f lood p l ain and t o adjust the
sum of thes e zonal cal cul ations for interference effects to provide the
overal l dis charge at the chosen flow depth . The adjustments required are in
essence empirical coef ficients derived from experimental work , e special ly
the definit ive results from the FCF at Wal l ingford . By carrying out the
analysis within a non-dimens iona l framework , general adj ustment factors were
anticipated that could be trans ferred to other sizes of channe l , and o ther
roughnesses , including of cours e field s cale systems .

3.3.2 The dimensional analysis given in Appendix 1 indicated an appropriate

l ine of attack , deducing the main parameters upon which the interaction
e f fect was expected to depend . These were reviewed ear l i er and , for a
smooth channe l and flood p lain as in most o f the tests at Wal l ing ford ,
reduce to :

( H-h) /H : re l at ive depth o f f lood plain flow to main channel flow

b c ' /H : width/depth ratio of main channel
b ' / ( H-h) : width/depth ratio o f f lood p l ain f low
F
52
 s : s ide s lope o f main channe l
e
s : s ide s lope o f f lood p lain
F
Re : main channe l Reynolds Number
c
Re : f lood plain Reynolds Number
F

The nomenclature is i l lustrated in Figure 2 . 1 and is defined in section 1 3

o f Vo lume 2 . Note that width/depth ratio i s a lso re ferred t o as • aspect
ratio ' .

3.3.3 The experimenta l constraints meant that with a fixed channel s l ope
and s ingl e value of main channel depth , the two Reynol ds Numbers depend
direc t ly on f l ow depth and hence on ( H-h) /H . Reynolds number c ould not be
varied independently but in any event the main viscous influence would be
accounted for in the c omputation of basic zonal f lows . The width/depth
ratip ( a l s o referred to as the aspect ratio) of the main channel wa s not
adjusted { except as a secondary consequence o f changing the main channel
side s lopes ) , and this was a significant constraint . In most tests the
f l oo d pl ain side s l ope was fixed at 1 in 1 , though thi s was not an important
factor . Thus the experiments on smooth channel s are fully covered by a
reduced set o f dimensionless independent variab l es : relat ive depth ,
ratio o f f l ood p lain width to main channel width , and channel s ide s lope .
The experiments with roughened f lood p lains bring in an additional
parameter , acc ommodated as a ratio of main channel and flood p l ain friction
factor s .

3.3.4 The influence on the flow o f the interaction between main channe l and
flood p l ain f l ows depends on how comparable the hydrauli c conditions in
the se zones are : i f velocities and depths are very s imilar , then we can
expect interaction e ffects to be small : if they are very dis s imilar , then
maj or effects are to be expected . The degree to which the different zones
exhib it flow s imilarity is referred to as their " coherence" : the greater
the ir coherence (unity being the maximum value ) the more l ikely is the
hydraul ic s o f the section to approach s imp le channel (neg ligib l e
interaction) conditions . The development o f this concept is exp lained in
Appendix 2 . 3 . Several di fferent formulations were cons idered , a lthough a l l
incorporated the basic definition :

3.3.6 I f the Manning e quation applies , and perimeter weighting o f the

frict ion factor is app l ied , then the coherence equation becomes :
53
" Coherence is the ratio o f the conveyance calculated as a single
cross-s ection to that calculated by summing the conveyances of the separate
flow zones "

The se calculated values come from a s tandard friction formula before making
any allowance for interference e f fects .

3.3.4 The treatment as a s ing l e section requires some a ssumption about the
overa l l resistance , and for thi s purpose the perimeter weighted friction
factor is used , deduced from the separate ( and calculab l e ) values for main
channe l and flood p l ains .

i =n i =n i =n
L A I [ L A iL ( f P ) ]
i i i
i = 1 i i= 1 i= 1 . • • 3.1
COH =
i =n
L [ A . I (A . /f . P . ) ]
J. J. J. J.
i= l

In the above i identifies each of the f l ow zones , and in the basic case o f a
trapezoidal channe l with two f l ood p lains n = 3 . Note that this i s the form
denoted COH in Appendix 3 .
3

3.3.5 For a conventional compound cross-section geometry , the coherence o f

the section may b e expressed i n terms o f the geometric ratios : let
A* = N A /A ; P * = N P /P ; H * = ( H-h) /H ; and f* = f / f ' where N is the
F F C F F C F C F
number of flood p lains . Then

( 1 + A* ) / [ ( 1 + A* ) / ( 1 + f * P * ) ]
COH = 3.2
A * / ( A* / f * P * )
. . •
1 +

In this form it is obvious that as A * becomes large ( deep flow on flood

p l ain) then COH approaches unity , for equal roughness of main channel and
flood p lain (when f * approaches unity as the depth increases ) . Also when A *
i s very sma l l ( f lood plains just inundated) COH approaches 1/ ( 1 + f*P * ) . As
A* and P * depend on H * , then for a g iven geometry COH also depends on H * .

54
 ( 1 + A* ) J ' l // ( 1 + P* ' ' 3 n* Z /A* 1 ' 3 )
. . . 3.3
1 + A* s ' 3 /n*P* 2 ' 3

3.3.7 As wil l emerge l ater , the most general formula for channel coherence
( equation 3 . 1 ) provides a use ful co-ordinating parameter in the analysis o f
the experimental resul t s . Its benefit is that i t brings together in one
parameter mos t o f the factors expected to influence the hydraulics o f
co�ound channel s , and so can take the p l ace o f relative depth as an
indicator o f how l ike a s ingl e channel the performance might prove . A
coro l lary is that the c loser to unity COH approaches , the more l ikely it is
that the channel can be treated as a singl e uni t , using the
overal l geometry . As f* is included in the general definition of COH ( see
equ 3 . 2 ) , dis s imil ar roughnesses are within its scope . COH is thus· an
independent variab l e calculated from the known geometry and basic resistance
function that may take the p lace of other independent variables ( re l ative
depth , side s l ope etc . ) .

3.3.8 The dimensional analysis o f Appendix 1 envisages several alternative

dependent variables that could provide the adjustment needed to the zonal
cal culation of discharge . These alternatives were tested , to see which
provided the most appropriate expl anation of the test results and which
would therefore provide a good basis of des ign . Of course in the design
s ituation , the zonal cal culations would be based on the engineer ' s favoured
resistance function with resistance coefficient appropriate to the known
sur face texture . The zonal calculations of discharge for the research
results at any given flow depth are wel l specified as a result o f the series
o f tests to establ ish the resistance functions for the smooth and rough
conditions . The relevant smooth equation was a s fol l ows :

1 / l f· = 2 . 02 log ( Re / f ) - 1 . 38 . • • 3.4

and the calculation for rod roughness is expl ained in Appendix 2 .

3 .3 .9 Ful l er detail s o f the analyses o f experimental results wil l be found

e lsewhere (Technical Report Number 4 , Oct 1990 , unpubl ished information) .
In what fol l ows , the main thrust o f the analysis is described but only the

55
succes s ful l ines o f attack in what was a compl ex sequence o f testing o f
alternatives will b e detai l ed .

3 . 4 . Analysi s of experimental result s

General

3 .4. 1 Tabl e 3 . 1 l i sts the tests carried out on straight a ligned channel s at
Wal l ingford . It shows their geometries and the number o f stage/discharge
results available in each s eries , with f l ow above bank height . Thes e
experiments were designed s o that they could b e grouped i n various ways t o
demonstrate the e ffect o f relevant parameters in turn . Thus series 1 , 2 , 3
and 5 cover a range o f different flood p l ain widths , with main channel bank
s lope o f 1 : 1 . This bank s lope also appl ied to the flood p lain edge with the
exc�ption of series 1 , at maximum f l ood plain width , when in any case its
influence would be minimal . Tests 2 , 8 and 1 0 kept flood plain width
constant but covered main channel bank s lopes o f 0 , 1 and 2 ( horizontal/
vertical ) . The flood p lain was edged at a s l ope of 1 in a l l these cases .
Tests 2 and 6 provide a comparison between the symme tric case o f twin f l ood
p lains and the asymme tric case o f a s ingl e f lood p l ain . Tests 2 and 7 , 8
and 9 , 10 and 1 1 are pairs with the odd numbers having rod roughened flood
p lains to compare with the even numbered smooth case s .

3 . 4�2 The basis o f analysi s was through a comparison o f the total discharge
calculated from zones separated by vertical divisions with the measured
dis charge at that depth of flow . Alternative parameters were considered to
estab l ish their relevance and s ignificance . Flow depth was represented in
three alternative ways :

- relative depth , (H-h) /H ; H *

ratio o f flood plain flow depth t o depth o f main channel , ( H-h) /h
- channe l coherence , COH , as defined in para 3 . 3 . 5 and e qu 3 . 2 .

The second o f these parameters for depth o f flood p lain inundation was soon
abandoned as the first , defined as relative depth H * , gave c learer , more
linear functions .

3.4.3 The discrepancy between the basic calculat ion and the measured flow
was parameterised in five ways :
56
- adj ustment factor , measured discharge/basi c calculation , DISADF
- prop ortionate discharge de ficit , (Q - Q ) /Q D ISDEFP
CALC MEAS CALC '
- dis charge de ficit as proportion of bank ful l f l ow , DISDEFBF
- discharge de ficit normalised by the calculated velocity di fference and the
product of f lood p lain f l ow depth and main channe l depth ,
Q * = {Q L - Q ) / (V -V ) (H-h) h
l CA C MEAS C F
- . the same apart from using the product o f total f low depth and main channe l
dep th , Q * = ( QCALC -QMEA ) / ( VC -VF ) H h
2 S

The philosophy behind thes e parameters was developed in Appendix 1 : Q* and

l
Q* incorporate the main channel depth , h , on the basis that width o f the
2
zone o f influence would be related to channel depth rather than channel
width i f the system was e ffective ly wide ( 2b/h and 2B/H very l arge ). . h is
coupled with ( H-h) or H to yie l d a p l ausib le cross-sectional area of
inf luenc e , to c ouple with the velocity difference between main channe l and
flood p l ain as an indicator o f the scale o f inf luence - and , o f course ,
providing the requisite dimens ions . Q*
and Q* are both very similar
l 2
non-dimensional expressions for the discharge deficit due to interference
e ffects ( hence the * subscript ) but Q * was found more useful : Q* has been
2 l
discarded there fore from what fol lows .

Regions of flow

3.4.4 It is wel l estab l i shed , mainly but not only from the FCF results ,
that the magnitude o f the interference between the channel flow and main
channel f l ow shows different trends as the f l ow depth varies . As an
examp l e , Figure 3 . 2 shows the stage-discharge results for geometry 2 , with
B/b = 4 . 20 . The test result s were first assemb l e d in order o f increasing
depth and then running averages of three were taken to smooth out s catter
arising from random experimental errors . ( This smoothing process was
ge�eral ly used : a l l data shown is of this type unless otherwise stated . )
This figure shows the factor by which the sum o f the calculated zonal flows
has to be multipl ied to agree with the total discharge measured in the
supply pipes ( DI SADF ) , plotted against relative depth. The flow passes
through three distinct regions o f behaviour :

57
- Region 1 is at relative ly sha l low depths where the interference
e f fects progressive ly increase with depth up to re lative depth 0 . 2 , when
the " loss" of conveyance is over 1 0 percent .

- Region 2 covers depth rat ios to 0 . 4 for this particular geometry ,

with the inter ference effect diminishing towards a discharge loss o f about
4 percent .

Region 3 occurs with further increase in depth , which again

increas es the interference e ffect , presumably because of some change in
secondary currents .

- Also shown on Figure 2 is COR , and it wil l be s een that DISADF a lways
l ies between COR and unity , ie a lways exceeds the " s ing le channe l "
computation but i s less than the sum of the zonal computat ions . It seems
possible that , had experiments been continued to greater depths , they
would have fol lowed the COR function , ie a s ing l e channel computation
becomes appropriate at considerabl e depths o f flood berm inundation . This
forms Region 4.

3.4.5 The depth l imitations between the regions shown in the samp le p lot of
Figure 3 . 2 are not genera l ; they depend on various parameters , and di ffer
cons iderably with rougher f l ood plains : nor can it be as sumed at this s tage
that if ( R-h) /h > 0 . 5 , the inter ference e ffects are neg l igible . What is
clear , however , is that different flow regions exist , and di fferent design
formulae are required for each zone , as we l l as the means for estab l ishing
which region a particular design case wil l lie within .

Infl uence of flow depth and flood plain width

3.4.6 The discharge adj ustment factor , D I SADF , has the merit o f
il lustrating directly the magnitude of the inter ference effect , and its
var iation with f l ow depth and f lood plain/main channe l width ratio . D ISADF
is shown plotted against the relative depth , ( R-h) /R , in Figure 3 . 3 . The
four cases tested show re lated variations of DI SADF with depth ratio , but
the trends are discontinuous due to the different regions of flow . Taking
test 2 , B/b c = 4 . 2 , from (R-h) /R = 0 to 0 . 20 , ( region 1 ) , D ISADF reduces
from unity when the flow first reaches the flood plain to a minimum value of

58
0 . 895 : in other words , at this minimum condition the actual flow is some 1 1%
l e s s than calculated by adding the separate zonal figures . As ( H-h) /H rises
from 0 . 20 to 0 . 40 , ( region 2 ) , D I SADF tends back towards unity , rising in
fact to 0 . 965 : in other words the zonal cal culation becomes more accurate ,
but never better than within s ay 3% . There is a kink in the trend at
( H-h ) / H = 0 . 40 , towards lower values o f DI SADF again , with a value o f 0 . 94
when ( H-h ) / H = 0 . 50 . ( Region 3 ) . This complex pattern i s we l l e stab l i shed :
the other cases show rather similar features .

3 . 4. 7 I t i s o f interest at this stage to s ee what might be learned from the

channel coherence , as defined earlier . Values o f COH for these cases have
been p l otted on Figure 3 . 4 , to the same scales as Figure 3 . 3 - but note the
di f ferent range o f depths . A comparison with the experimental results shows
that in a l l cases D ISADF l ie s between COH and unity. This means that a
sing l e channe l basis o f cal cul ation provides a lower boundary to the channel
conveyance and the zonal cal culation provides an upper boundary . What is
required i s a group of r e lationship s , covering the di f ferent flow regions ,
to show wher e between these l imits the true value lies .

3 .4.8 Figure 3 . 5 i l lustrates the di f ferent regions o f flow in terms o f

relative depth , ( H-h) /H , with the bounding value o f D ISADF = COH . This i s a
gener a l ised diagram so is un- s ealed . The four flow regions are indicated ,
though in some cases 2 and 4 might a lmost link together as 3 diminishes .
Start ing at sha l low flood p l ain depths ,

Region 1 is at relatively shallow depths on the fl ood p lain , with

inter ference increas ing broadly in proportion to flood p lain flow
depth . The extent o f region 1 depends on the flood p l ain width , narrow
flood p lains permitting it to extend to deeper f l ows .

Region 2 is where the trend o f DI SADF i s similar to but lies below the
COH funct ion . The sect ion is behaving more l ike a s ingl e channel , but
interference between f lood p lain and main channel adds to resistanc e .

Region 3 which is mos t apparent with wide flood p l a ins , indicates a

change o f flow pattern ( in secondary currents for examp l e ) which gives
rise to diminishing interference .

59
 Region 4 is at relative depths l arge enough for it to be possib l e to
treat the cross-se ction as a sing l e zone , with perimeter weighting of
the friction factor , when calculating the stage/discharge function .
This is so when COH > 0 . 9 5 or thereabouts .

3,4,9 D I SADF has been p l otted against COH for these four cases o f varying
flood p lain width in Figure 3 . 6 . The l ine D ISADF = COH is the result given
by a single channel calculation , with overal l friction factor calculated
from the perimeter weighted zonal values . The results for greatest depth
approach this condition , but have not extended to sufficient depth to
confirm it . It appears probab l e that a sing l e channel calculation becomes
acceptab l e at a depth ratio ( H-h) /H exceeding 0 . 5 , ie when the depth of flow
over the flood p lain is somewhat greater than the depth o f the main channel .
( It wil l emerge l ater that this very definite ly does not apply when the
f lo�d p lains are rougher than the main channe l ) . The different trends in
the flow regions are also identifiab l e in this p lot . The four regions wil l
b e considered in turn , beginning with the performance r e levant with fairly
sha l low flood p lain flows .

Region 1

3 . 4 . 10 At very sha l l ow depths on the flood plain , the Reynol ds number is

below the usual values considered necessary for turbul ent flow : at a f lood
plain depth of O . l Om ( H* = (H-h) /H = 0 . 0625 ) , Re 4500 , so results for H*
F
=

< 0 . 0625 have been disregarded .

3 .4. 11 Although the parameters D ISADF , D ISDEFBF and Q* 1 were also

considered , the Q* vers ion o f the discharge deficit was found most useful
2
in representing Region 1 . One o f ite advantages is that it provides linear
functions with the relative depth , H* = ( H-h ) /H , as in fig 3 . 7 , where
para l l e l lines provide a good fit to the experimental results :

B/b =
6 . 667 : Q* 0 . 89 + 9 . 48 H * 0 • • 3.5
2

B/b 4 . 2 : Q* 0 . 1 4 + 9 . 48 H * . . . 3.6
2

B/b 2 . 2 : Q* - 0 . 5 8 + 9 . 48 H * 3.7
2
. . .

60
These are approximately linear with B/b :

Q* = - 1 . 240 + 0 . 3 29 B/b + 9 . 48 H * . • . 3.8

2

The lower l imit o f appl ication should probably ensure that Q * is not
2
al lowed to be negative at sha l low depths : there woul d never be an addition
to the calcu lated total discharge . The upper l imit of app l icability of
these Region 1 functions wi l l be given by intersection with the function for
region 2 , which is considered next .

Region 2

3 . 4 . 12 Region 2 is at increased depths where the interference e ffects

diminish progressively , and extends to the depth at which the discharge
adjustment factor tends to kink back towards lower values again . The
discharge adjustment factor was seen to have a somewhat simi lar t rend to
the channel coherence in region 2 : for the four width ratios treated here ,
figure 3 . 4 shows that the COH values interlace somewhat , and the
experimental values of D ISADF interl ace in much the same way in flow
region 2 ( Fig 3 . 3 ) . It is a strange finding that on the p lots agains t H * ,
the COH functions provide a convincing fit to the D I SADF results ,
provided they are overlain with a vertical shift in (H-h) /H o f about
0 . 15 . There seems to be no phys ical expl anation of this : it is a rather
obscure empirical resu lt .

3 . 4 . 13 In its functional form , the channe l coherence may be wr itten as :

COH ( H * , channel geometry) = function ( H * , A* , P * , f * ) • • • 3 .9

Thus the ,empirical finding implying a shift of 0 . 15 in H * can be wr itten

as_:

D I SADF ( H * , channel geometry) = COH ( [ H * + 0 . 1 5 ] , channel geometry) • . • 3 . 10

This states that a cal cu lation o f COH at H * + 0 . 1 5 does duty as DI SADF for
H* . The lower l imit o f appl ication o f the above ( region 2 o f the
per formance ) is provided by its intersection with region 1 : the upper l imit
is formed by its intersection with region 3 .

61
Region 3

3 . 4 . 14 Al though this l imited region o f a return to increasing inte r ference

with flow depth is not c learly de fined in a l l tes t series , it appears from
Figure 3 . 6 to be best exp lained as :

D I SADF = 1 . 567 - 0 . 66 7 COH . . . 3. 11

This result may have b een influenced by one or too extraneous results . so
this point wil l be reconsidered l ater in the light o f other results .

Region 4

3 . 4 . 15 Region 4 is e s s entially defined as b e ing when the channel is

sufficiently coherent for it to be treated hydraulically as a one unit : in
o ther words , stage discharge computations need not be buil t up from separate
zonal calculations . The de finition e f fectively provides its governing
equation :

D ISADF = COH 3 . 12

and this has been shown on figure 3 . 6 . The FCF tests did not extend to
depths sufficient to provide good confirmation but none of the data
indic ated that D I SADF ever exceeded COH .

Infl uence of channel side s l ope

3 . 4 . 16 Three tests were c arried out to examine the effe c t o f channel side
s lope . test series 02 , 0 8 and 1 0 , with s e r espectively 1 , 0 and 2 . (s = 1
F
in each case ) . These were a l l with the same actual flood p lain width as
we l l as the same main channel bed width , b /b = 2 . 25 /0 . 7 5 = 3 . In
F
consequence the ratio B/b varied , being respectively 4 . 2 , 4 . 0 and 4 . 4 .
Considering the shal lower range o f flows o f region 1 firs t . figure 3 . 8 shows
the Q* results . The range of the data i s narrow , showing that the
2
influence o f bank s l ope on Q* i s modest . Equation 3 . 8 has been shown on
2
Figure 3 . 8 ( fu l l lines ) for the bounding values o f B/b , 4 . 0 and 4 . 4
corresponding to tests 8 and 1 0 respective l y . However , the exper imental
data tend to have the inverse correlation : test 8 agrees better with B/b =

4 . 4 and test 1 0 with B/b = 4.0. This sugges t s that an adj ustment is
required to al low for dependency on bank s l ope .

62
3 . 4 . 17 The direct correlation o f B/b with bank s lope can be inverted ,
however , by using a width ratio based on hal f top width o f main channel , we .
Equation 3 . 8 then become s :

Q* = - 1 . 240 + 0 . 39 5 B/we + 9 . 48 H* 3 . 13
2
. • .

and this is shown by broken l ines on Figure 3 . 8 for the bounding values o f
B/we , 4 . 0 and 3 . 14 for tests 8 and 1 0 respectively ; but now i n the desired
sequence to agree with the experimenta l trend . This suggests that
expressing the width ratio in terms o f the top width o f the main channel
rather than its bed width is more appropriate .

3 . 4 . 18 Figure 3 . 9 shows D I SADF against H* for a l l f l ow regions but within

region 2 (H* broadly from 0 . 27 to 0 . 3 7 ) the p l o t s for the three side s l opes
come c lo s e together . Region 1 c learly differs but that region is best
covered by Q* • The upper bound o f region 2 also s eems to vary somewhat
2
with s e as do regions 3 and 4 when p lotted as D I SADF against H* .
Figure 3 . 10 shows the calculated values o f e oH p lotted to the same scales .
Series 02 and 10 show good agreement when shifted vertical ly by 0 . 15 in H*
value , but it appears that series 08 , for s e = 0 , rectangular main channel ,
requires a shift in H* o f , 0 . 09 rather than 0 . 1 5 for good agreement . This
is not surprising perhaps : the behaviour in terms of momentum trans fer may
indeed be changed with a s quare bank top from that applying with a more
gentl e transition between channel and f l ood p lain .

3 . 4 . 19 Looking at Figures 3 . 9 and 3 . 3 togethe r , there is a question about

how best to represent Region 3 . Although test series 0 1 and 02 showed a
pronounced backward kink in D I SADF in region 3 , this is not confirmed by
tests 10 and 8 . I s the backward kink in tests 0 1 and 02 a somewhat spurious
result of j ust a few data points being away from the real trend? The data
far tests 08 and 10 indicate not a backward kink so much as a vertical
trans ition at about D I SADF = 0 . 9 4, whi l s t tests 03 and 05 (narrow flood
p l ains ) hardly show a region 3 , going fair ly directly from region 2 to 4 at
similar value of D I SAD F . 0 . 95 5 would be an appropriate average for region 3
for tests 0 1 and 02 .

63
3 . 4 . 20 It may be more appropriate therefore to take region 3 as providing

DISADF = 0 . 95 • • • 3 . 14

the upper bound o f the region being when COH = 0 . 95 , and the l ower bound
defined by comparing with the region 2 cal culation : i f the region 2
calculation gives DISADF > 0 . 95 , then region 3 ( o r 4 ) app l ie s . The ques t ion
of which formula is best for region 3 , 3 . 14 or 3 . 18 , wil l be returned to
l ate r . There i s no evidence that e ither region 2 o r 3 i s sensitive to bank
s l ope ,

Influence of asymme try

3 . 4 . 21 Figure 3 . 1 1 shows the comparable symme tric and a symme tric tests , 02
and .06 , in the form o f Q* against H* . Equation 3 . 8 was deduced from the
2
tests with varying f l ood p l a in width and was then tuned to agree with the
tests with d i fferent s i de s l ope s , equation 3 . 12 . Provided B is s t i l l
defined as hal f the channel p lus f lood p lain width at f lood p lain e l evation ,
the s ame equation a l s o takes care o f a symme try , the relevant values o f B/w
c
being 3 . 50 for the symme tric case ( 02 ) and 2 . 25 for the a symme tric case
(06) . Q* thus proves to be a robust parameter for region 1 , being dependent
2
l inea r ly on H* and B/wc • but not sensitive to s e or the presence of one or
two flood p lains .

3 . 4 . 22 Figure 3 . 12 shows DISADF against H* . for comparison with the "COH

shift" method . The asymme tric tests series 06 is included in Figure 3 . 10 ,
and demonstrates that the cal cu lation o f COH is a f fe cted by the number o f
f lood p l a ins . However , the experimental data in Figure 3 . 12 show l ittle
difference between the two cases in r egion 2 and the conclusion i s that the
shi ft in .H*required is different when the channel is asymme tric : 0 . 10 woul d
be the appropriate value to use i n the procedure o f equation 3 . 10 .

3 . 4 . 23 With asymme try , Figure 3 . 12 suggests that region 3 may be treated as

a zone where D ISADF = constant and as b e fore 0 . 95 i s an appropr iate value .
Region 4 , ( D I SADF = COH , hence apply s ingl e channel procedures ) , was not
really entered in the a symme tric tests , as in a l l other case s .

64
3.5 Separation of main channel and flood plain e ffects

3.5.1 The detailed information obtained on velocity distribution provided a

basis for assessing the discharges over the main channel and f lood p l ain
zones separately . The sum o f thes e "velocity traverse" discharges general ly
agreed to within one or two percent of the discharge measured by the orifice
meters in the supp ly l ines , but in general the veloc ity traverse data were
then adjusted to agree with the total discharge given by the orifice meters .
Because o f the extensive data set obtained at each depth to cover its other
uses in the research programme , detail ed cros s s e ction information was
obtained at relatively few depths , typical ly eight , the lowest o f which was
in any event suspect because of rather l ow flood p lain Reynolds Numbers .
Because o f the fewer depths covered and their r e l at ively wide spacing , no
averaging o f data was permi s s ib l e .

3.5.2 Figure 3 . 1 3 shows the discharge adj ustment factors required for the
flood p lain and main channel flows calculated separately , and typifies the
data p lots showing the separate influences o f f l ow interaction on the two
zones . This figure shows the three widths o f smooth flood p lain tested .
Region 1 o f the flow behaviour identified from the overa l l analysis exten�s
up H* = 0 . 30 , 0 . 27 and 0 . 23 for series 0 1 , 02 and 03 respectively . Region 2
fol l ows up to H* = 0 . 3 9 approx , with region 3 extending towards H* = 0 . 5 or
so . Unfortunately regions 2 and 3 were not wel l covered by the separate
zonal information : there i s perhaps only one data set near the boundaries
o f the higher f l ow regions , with insufficient evidence to establish trends
within them .

3.5.3 Figure 3 . 19 c learly shows that the main channel discharge is hampered
by the flood p lain interference , to an increasing extent as flow depth rises
through r_e gion 1 . The maximum reduction in main channel flow is between 1 2
and 18% depending on B/b . Interference enhances the f lood p l ain flows by a
diminishing amount in region 1 , reducing to some 3 to 13% ( depending on B/b)
at the upper l imit o f region 1 . Thus region 1 i s characterised by
increasing e ffects on the main channel but decreasing e ffects on the f lood
p lain as the flow depth increases .

3.5.4 Through regions 2 and 3 , the separated f l ows confirm the general
behaviour estab l ished from the stage-discharge data . On the flood p lain ,
interaction e f fects appear to increase with increas ing depth up to H * = 0 . 40
65
( re gion 2 ) and decrease again to the maximum depth tested , whilst in the
main channe l the reverse happens . Looking at the separated zones a l so
demonst rates that at maximum depth there is sti l l significant interaction
between the zones , even though the total discharge is by then c l o s e to that
which woul d be predicted for the comp lete section using the perimeter
weighted friction factor .

3 . 5 .5
Figure 3 . 14 shows the non-dimensional discharge deficit , Q* ' for
2
both zones , and may be compared with the overal l data in Figure 3 . 7 . The
main channel component dominates , being an order o f magnitude greater than
the flood p lain component . Both vary l inearly with H * in region 1 ( a s did
the composite value ) and the r e levant equations are :

3 . 15

. . . 3 . 16

and so for consistency with equation 3 . 13 :

Q* = - 1 . 240 + 0 . 39 5 B/w + 1 0 . 7 6 H *
2C C

This has been added to Figure 3 . 14 and shows good agreement . Thus the
separate discharge data confirms the overal l data and shows the divis i on o f
interference e f fect between main channel and flood p lain .

3.5.6 Because of the sparse data in those regions , neither D I SADF nor Q*
2
information for the separate f l ow zones permits any deductions to be made
for regions 2 , 3 or 4 , other than to confirm that the region 1 conclusions
can not be extrapolated to greater depths of f l ow , and to indicate a
somewhat . stepped r e l at ion . Plots using COH as the vertical scale were a l s o
prapared f o r the separate data , but added nothing o f value to the p icture .
COH was useful especial ly in region 2 , 3 and 4 , but there i s insufficient
separated data to confirm the equations derived from the total discharge
information .

66
3 .5.7 The results for several main channel bank slopes are shown in Figure
3 . 1 5 as D I SADF against H * . Adj ustments required to both zones show an
influence o f bank s l ope . Test 08 had vertical banks which increased the
proport ionate e ffect on the main channel f l ow compared with tests 02 ( s c= 1 )
and 1 0 ( s c = 2 ) in region 1 , becoming less clear cut at the increased depths
of regions 2 and 3 . Discounting the data at minimum flood plain depth , the
vertical bank case diminished the influence on the flood p lains at some
depths , but the trend with bank slope is not obvious . One interesting
feature of Figure 3 . 15 is the confirmation for a l l bank s lopes of a quirk in
flood p lain behaviour at H * = 0 . 40 , about at the transition from region 2
to 3 .

3.5.8 Figure 3 . 1 6 may be compared with figure 3 . 7 : Q* versus H * for these

2
Any influence of s e on Q *
three bank s lope s . i s satis factori ly eliminated
2C
by expressing the width ratio in terms of main channel top width . Equation
3 . 16 is shown for the two bounding values o f B/w and represents the
C
region 1 data wel l .

3 .5.9 i s also an i lluminating parameter for the asymme tric case : test
Q*
2
06 had one flood p l ain but was otherwise comparable with test 0 2 . The
results from the separate zones are shown on figure 3 . 1 7 , and show that the
discharge deficit on the flood p lain is the same whether there is one or
there are two . There is some difference in main channe l influence however
( despite some extra scatter in the asymme tric data) : the main channel
discharge de ficit i s reduced for the s ingle f l ood p lain , though not halved ,
but this appears ful ly exp lained by the difference in B/wc .

3 . 5 . 10 Because there are few velocity/area data sets in regions 2 and 3 ,

it has not been possib le to establish formulae for adj usting the separate
zona l fl9w ca l culations , though there had been sufficient stage/discharge
m�asurements based on ori fice meter readings in the supp ly pipes to
e stablish adjustments to total flow in those higher regions of f low . There
may , however , be occas ions when an estimate of the flows in the s eparate
zone s is required for Regions 2 , 3 and 4 : this will be so if any sediment
calculations are required ( see Chapter 9 ) . Figures 3 . 1 3 (various width

67
ratio s ) and 3 . 1 5 (various bank s lopes ) suggest how this problem may be
tackled , if only approximately. Despite some scatter in the plotted data ,
there does not seem to be much variation in D ISADF , the adjustment to the
C
main channel f low , at depths above the Region 1 l imit . Hence it would be
reasonabl e to retain the value of D I SADF at the l imit of region 1 through
C
regions 2 and 3 . I t should be mentioned that i f the flood p lains are
appreciably rougher than the main channel , the problem doe s not arise , as
flow s tays in Region 1 for which D ISADF i s calculable .
C

3 . 5 . 11 I t might be thought that these tests with smooth channel and flood
plains were unreal istic as mos t practical s ituations involve rough
perimeters . However , they should b e viewed as simuiating equal roughnes s on
flood p l a in as in main channel : the above tests represented equal values o f
Manning ' s n in a l l e lements o f cross-section . Th e methods o f analys is used
get round any l imitation that might have arisen through using smooth , rather
than rough , surface s ; and also remove any concern over scale effects .

3 .6 Influence o f f lood plain roughnes s differing from main channel

roughness

Form of addi tional roughness used

3.6. 1 In certain of the tests at Wal l ingford the flood p lains were
roughened by surface piercing rods . The basic pattern used consisted o f a
triangular distribution o f ang l e 60° , designed to have a density o f 1 2 rods
per m2 • The system of roughening is i llustrated in P late 2 . Al l the
roughened f lood plain experiments were carried out with a f lood p lain bed
width of 2 . 25m. A sma l l number o f tests were made with reduced roughness ,
by omitting every other rod from a lternate rows . The rods were o f 25mm
diameter , _ made of timbe r but e f fectively smooth cylinders .

3 .6.2 Prel iminary tests were made to determine the actual resistance o f the
rod roughne s s . The hydraulic resistance i s the sum o f the drag o f the rods
and the friction arising from the smooth cement mortar finish of the solid
channel sur face , with a l l owance for the blockage e ffect o f the rods . The
friction aris ing from the channel boundary was assessed from the modified
smooth turbulent equation used in the analyses already des cribed .

68
The cal ibration analysis , and the resulting equations and calcul ation
procedures , are given in Appendix 2 . The method is general in the sense
that it could be app lied equa l ly wel l to the a lternative rod spacings used ,
and could take account o f different numbers of rods in a lternate rows .

3.6.3 The rod roughnes s provided much higher friction factors on the f l ood
p lain than in the unroughened main channel , a radically different case from
thos e a l ready des cr ibed . This is i llustrated in Figure 3 . 1 8 , which shows
the ratio of flood p lain to main channel friction factor for the two
conditions ( tests 0 2 and 07 : s e = s = 1 ; B/b e = 4 . 2 ) . With smooth flood
F
p lains , f / fe varies from 3 a t very l ow depths reducing progressively to 1 . 2
F
when H * = 0 . 5 . With rough f l ood p lains , the variation i s from 4 . 6
increasing with depth up to a lmost 2 0 at H * = 0 . 5 . S o a lthough in xhe
smooth case increasing depth he lps br ing the frictiona l character istics o f
the zones together , i t does the reverse i n the rough case : they become even
more disparat e .

3.6.4 It fol lows that the variation o f e oH with depth also differs
radical ly. This is i llustrated in Figure 3 . 19 , where channel coherence is
p lotted against relative depth for this geometry. As we s aw earl ier , the
smooth compound channel becomes more coherent with depth , with e oH
increas ing from 0 . 43 at shallow depths on the flood p l ain to a value of 0 . 94
at H * = 0.5. On the other hand , in the rough case it increases from 0 . 3 6 to

0 . 45 in the range H * = 0 . 06 to 0 . 2 1 and then varies around 0 . 4 1 / 0 . 42 as the

depth increases to H * = 0.5. The variation is a l ittle irregular , because
there are complex features at work: the balance between surface drag on the
boundary and the form drag of the rods ; the degree of b lockage ; variation
of Reynolds Number , etc . However , the overal l range of e oH for the rough
f l ood p lains is restr icted and no matter what depth of f l ow occurs the
channel coherence remains low .

3 . 6. 5 The roughened f lood p l ain tests were interspersed with the series o f
tests into di fferent main channel bank s lopes and so cover a sing l e f lood
p la in width , though coupled with three va lues of s e . They are li sted in
tabl e 3 . 1 together with other relevant information . In one o f these cases ,
s = 1 , s ome tests were also made with reduced roughness . The measurement
e

69
procedures were much the same as with a l l other series : in mos t cases there
were many s tage discharge measurements , but fewer velocity traverses on
which to assess the separate f l ows in the main channel and flood p l ain
zones .

Experimental resul ts

3.6.6 The first question to resolve i s whether the compound channel with
extra roughness on the flood p lains shows s imilar differentiation between
regions o f behaviour as the depth increases . To compare the smooth and
r ough flood p lains , results for tests 07 ( a and b for the two roughness
densities tested) and test 02 ( s ame geometry but smo oth flood p lains ) are
p l otted in Figure 3 . 20 in the form discharge adj ustment , D ISADF , factor
against relative depth , H* . The f low regions for test 02 are indicated but ,
bear�ng in mind that region 1 i s the zone o f increasing interference with
depth (reducing value of DISADF) , there is no evidence that the tests with
rod roughness on the f l ood p lains ever entered region 2 , let a l one 3 or 4 .
In fact the result s show progressively increasing interference e f fects up to
the maximum depths covered , reaching the very severe condition approaching
40% loss of conveyance when the depth on the f lood p lains equa l s the dept�
o f the main channel . As expected , the tests with reduced density o f rods ,
07b , show somewhat less loss o f capacity than those with ful l density . If
flood p lains are much rougher than the main channel , this clearly has a
major effect on flow .

3.6.7 Figure 3 . 2 1 shows the discharge deficit normal i s ed by bank ful l f l ow

for these comparable tests . I t is noteworthy that at the maximum depth
tested , the loss of capacity compared with the traditiona l ly recommended
text book procedures exceeds bank ful l f l ow !

3 .6.8 Q*
was found o f especial value in Region 1 in terms o f fitting
2
general equations for the smooth channel s tested , and this is p lotted in fig
3 . 22 , with the friction factor ratio s , f / fc • added . Also shown is the
F
equation for region 1 for smooth flood p l ains , equ 3 . 1 3 . I t would be
feasible to adjust the s lope o f this smooth function to provide a
satisfactory fit to the rod roughness data . In fact the equation ,

70
Q* = 1 4 . 0 H* • . . 3 . 18
2

is appropriate for run 07a . However , as in neither set of tests did the
friction factor ratio remain cons tant , there is no obvious way to correlate
the s lope of this graph with flood p lain roughnes s . Fortunately , by
considering the main channe l and flood plain effects separat e ly , this
probl em is circumvented . Four (unaveraged ) results from test 07b are also
shown : these were with reduced flood p lain roughness and lie between the
smooth and ful l roughness results , as might be expected .

3.6.9 Figure 3 . 23 g roups comparab l e sets o f smooth and rough flood p l ain
tests , 0 2 and 07a with s 1 and 1 0 and 1 1 with s
e = 2 , with Q* 2 plotted
=
e
agains t H* . Remembering that region 1 extends to an H * value of aQout 0 . 2
only with smooth flood plains but extends over the ful l depth range when
they are rough , c lear distinctions can be s een in this figure . First l y ,
a lthough interference e f fects signi ficantly increase the flood p lain
discharge when they are smooth , any increase is quite neg l igible when they
are rough . Presumably this is because the faster main channel flow is not
abl e to penetrate so readily - or exchange some of its extra momentum so
readily - when the flood p lains are very resis tant . Secondly , there is
relatively little difference between the inter ference e f fects on the main
channe l flow between the smooth and rough cases when expressed in the form
of Q * : the region 1 data fol low s imi lar trends in the two cases , a lthough
2
the greater extent o f the data from the roughened f lood p lains woul d suggest
some modification to the best fit equation .

3 . 6 . 10 Discounting any addition to the fl ood plain f l ow permits the main

channel ve locity traverse data on Figure 3 . 23 to be considered together with
the ori fice p l ate discharge measurement in Figure 3 . 22 : they both de fine
the discharge deficit in the main channe l when normal ised as Q* • Taking
2
into account also the features determined earl ier from the wide range o f
tests with smooth flood p lains , the modi fied ver sion o f equation 3 . 22 for
rough f lood plains become s :

Q* = - 1 . 240 + 0 . 3 9 5 B/w + 1 3 . 0 H* . . . 3 . 27
2 e

71
This i s shown on Figure 3 . 23 . The test 1 1 results show more variation and
may be l e s s r e l iab l e : flood plain f l ows were not measured and so the main
channe l flows could not be corrected in any way to agree with the orifice
meter readings .

3 .6. 11 Tests 07a and 07b were made with main channel s i de s lopes o f 1 in 1 .
Two other cases were a l s o t ested : 09 with vertical s ides ( s e = 0 ) and 1 1
with s e = 2 . Q* i s shown against H* for thes e three cases in fig 3 . 24 .
2
Test 09 conforms very c losely t o the smooth channel vers ion o f the function ,
e qu 3 . 17 : tests 07a and 1 1 results are very s imilar and fol l ow the
modification in equ 3 . 19 . A lthough there is some variat ion from the
equation for Q* e as a function of H* derived from the smooth f lood p l a in
2
tests , the degree o f agreement i s quite remarkabl e when one consid�rs that
tests over a range o f frict ion factor ratios from 1 . 5 to about 3 have been
extrapolated to f / fe = 20 without any maj or revision . Q* is a powerful
F 2
parameter for describ ing region 1 f l ows , being firmly correl ated with B/we
and H* but independent o f a l l other variab l e s covered in the Wal l ingford
tests .

3 . 6 . 12 A feature yet to be estab l ished i s the l imiting condition for region

1 to apply with rough f l ood p lains . The upper l imit was previous ly
considered to be where the predictive e quation for region 2 woul d give a
higher discharge than that for region 1 . The function for region 2 that
e quates D ISADF to eOH with a shift in "* ' equ 3 . 10 , woul d a l so succeed with
the rod roughened f lood p lains in the sense that it woul d yiel d l ower
discharges than the region 1 funct ion ( eOH being around 0 . 4) and so be
discarded for f l ows in reg ion 1 .

3 . 6 . 13 The overal l conclusion from the rod roughnes s experiments was that
the f l ow l ay in region 1 in a l l cases . It could be represented reasonably
we� l by the Q * e function e stab lished with smooth f lood p l a ins , but differs
2
in that Q* is neg ligib l e ( no addition to flood p lain discharge ) . We are
2F
l eft with the requirement o f a criterion for determining when to neglect
Q* : it has to be phased out with increasing friction factor ratio . This
2F
can be achieved by amending e quation 3 . 15 to :

72
 • • . 3 . 20

This reverts to equation 3 . 1 5 when f / f = 1 . 5 6 , a typical smooth f lood

F C
p l ain value towards the l imit of region 1 , but becomes neg l igib le with rod
roughne s s . Whether the assumed l inear phas ing out i s correct woul d require
more - and very accurate - data , but as the term is a minor part of the
required di s charge adj ustment , the point is o f l ittle consequence .

3 . 6 . 14 Equation 3 . 1 7 for smooth fl ood pl ain tests may be l inked to equation

3 . 1 9 for rough flood plains to give a general function for Q* ' at the same
2C
t ime taking account o f the comment in par a . 3 . 6 . 1 1 concerning rough f l ood
p l ains associated with vertical main channel banks . This is achieved by the
fo l l owing functions :

Q* = - 1 . 240 + 0 . 3 9 5B/w + G H* . . . 3 . 21
2C C

where

For s e � 1 . 0 :

• • • 3 . 22

For s e < 1.0:

• • • 3 . 23

3 . 6 . 15 There i s a n obvious gap in the avai l ab l e results from the

Wal lingford test facil ity. The tests with roughened f lood p l ains gave such
a high disparity between the friction factors on the f l ood plain and in the
main chan�el , which moreover increased with f l ow depth , that no region 2 or
3 �ata were obtained except with equal roughnes ses . There must , however , be
many practica l cases where a modest dif ference in roughness exists and so
transition to region 2 behaviour ( diminishing interference e f fe ct s ) would be
expected , and ultimately to regions 3 and 4 perhaps . Although the abi lity
of the parameters including V - V to accommodate the ful l range o f
C F
conditions tested has been demonstrated in region 1 , any comparable

73
demonstration in the higher regions e ludes us because the data are not
available from the FCF research . Other sources o f data covering modest
differences in roughness wi l l be cons idered l ater ( Chapter 5 ) .

3.7 Hydraulic des ign formul ae

3 .7 . 1 The purpose of the above analyses was to deduce formulae that could
be used to predict the f l ow in compound channels , i . e . e stabl ish their
stage/discharge functions . Alternatives were considered and progressively
developed to cope with the range of conditions tested . The results are
summar ised above : they were detailed in Technical Report number 4 , October ,
1990 (unpub l ished) . Because o f the comp lexity o f t he flow behaviour ,
involving different regions of behaviour , there i s obvious ly no s ing l e
formula to cover a l l conditions . Moreover , the preferred form o f equation
and the parameters it depends on differ from one region to the next , and a
logical method has to be estab lished for determining which f l ow region
app l ie s in any given case . In general , the equations are s impl e in form ,
with l inear variation with the governing parameters . App lication in
practice wil l probably utilise a computer program that includes the logic
for determining which region of f low applies . The fol l owing summa rises the
e quations so far deduced :

Region 1 .

3.7.2 This i s the region o f relatively sha l low depths where inter ference
e ffects increase progres s ively with depth . Q* was shown as a s imple l inear
2
function of B/b and H * in Figure 3 . 7 . When converted into a form imvo lving
the ratio of overa l l width at flood p lain l evel to main channel top width ,
it was found to be independent of s e and asymme try . I t needed some
modi fication to cope with the high f lood p l ain roughness tested , but was not
very sens itive to the friction factor ratio . The velocity area measurements
separated Q* into its f l ood p lain and main channel components , and these
2
were in turn adj usted to a general form covering both the smooth and the
roughened flood p lain results , equations 3 . 20 to 3 . 23 . This group o f
equations covers a l l the test conditions , and s o form the predictors for
region 1:

74
REGION 1 :

. • . 3 . 24

Q* = - 1 . 240 + 0 . 3 9 5 B/w + G H * . . . 3 . 25
2C C

For s e ;Ji: 1 . 0 :

G • • • 3 . 26

For s e < 1.0:

. • • . 3 . 27

I t w�l l emerge from s tudying data from other sources in Chapter 5 that these
Region 1 functions require amendment before appl ication to compound channe l s
where the main channel width t o depth ratio differs from the figure o f 1 0
that app l ied t o al l the FCF work . A modi fication i s put forward that
e f fectively rede fines Q* , and hence Q* and Q* ' to a l low for the e f fect
2 2C 2F
of main channe l aspect ratio where i t d i f fers from 1 0 .

Region 2 .

3.7.3 This i s the zone o f greater depth where the interference e ffect
diminishes again . The upper flow regions are where the channel coherence ,
COH , proved promi s ing . For Reg ion 2 , the procedure involved calculating COH
with a shift in H * and equating the discharge adj ustment factor , DISADF , to
it , and this was found to be appl icab le to all geometries tested and for
both smooth and rough flood plains . The procedure was stated in functional
form in e_qu 3 . 10 , but thi s needs some adjus tment t o take into account a
di fferent "shift" in H* to provide a satis factory fit for s e = 0 and for the
asymme tric case .

DISADF ( H* , channel geometry and roughnes s )

= COH ( [H * +shift ] , channe l geometry and roughne s s ) • . • 3 . 28
3

75
where for s � 1.0
e

shift = 0 . 05 + 0 . 0 5 N 3 . 29
F
• • •

for s < 1.0,

e

shif t = -0 . 0 1 + 0 . 05 N + 0 . 06 s e 3 . 30
F
• • •

In the above N i s the number o f flood p lains . Function 3 . 28 above has not
F
been e stab l i shed for region 2 with di fferent roughnesses on flood p l ain and
in main channel , but it i s hoped that it proves o f more general
app l icab i lity than prescribed by the FCF tests . The test s eries did not
cover asymme tric conditions with s e < 1 , so that is a l so a gap in .
confirmation o f equ 3 . 3 0

Region 3 .

3.7.4 This is a relatively narrow region o f f l ow , for which equ 3 . 1 1 was

derived , giving D I SADF as a function o f eoH . The data are somewhat

uncertain in e stablishing this as a region of increas ing inter ference

e f fects , and it was suggested that when taking a l l results together a
constant value o f DI SAD F would not be unreasonab l e . There were , however ,
these two alternatives with uncertainty over which might be the best fit :

D I SADF 1 . 5 6 7 - 0 . 66 7 eoH . . . 3. 31

or : D I SADF = 0 . 95 . . . 3 . 32

Region 4 .

3 .7 . 5
• Thi s i s the region where the coherence o f the cros s-section i s such
that it may be treated as a s ingle section , with perimeter wei ghting o f
friction factor s , when cal culating over a l l fl ow . This is equivalent to :

DI SADF eOH • . . 3 . 33

76
It should be remembere d , however , that this does not mean that interaction
e f fects are negl ig ib l e : the main channe l di scharge may be appreciably
reduced compared with the basic zonal calculation , see paragraph 3 . 5 . 10 .

Choice of region .

3 .7 .6 The logic behind the s e lection o f the appropriate predictive equation

is dependent upon the calcu lation of discharge for a l l regions in turn ,
re ferred to a s Q ' Q ' Q and Q respectively . The choice o f the
RI R2 R3 R4
appropriate region and hence appropriate total discharge proceeds as
fol lows :

Region 1 or 2 ?

•••
3 .34

Region 2 or 3 ?

3 . 35

Region 3 or 4 ?

3.7.7 The cal culation o f Q l etc begins with the basic computation o f the
R
separated main channel and f lood p l ain flows , using an appropriate
resistance formula and associated roughness va lues . It then uti lises the
equations summarised above , together with the respective definitions o f the
dimensionless groups used , Q * , Q* and DISADF , to adj ust that basic
2F 2C
calculat ion for the interference e ffects arising from compounding . It is
advisab l e to calculate for all four regions at any given depth , unless there
is firm information about which region or reg ions might apply. The logic
route g iven in the previous paragraph then s e lects the appropriate region
and corresponding evaluation of discharge . As there are two a lternative
equations for D ISADF in region 3 , both might be considered in turn . This
wil l not only give s l ightly dif ferent a lternative values for Q , it wil l
R3
a l so re sult in the boundaries o f this f l ow region changing in the two cases .
The r e lative merits o f these residual a l ternatives will be discussed again
l ater .
77
3.7.8 It wil l be appre ciated that the l ogic o f selection between the
equations for dif ferent regions does not provide any transition between
them . This accords with close examination of the individual test resul t s :
t here is l ittle evidence of a curved transition between the zonal equations ;
the switch is quite sudden .

3.7.9 I f a separate assessment of discharges and mean ve locit ies in the

main channe l and flood plain are required , they may be calcu lated in Region
1 using equations 3 . 24 and 3 . 2 5 with the de finitions o f Q*
and Q* . The
2C 2F
separate assessment o f zonal flows in the higher flow regions is not so we l l
covered by the empir ical methods above , but paragraph 3 . 5 . 10 indicates the
way forward .

3 . 7 . 10 I t remains t o check the per formance o f this set o f predictive

equ�tions by r eference back to the expe rimenta l data . The formulae were
added to the program used in data analys is , and the percentage discrepancies
between the individual results and the predicted discharge s for the observed
depths , geometries etc were assessed . These discrepancies were subj ected to
statistical analys i s , to obtain mean errors and the standard error o f
estimate . The former statistic indicates the over a l l goodne s s o f fit , a�d
the latter the variab i l ity . This variabil ity can have two components : any
imper fection in the trend of the predict ive equations and a l s o the
inevitab le experimenta l scatter due to random errors o f measurement . Tabl e
3 . 2 . summa rises these result s , with statistics for groups o f experiments as
we l l as for the total set , inc luding thos e with roughened f lood p lains .
Tab le 3 . 2 is in two sections , the first uti l i s ing equation 3 . 3 1 for region 3
and the second equation 3 . 3 2 . Within region 3 , the former provides the
better fit : mean error 0 . 1 3 % compared with - 0 . 28% taking a l l results
together ; standard error o f that estimate 0 . 7 0% compared with 0 . 9 1% . Either
wou l d be . acceptabl e in an engineering context .

3 . 7 . 11 The l east satis factory group o f results is with rod-roughened flood

p lains . They a l l l ie in reg ion 1 and a lthough the mean error of 0 . 07%
indicates high accuracy on average , the s tandard error o f 1 . 46% is the
highest of any region or any grouping . Bearing in mind that the method
essentia l ly calculates the adj ustment to a basic calculation summ ing the
individual nominal discharges on flood plain and in the channe l , the

78
adj ustment itse l f is much greater with the very rough f lood p la ins us ed in
thi s test series than with smooth flood plains . This probably explains the
increased error o f the overa l l adj ustment : it i s easier to predict
accurately when the correction to be made is s ay 10% than when it is as high
as 30 or 40% . For the smooth f lood p lains i . e . equal Manning ' s n for f lood
p l ains as for main channel , the mean errors for the various groups o f tests
are all under a third of a percent ; and the variab i lity ( st andard error o f
estimat e ) under hal f a percent . The former shows the excellence o f the set
o f predictive equations in fitting the experimenta l trends ; the latter could
hardly be bettered in terms o f consi stency of l aboratory measurement . The
complete data set is f itted a lmost exact ly on average by these predictive
methods : mean error -0 . 00 1% . The variab i l ity of 0 . 8% is highly
sat i s factory , bearing in mind that perhaps 0 . 5% arises from the experimental
observations themselves , and that the one set of equations i s appl ied to
both s imilar and very dissimi lar roughness conditions , t o asymme tric as wel l
a s symme tric cases , t o a range o f f lood plain widths and channel bank
s l opes , over a range o f flow depths covering four different regions o f
flow .

3 . 7 . 12 It wi l l be appreciated that these checks complete the circle o f

dimensiona l analys i s , experiment , empirical assessment , establ ishment o f
formulae and back-checking . Validation o f thes e predic tive equations
against independent experimental data for di fferent geome tric and roughnes s
conditions i s a necessary part o f estab l i shing their genera l ity , o r indeed
their l imitations and this is dealt with in Chapter 5 .

3 . 7 . 13 For a direct expos ition o f the des ign procedure s , readers are
referred to the "Summary and Design Method" , and also to the exampl e given
in Appendix 6 . The equations l i s ted above are not the fina l , genera l ,
ver s ions : they wil l be found in Chapter 1 0 , Section 10 . 1 .

79
 TABLE 3 . 1

Stn2!ARY OF [Link] UNDERTAKEN ON THE SERC - FCF : SERIES A t STRAIGHT

All dimensions are in metre s : see fig 2 . 2 for nomenc l ature : for the asymme tric channel . B =

hal f total width at f lood plain elevation : the numbers o f results refer to above-bank
stage-dis charge measurements .

TEST ROUGHNES S bwc hwF b B we se SF B/b B/wc No of NOTES

NUMBER Channel Flood p l . results

01 Smooth Smooth 1.5 4 . 10 0 . 75 5 . 00 0 .90 1 0 6 . 67 5 . 56 27 Symme tric

Two F P ' s
11
02 Smooth Smooth 1.5 2 . 25 0 . 75 3 . 15 0 . 90 1 1 4 . 20 3 . 50 29
(X)
0 "
03 Smooth Smooth 1.5 0 . 75 0 . 75 1 . 65 0 . 90 1 1 2 . 20 1 . 83 22

- - - -
04 Smooth 1.5 0 . 75 0 .90 1 - - - No F P ' s
( ca l ibrat ion)
05 Smooth Smooth 1.5 0 . 9 75 0 . 75 1 . 87 5 0 . 9 0 1 1 2 . 50 2 . 08 8 as R Main

06 Smooth Smooth 1.5 2 . 25 0 . 75 2 . 025 0 . 90 1 1 2 . 70 2 . 25 20 Asymme tric

One F P
07a Smooth Rough 1.5 2 . 25 0 . 75 3 . 15 0 . 90 1 1 4 . 20 3 . 50 22

07b Smooth Rough 1.5 2 . 25 0 . 75 3 . 15 0 . 90 1 1 4 . 20 3 . 50 4 Reduced rod

density
08 Smooth Smooth 1.5 2 . 25 0 . 75 3 . 00 0 . 75 0 1 4 . 00 4 . 00 25 Rectangular
main channe l
"
09 Smooth Rough 1.5 2 . 25 0 . 75 3 . 00 0 . 75 0 1 4 . 00 4 . 00 10

10 Smooth Smooth 1.5 2 . 25 0 . 75 3 . 30 1 . 05 2 1 4 . 40 3 . J.4 19

11 Smooth Rough 1.5 2 . 25 0 . 75 3 30

.• 1 . 05 2 1 4 . 40 3 . 14 16
TABLE 3 . 2

STATISTICAL ANALYS IS OF PERFORMANCE OF PREDICTIVE EQUAT IONS IN REPRESENTING

RESULTS FROM THE SERC-FCF: SERIES A, STRAIGHT

KEY : No - number o f tests in serie s ; SEE% - S tandard error of estimate ,

percent , ie r .m . s . o f variation about average discrepancy ; ME% - Mean error ,
ie average percentage discrepancy between prediction and experiment .

I - Using equation 3 . 3 1 for region 3 .

Test Flow region : -

series 1 2 3 4 Al l Group

1, 2, No 33 34 8 0 75 Varying B/b
3, 5. SEE% 0 . 225 0 . 3 60 0 . 35 4 0 0 . 3 08
ME% -0 . 4 1 2 - 0 . 044 -0 . 2 1 6 0 - 0 . 225

8 . 10. No 17 13 5 3 38 Varying s e
SEE% 0 . 13 7 0 . 39 2 1 . 047 0 . 403 0 . 46 7
ME% 0 . 122 0 . 43 7 o.721 0 . 25 7 0 . 3 19

6 No 11 3 2 1 17 Asymme tric
SEE% 0 . 3 12 0 . 205 0 . 63 6 0 . 37 0 0 . 355
ME% 0 . 07 4 - 0 . 00 1 0 . 02 2 0 . 429 0 . 07 5

7 a . 7b , No 44 0 0 0 44 Rough flood p l .
9 . 1 1 . S EE% 1 . 463 0 0 0 1 . 463
ME% 0 . 072 0 0 0 0 . 07 2

All No 105 50 15 4 174 Al l cases

SEE% 0 . 96 2 0 . 3 62 0 . 69 7 0 . 395 0 . 80 1
ME% - 0 . 07 2 0 . 08 4 0 . 1 28 0 . 3 00 - 0 . 00 1

I I - Us ing equation 3 . 3 2 for region 3 .

1 . 2. No 33 29 13 0 75 Varying B/b
3. 5. SEE% 0 . 225 0 . 27 4 1 . 126 0 0. 521
ME% -0. 412 0 . 06 8 - 0 . 984 0 -0 . 3 26

8 , 10 . No 17 11 8 2 38 Varying s e
SEE% 0 . 137 0 . 303 0 . 59 6 0 . 486 0 . 35 0
ME% 0 . 122 0 . 35 1 0 . 629 0 . 4 46 0 . 3 12

6 No 11 3 3 0 17 Asymme tric
SEE% 0 . 3 12 0 . 205 0 . 36 1 0 0 . 3 05
ME% 0 . 07 4 -0 . 00 1 0 . 03 26 0 0 . 105

7 a . 7b . No 44 0 0 0 44 Rough flood p l .
9 . 1 1 . SEE% 1 . 46 3 0 0 0 1 . 46 3
ME% 0 . 07 2 0 0 0 0 . 072

All No 1 05 43 24 2 174 Al l cases

SEE% 0 . 96 2 0 . 27 7 0 . 90 6 0 . 486 0 . 83 3
ME% - 0 . 07 2 0 . 133 - 0 . 283 0 . 446 -0 . 0 4 4

81
 Sump
\\
4 x 0.1 1 3m3/s pumps \\
\\
\\
\\
\\

'· Approach length

25m ·'· Testsection
1 5m
Tailgate
control

Fig 3.1 Layout of SERC facility at Wallingford

/
0.6 -

/ Region 4
- /
•
/ •

c� •

•
Region 3

0.4 -

� •
•

H-h _.....- •

H --- •
• Region 2
-
•
•
•
•
•
0.2 - ••
•
•

•
• Region 1
•
- •
•
•
•
•
I I I I I I I I I I I I I I
0
0.85 0.90 0.95 1 .0
DISADF
JBW/15110-91/30

Fig 3.2 Discharge adjustment factor (ratio of measu red discharge to sum
of zonal calcu lated flows) versus relative flow depth : also COH to
same scales; test series 02, averages of three.
 0.50 (),

�6
0.45 t-
0 6
0.40 t- 0
0 0
X

� �+
+
6
6
Xo +
6
0.35 t-
+
+
0.30 -
+ +
:::r::
:2 0.25 -
00
'

:::r:
-:

0.20 -
+
+ >f' O
0.15 t- -f=l-
+
+
0 0
0 . 1 0 t-
++
+
0
A 6
0.05 f-
6
'---�� ---�
� ����� ---�
1 �1--�l ---'--
l �I --� I --�
I ---'- I --�
I �I --�
0.00
0.85 0.86 0.87 0.88 0.89 0.90 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 .00

0
DISADF
Averaged
data + Test - 01 6 Test - 02 Test - 03 X Test - 05

Fig 3.3 Effect of varying flood plain width on discharge adjustment

factor - DISADF, as function of (H-h)/H = H*

0.6

limit of experiments
0.5

� 0.4

:::r::
;
:::r::

0.2

0.0
0.85 0.90 0.95 1 .00

0
Coherence - COH3
+ Test - 01 6 Test - 02 Test - 03 X Test - 05
JBW/16110-91130

Fig 3.4 Calculated co herence - COH , Wal lingford channel, various 8/b ratio
 Alternative Region 3

Performance
f"'nction

DISADF

Fig 3.5 Illustration of d ifferent regions of flow

0.95

[Link]

Region 3
DISADF 1 .567-0.667COH3
0.80
=

0
0
X
0.70 0
+ 0
0.65 +
0
0.60 ++

+
0.50
0.85

+ Test - 01 0 Test - 03 X Test - 05

JBW/1 7/10-01130

Fig 3.6 Resu lts for fou r plain widths, DISADF against COH (ru nning
averages)
 0.50
0
0
0.45
0 I:!.

x �+ + +
o
+ +!:!. 0
0.40 I:!.
I:!.

0.35 +0 I:!. X

o -n_ x
0.30
$_
..c

t
0.25

0.20

0.1 5

0.1 0
ReF ::a 4500 ·

0.05

0.00 �----�-----L--L---�--�
0.0 0.5 1 .0 1� 2� 2� 3� 3.5 4.0
Averaged OSTAR2
data + Test - 01 1:!. Test - 02 0 Test - 03

Fig 3.7 Averaged resu lts for four flood plain width s : Q*2 against H*:
Q*2 = DISDEF/[(Vc-VF)H h]
0. 50

+I:!. I:!. +
0.45 + 0
+ I:!.
l:!. o
++
I:!.
0.40 I:!. 0
+ A

0.35 + <!;>.

0.30

$_
..c

t
0.25

0.20

0.1 5

0.1 0

0.05 0
+
0.00
0.0 0.5 1 .0 1 .5 2.0 2.5 3.0 3.5 4.0
Averaged OSTAR2
data + Test - 02 A Test - OS 0 Test - 1 0 JBW/18110·91130

Fig 3.8 Averaged results for three side s lopes: Q*2 against H*
 0.50
+ 't
0 Lt
0.45 - +
c9
0.40 -
+

£:.. 0 +
�� +
t

0.35 -
£:..
£:.. 0 +
0.30 - e:.. o d-
+o
+ o
$.
.s.:
.
0.25 - + 0
�
£:..
+ 0
•

£:..
+ 0
0.20 - ++
£:.. + 0
£:.. Q-
0.1 5 -
+ 0
+ +
0.1 0 - + 0
+ +
+
0.05 -
+
· e:.. + 0
+
I I I I I I I I I I I I I I
0.00
0.85 0.86 0.87 0.88 0.89 0.90 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 .00
DISADF
Averaged
data + Test - 02 £:.. Test - 08 0 Test - 10

Fig 3.9 Averaged results for three side slopes: DISADF against H*

0.0
0.85 0.90 0.95 1 .00
Coherence - COH
+ Test - 02 X Test - 06 asymmetric £:.. Test - 08 0 Test - 1 0
JBWI19110·01130

Fig 3.1 0 Calculated cohere nce - COH, for FCF, various Se values; also
asymmetric case
 0.50
ll.
+
+
0.45

+ ll.
0.40
+
ll.
0.35 +

0.30
J:
;::::...
� 0.25
.c

0.20

0. 1 5

0.1 0

0.05
+
0.00
0.0 0.5 1 .0 1 .5 2.0 2.5 3.0 3.5 4.0
Averaged QSTAR2
data + Test - 02 ll. Test - 06

Fig 3.1 1 Asymmetric and sym metric cases (averaged data) : Q*2 against H*

0.50 .-------.
.p +
0.45 - ll.
+
ll. +
0.40 - +
+
0.35 -
ll. +
+
0.30 -

�
:2 0.25 1-
�
1

0.20 1-

0. 1 5 1-

0.1 0 1-
+
+
0.05 1- +
+
OQ I I I I I I I I I I I I I I
Q
0.85 0.86 0.87 0.88 0.89 0.90 0.91 0.92 0.93 0.94 0.95 0.96 0.97 0.98 0.99 1 .00
DISADF
Averaged
data + Test - 02 ll. Test - 06 JBW/20110-91130

Fig 3. 1 2 Asymmetrical and symmetrical cases (averaged- data) : D ISADF

against H*
 0.50

0.45

0.40 •

0.35

0.30
J:
;::::...
Region 1 Region 1
-7 0.25 XA
0 •
�

0.20 + ll 0 X .A. •

0.15 ll 0 .A. •
+ X

ll
0.10 + 0 X •

0.05 •
ll

Fig 3.1 3 Adjustments factors for separate zones : various width ratios

0.50

0.45 ·

0.40 X .A. • +

0.35

0.30 X

�
-7 0.25
�

0.20

0.1 5

0.1 0

0.05

Fig 3. 1 4 a*2 against H* for separate zones : various width ratios

 .....
+
0.50

0
X

•
0.45 ,_.

0.40 - 6 0+ -"<

0.35 -

0.30 ,- 6 0f. A ex

$_ •
-7 0.25 '- A A
�
p X

0.20 -
A ;a A • X

0.15 -
t:;. + 0 e A X

A + 0
X
0.1 0 -
• A

0.05 A • A
q. X
-

I I I I I I I I I I I I I
0.00
DISADF
0.60 0.85 0.90 0.95 1 .00 1 .05 1 .1 0 1 .1 5 1 .20 1 .25 1 .30 1 .35 1 .40 1 .45 1 .50 1 .55
�r::dual t fe Test - 01
i fe Test - 02 � fe Test � 03

Fig 3.1 5 DISADF against H* for separate ·zones : various bank slopes - se

+
0.50

0
X

•
0.45

0.40 .. .

0.35

0.30 -
:r:::
;::..,
-7 0.25 �
e

0.20

0.1 5

0.1 0

0.05

0.00

me Test - 02QSTAR20 me
-2.0 -1 .5 -1 .0 -0.5 0.0 0.5 1 .0 1 .5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
6.
Individual
data t fe Test - 01
A f • f
Test - 03 JBW/22/10-91130

Fig 3.1 6 a*2 against H* for separate zones: various ban k slopes - Se
 0.5� ,-------,--i::>.,
X +
0.45

0.40 J.. X +

0.35

0.30
:I:
;::...
-7 0.25
�

0.20
Eq 3.24
0.15 Individual
data

0.10 + me
Test - 02
X fp

0.05 i {pc Test - 06

0.00
-2.0 -1 .5 -1 .0 ·0.5 0.0 0.5 1 .0 1 .5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
QSTAR2

Figure 3.1 7 a.2 against H. for separate zone: sym metric/asym metric
comparison
0.5 r-
-....-----.

0.4

0.3
-.5
. 0
:I:
�
(f)

0.2

0.1

0
0 4 8 12 16 20 JBW/3.18/4-9211.0

Figure 3.1 8 Variation of flood plain/mai n chan nel friction factor ratio for
smooth and rod-roughened flood plai ns, B/b = 4.2, Se = sF = 1
 0.5

"0
CD
c:
CD
.c:
01
0.4 ::J
e
"0
0
a:

0.3

:J:
.

0.2

0.1

0
0 0 .2 0.4 0.6 0.8 1 .0
COH

Fig 3.1 9 Variation of channel coherence - COH, with smooth and rough
flood plai ns, 8/b = 4.2, Se = s F = 1
0.50 + A
ll t') oo 0
+ + c:
+ 0
0.45 -
ll ·c;, 0
ID
a:
0
0.40 -
++t 0
0
ll
0.35 - C\1 0
c:
0
·c;,
0
ID
0.30 - "4-+ a: CD
+ 0
:J:
:2 0.25 -
+ +
oo
;;. ++
I

0.20 -
+ CQn

0.15 -
++ 0
+ 0
�
0.10 -
+
0+
...-

c:
0
·c;, 0
0.05 ID
a:
-

q:>
I I I I I I I I I
+
. 0.00
0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1 .00
DISADF
Individual
d ata + Test - 07a ll Test - 07b 0 Test - 02 JBW/2o411 0-11 1 130

Fig 3.20 DISADF against H*: comparison between smooth and rough
flood plai ns
 0.50
A +
A +
0.45 r-
A +
A +
0.40 r-
e +
A
+
0.35 f- A
+
A
A +
0.30 - +
A +
:::c A
-- +
.:c: 0.25 - A
;-£. A
t
A +
0.20 - & +
A
A + Averaged
0. 1 5 -
+ d ata
A
+
0. 1 0 - i + Test - O?a
A A Test - 02
t"
0.05 �
J.
I I I I I I I l l
0.00
0.0 0.2 0.4 0.6 0.8 1 .0 1 .2 1 .4 1 .6 1 .8 2.0
DISDEFBF

Figure 3.21 DISD EFB F against H•: comparison between smooth and
rou h flood lains
0.50
(1 . 3 )
A
A
0.45
A X

(1 .3) A
0.40
e
A
X
0.35 A
�
Cl;)'
( 1 .4) A
A <vo.
0.30
( 1 .45) A Averaged
:::c A data
:2 0.25
;t + Test - 07a
0.20 A Test - 02

x Test - 07b
0. 1 5

0. 1 0

Added figures show sample

0.05 friction factor ratios

0.00 �--��--�---�-
0 2 3 4 5 6 7 8 9 10
QSTAR2 JBW/3.22/4-9211.0
.

Fig ure 3.22 a.2 against H.: compariso n between smooth and rough
flood p lains
 •

0
0.45

0.40 X 0 0

0.35

0.30 0

:E
:E' 0.25
t
Smooth
Eq 3. 1 6
0.20

0.1 5

0.1 0

0.05

0.00- L---�----�--�
-1 0 1 2 3 4 5 6 7 8 9
me 0 me
o
6.
Individual + me me
QSTAR2
data x fp Test - 07a .& fp Test - 1 1 • fp Test - 02 0 fp Test - 1 0

Fig 3.23 a.2 against H. for separate zones: comparison between smooth
and roug h flood plains, for chan nel side s lopes, Se of 1 an � 2

0.45

0.35

0.30

:E
:E' 0.25
t
0.20

0.10

[Link] L---�----�--L--�
0 1 2 3 4 5 6 7 8 9 10
OSTAR2
Individual JBW126110·111fJO
+ Test - 07a Test - 09 Test - 1 1

Fig 3.24 a*2 again st H.: rough flood plains with various side slopes - Se
4. SKEW CHANNE LS

4.1 The importance of momentum tran s fer with non-al igned flow .

4. 1. 1 I f the main channel is not aligned with the f lood p lains , there wi l l
b e lateral trans fer o f discharge between the main channel and flood plains ,
and o f cour s e this imp l ie s transfe r o f momentum . At the j unction between
the contracting f lood p l ain and main channe l , flow with relat ively low
momentum wi l l be added to the channe l , rather l ike that o ccurring at a
side-channe l spil lway . This flow requires some additional energy to
acce lerate towards the general velocity in the main channel . Conver s e ly at
the j unct ion between the main channel and the expanding flood p lain , flow
wi l l be l eaving the channel with excess ve locity and momentum compared to
that general ly developed on the flood p lains . This excess velocity wil l be
di s s ipated as it proceeds downstream on the f lood p lain .

4.1.2 This direct exchange of momentum is a somewhat di fferent mechanism

from the indirect exchange which o ccurs due to interfacial shear adj acent to
the banks of aligned compound channe l s , and is also asymme tric . One might
anticipate radical ly different s econdary circulations therefore . even [Link]
the ang le o f skew i s quite modest . The detailed mechanism of energy loss
wi l l also differ in the two cases . The question o f prime interest here
is the influence of skewnes s on the overal l head loss as exemp l ified by the
s t age-discharge function .

4.1.3 Compound channels where the main channel is at an ang le t o the flood
p lains formimg the val ley f loor can not be o f unlimited length : with l arge
ang les of skew the channe l soon reaches the other side of the valley f loor ,
so the main channe l mus t de f lect , perhaps curving round to the opposite
direction of skew . Several such reversals woul d be re ferred to as
[Link] . I t i s only with sma l l angles o f skew that one may treat a
l imited reach o f river or artificial drainage channel as a variant o f an
aligned straight compound channel , rather than as a meandering system where
the influence of bends is an integra l part of the overa l l system hydraulics .
Also , only sma l l ang les of skew can be studied in the l aboratory separate
from the influence of intervening bends .

82
4.2 Res earch on skew compound channel s

4.2.1 The only l arge scale programme o f res earch on skewed channe l s to
date is that carried out by the Bristol research group ( E l l iott and S e l l in ,
199 0 ) . Their main work on the FCF covered angles o f skew up to 9 . 2 °
( see P l ate 3 ) and included extensive investigation into velocity fie lds ,
shear stresse s , Reyno lds stresses and secondary currents for a range o f
angles , with a l ternative bank s lopes . Most of the tests were with smooth
channe l and f lood p lains , moulded in cement mortar . One set of tests was
with roughened f lood p lains . The roughnes s used was a variation of that
des cribed in Appendix 2 , with a pattern of 25rnrn diameter rods on the f lood
p l ains , extending through the depth of f low . The basic resistance function
for the smooth condition is a modi fied form of the smooth-turbulent
equation . The resistance o f the rod roughnes s was developed in Appendix 2 ,
and i s a combination o f the form drag due to the rods and the sur face drag
of the wetted perimeter , as used for the a l igned channe l s . The functions
der ived were sufficient ly general to apply to the particular dens ity used in
the skew channe l tests , where the total numbers of rods in successive
transverse rows ( le ft f lood p l ain plus right f lood p l ain) were 12 and 7 .
The overa l l density was 9 . 3 rods per m2 o f f lood p l ain .

4.2.2 For present purposes , only the stage dis charge results are
cons idered . The main channel was of aspect ratio 10 throughout , 1 . 5m bed
width and 0 . 15m deep . See fig 4 . 1 for cros s-section and p l an view of the
test faci lity . The s ide s l ope app l ied to the temporary wa l l s forming the
edges o f the f lood p lains was 1 in 1 throughout . The overall width of the
val ley f loor was kept constant , and was set at the maximum attainable in the
faci l ity with a skew of 9 . 2 ° . The test series had to be intersper sed with
other work in the FCF , so although ang les o f 2 . 1 , 5 . 1 and 9 . 2 ° were
considerep , and three main channe l s ide s lopes , the coverage was sel ective ,
as shown in the fo l lowing table .

83
TABLE 4 . 1 SKEW CHANNEL EXPERIMENTS IN FCF

Series No Skew se 2B Average B/b Remarks

of ang l e , FP width
tests �� m m

14 14 5.1 1 5.6 1 . 90 3 . 733 Smooth F Ps

"
15 16 9.2 1 5.6 1 . 90 3 . 73 3
"
16 11 5.1 0 5.6 2 . 05 3 . 733
"
17 9 2. 1 0 5.6 2 . 05 3 . 73 3
"
18 12 5. 1 2 5.6 1 . 75 3 . 733
19 7 5.1 1 5.6 1 . 90 3 . 73 3 Rough F Ps

4.2.3 The main point o f interest is the degree to which the ang le o f skew
affects the interaction between the main channel and f lood pl ains . The
basic case with which to compare is that o f an a l i gned compound channe l , and
a l though a B/b ratio of 3 . 7 3 3 was not one o f those tested in the main series
o f runs , it is within the r ange tested and so ful l confidence can be
expressed in the cal culation of the a l igned channel stage discharge
function , for comparison with the skew channel results . These are shown in
Figure 4 . 2 and 4 . 3 as di scharge adj ustment factors and relative discharge
de ficits agains t relative depth . D I SADF is the factor by which the the
basic computation of the sum of the main channel and f lood plain flows
( be fore a l l owing for interference e f fects ) has to be mul tip lied to obtain
the true predicted flow for a l igned systems , or the actual measurement with
skewed systems . The predictive functions used are those derived from the
main series of tests de scribed in the previous chapter , which are accurate
to better than 1% .

4.2.4 Figures 4 . 2 and 4 . 3 ( upper) show that in a l l cases with smooth f lood
plains the general trend of D I SADF for skew channels is simi lar to that for
a l igned channe l s , but the interference e f fect is somewhat enhanced , i . e .
the departure from the basic summation o f calculated flows is rather
greater . The indications are that the flow goes through the same regions as
depth increases as were observed with al igned systems . In te rms o f its
e ffect on the overal l stage di scharge function , the interference effect
increases with depth at modest relative depths ( region 1 ) , then diminishe s
at greater depths ( region 2 ) and probably goes through the transition region

84
3 be fore approaching the region 4 condition where the dis charge may be
c a l cu lated from the overall channe l s ection geometry with perimeter
weighting of friction factors . With rod roughened f lood pl ains , an ang le o f
skew o f 5 . 1 ° adds a relative ly mode st but somwha t var iab le amount to the
interference e ffect , which from the value s of DI SADF p lotted
for this case is in any case considerab l e . Incidental ly , the high ratios of
calculated friction factors for main channel and flood p lain (up to 18) mean
that flow stays in region 1 in this case , whereas with smooth flood plains
flow had apparently progres s ed through the region s .

4.2.5 One way to express the extra inter ference e ffect brought on by
skewness is to factor the dis charge de ficit that would be calculated in the
absence of skew . The individual test results in each series show this to be
a reasonabl e approximation , certainly accurate enough in the context o f
hydraul i c des i gn . The average factors b y which the dis charge de ficit has t o
be mul tipl ied t o a l low for skew t o provide a good fit t o the experiments ,
together with their s t andard deviations (bracketted) , are lis ted below :

TABLE 4. 2 INCREASED EFFECT OF INTERFERENCE DUE TO SKEWNESS

Channel s ide Factor by which the cal cul ated discharge deficits
s lope , se for an a l i gned sys tem have to be increased : average ( SD )

Angl e o f skew

0 1 . 20 ( 0 . 2 1 ) 1 . 1 2 ( 0 . 19 ) Smooth FP
11
1 1 . 54 ( 0 . 14) 1 . 69 ( 0 . 20)
"
2 1 . 41 ( 0 . 16)
1 1 . 12 ( 0 . 1 5 ) Rough FP

4.2.6 The experimental ly determined di s charge adjustment factors were

typically of the order 0 . 85 to 0 . 9 3 in these tests , whi l e the calculated
aligned channel values were correspondingly o f the order of 0 . 90 to 0 . 9 5 .
Thus a 1% tol erance on the experimenta l values , say o f discharge , woul d
appear as a tol erance o f 10% t o 20% o n the adjustment t o discharge deficit
for skewness . The figures in the above t able are therefore qui te sensitive ,
not only to any experimental tolerances but also to any impre c ision in the

85
aligned channel predictions . The conclusions are therefore somewhat
approximate , but it appears that :

the general trend of interference e f fect with depth for skew channel s is
s imi lar to that for aligned channe l s , but enhanced somewhat
- the adjustment required to the discharge deficit may be treated as a
mul tiplying factor independent o f relative depth
- there is no conclus ive evidence of dependence on channel s ide s lope : the
results are somewhat scattered
for equal roughnes s on fl ood p lain a s in main channe l , the factors are
approximate ly 1 . 2 , 1 . 4 and 1 . 7 for ang les o f skew of 2 ° , 5 ° and g o
respectively.
when the f lood p l ain roughnes s i s considerably greater than the main
channe l roughnes s ( friction factor ratios exceeding say 5 ) then the
adjustment required may be somewhat reduced but the evidence is
inconc lus ive (but see Chapter 5 , section 5 . 4) .

4. 3 Extension of design method to skew channel s

4. 3 . 1 The information o n which t o base the des ign o f skew channel s i s very
l imited , though there is some data from sma l l scale tests to be re ferred to
in Chapter 5 . These other results cover a relatively narrow channel , but
even so do not extend coverage to an adequate range of channel geometries or
roughness combinations . Any recommendat ions for a l l owing for skew in
assessing the stage/discharge function o f compound channel s are therefore
tentative . What information is available suggests the fol lowing procedure :

- fol l ow the procedures for a straight channel of the same overa l l geometry ,
assuming that the varying f l ood plain widths may be averaged and equa l ly
divided to l e ft and right
- the a l i gned channe l discharge de ficit is the di fference between the
discharge calculated using the ful l predictive procedures and that
cal culated from the simp l e addition o f main channe l and flood p lain
discharges
- for reasonably comparable roughnesses on f lood p lain and in main channe l ,
the s e de ficits should be multipl ied by the fol lowing factors

86
Angl e , o 1 2 3 4 5 6 7 8 9 10

Factor 1 . 10 1 . 20 1 . 27 1.33 1 . 40 1 . 48 1 . 56 1 . 63 1 . 70 1 . 77

- in equation form , these factors may be represented as

D I SDEF D IS DE F * ( 1 . 03 + 0 . 0 7 4� ) for 1 <� < 10°

SKEW = ALIGNED
. . .6.1

- i f the roughness conditions are such that the flood pl ain friction factor
exceeds that in the main channel by a factor o f more than 5 , the above
skewness factors may be somewhat conservative
- no information is availab l e for greater angles of skew and extrapolation
is inadvisable

87
 Moveable floodplain walls Main channel Roodplain

2b
28

A-A Cross-section of SERC Skew Channel

A+
Roodplain
) q,
Main channel

I
Roodplain
I A +- L

Plan of SERC Skew Channel

JBW/2711 o-911:30

Fig 4.1 SERC Flood Chan nel Facility, skew channel layout and geometry
 Relative discharge deficit
0.1 5 0.10 0.05 0

se = 0, smooth

11> - 5. 1 ° 0
0.4

0.3

0.2

0.1

0
0.85 0.90 0.95 1 .00
Discharge adjustment factor, DISADF

Relative discharge deficit

0.1 5 0.10 0.05 0
0.5

se ... 1 , smooth +
0
11> 5.1 ° 0 +
0
=

0.4
11> .. 9.20 +

+ 0
+ 0 �

0.3 + 0
0
0 +
0
0.2 +

.p- 0

0 +

0.1

0
0.85 0.90 0.95 1 .00
Discharge adjustment factor, DISADF
JBW/28110-91130

Fig 4.2 S kew channel resu lts, DISADF and relative disch arge deficit
compared with predictions for aligned channels
 Relative discharge deficit
0.1 5 0.1 0 0.05 0
0.5

se = 2 , smooth 0
cl) = 5.1
°
0 0
0.4 0
0

0.3 0
0
0
.

J:

0.2
0

0
/ Regloo 1

0.1

0
0.85 0.90 0.95 1 .00
Discharge adjustment factor, DISADF

Relative discharge deficit

0. 1 5 0.1 0 0.05 . 0
0.5

se = 1 , rough
cl) = 5.1
° 0
0.4
0

0
0.3

. Region 1
J:
0
0.2

0.1

0
0.60 0.65 0.70 0.75 0.60 0.85 0.90 0.95 1 .00
Discharge adjustment factor, DISADF
JBW/29110.91/30

Fig 4.3 Skew chan nel results, DISADF and relative discharge deficit
compared with pred ictions for aligned channels
5 • OTHER SOURCES OF DATA FROM COMPOUND CHANNE LS

5.1 Allowance for width/depth ratio in generalis ed predictive functions

5.1.1 One o f the l imitations of the FCF at Wal l ingford is its restrict ion
to a sing l e aspect ratio of deep channel : in other words the ratio of its
bed width to depth had the fixed value o f 1 0 for all tests . D imensional
analysi s can give no guidance on the r e l at ive merits o f channel width or
channel depth as a norma l i s ing dimension in such a test series . The depth h
was inc luded in the def inition o f Q * on the basis that i f the interaction
2
with the f lood p lain was a l ocalised phenomenon not extending across the
ful l width o f main channe l , then it was inappropriate to use the channel
width to normalise the discharge deficit ari s ing from that interfe�ence :
channe l depth woul d be more appropriate . On the other hand , i f the main
channel was narrow enough to contain the ful l width of the interactive zone ,
then it wou l d be appropriate to include bed width in the normal ised form o f
discharge deficit , as exp lained i n Appendix 1 . In the event , analys i s o f
the FCF research shows that the interference e f fect i s both appreciable and
extensive so far as the main channe l f l ow i s concerned , and evidence from
se condary current measurements shows their potential for distributing the
influence over the whol e o f the main channel , at least up to the aspect
ratio of ten covered at Wal l ingford . The "wide channe l " as sumption behind
the use o f h rather than b in Q* , though legitimate in terms of dimensional
2
analysi s and providing a succe s s ful basis for the empirical analyses as in
Chapter 3 , is very much open to question because the FCF research could not
provide any distinction between a width-based or depth-based definition o f
the norma lised dis charge deficit for Region 1 .

5.1.2 I n a l l the work reported in Chapters 3 and 4 it would have been

equa l ly yalid from the dimens ional point of view to replace main channel
depth , h , by main channe l bed width , 2b : the former had the constant value
o f 0 . 1 5m , the latter the constant value of 1 . Sm . The empirical analyses
wou l d not have been a ffected by so doing : the goodnes s of fit would have
remained unchanged . Thus the predictive formula for reg ion 1 could equal ly
wel l be quoted in terms o f a new vari able , Q * ' where :
O

. . . 5.1

88
and , as

• • • 5 .2

then equations 3 . 2 4 and 3 . 25 may be re-written as :

• • • 5.3

and

Q O = - 0 . 1 2 40 + 0 . 0395 B/w +0 . 1 G H* • • • 5.4

* C C

(G remains as in equations 3 . 26 and 3 . 27 ) .

5.1.3 At this s tage it i s not known whether the or iginal depth-based

definition of Q* wil l prove satis factory when tested against other sources
2
of data , or whether Q* O above wil l be better . There is also the possibil ity
that the influence o f main channe l aspect ratio might be intermediate
between tho se l imit ing forms of equation , and this desirable flexibil ity can
be accommodated by re-de fining Q* as ( Q - Q ) / (V - V ) Hh ( ARF) where
2 CALC ME AS C F
ARF is an "aspect ratio adj ustment factor" . For predictive purposes ,
there fore :

. . . 5.5

For the aspect ratio o f 1 0 app licable to the FCF , ARF = 1 o f cours e : but to
yield equ�tion 5 . 1 above , ARF would have to be set to aspect ratio / 1 0 .

5.2 Other s ources o f research data

Scope and l imi tations

5 .2. 1 There have been many reports o f laboratory research on two- stage
channe l s ever s ince the l arge scale studies of the resistance to flow of
meande red channe l s carried out in the mid 1 9 5 0 s at the US Wat erways
Experiment Stat ion . Compound channe l s have enj oyed increas ing popularity as
a hydr aul i cs re sear ch proj ect in the last decade . Many o f the publ ications
on the sub j e ct contain experimenta l data listings , and it was hoped to

89
validate - or further cal ibrate a s necessary - the formulae derived from the
large scale work in the FCF at Wal l ingford , by using additional data for
di f ferent s ection geometries . The other sources of information which have
been uti l ised in thi s way were pub l i shed by the following authors
( re ferences are given in Chapter 1 2 ) :

Asano , Hashimoto and Fuj ita , 1985

Kiely , 199 1
Knight , Demetriou and Hamed , 19 8 4 ; Knight and Demetriou , 1983
Myers , 1978 , 1984 , 1985
Prinos and Townsend , 1983 , 1984

US WE S , U S Waterways Experiment S tation , 1956

Wormleaton , Al len and Hadj ipanos , 1982
Ervine and Jasem , 199 1 ( skew channel s )

5 .2.2 Thes e tests i n hydraulics l aboratories covered a wide range of

conditions , as indicated in Tab le 5 . 1 . The range o f main channe l aspect
ratios was from 1 . 3 to 3 0 ; width ratio B/b up to 3 0 ; gradients from 0 . 22 to
1 . 8/ 1000 ; and f lood p l ain roughnes s e s up to three t imes the main channe l
value , in terms o f Manning ' s n . However , much o f this research was carried
out in University hydraul ic s laborator ies , and so was of relatively smal l
scale . I n order to avoid severe Reynolds number e f fects e . g . possibly
laminar flow , as wel l a s measurement difficulties , depths had to be kept
relatively high compared with widths in mos t o f thi s pub l i shed research .
There i s thus a dominance o f work at channe l aspect ratios of the order o f

two , somewhat removed from the usual range o f practical compound channel s ,
and very di f ferent from r ivers with f lood p lains . The exception to this is
the research by Asano and c o l leagues , who used a range o f channe l aspect
ratios based on an ana lysi s of rivers in Japan . This is the only work at
aspect ra�ios greater than that used in the FCF at Wal lingford . The main
channe l s of Japanese rivers show a dominant width/depth ratio between 20 and
3 0 ; and a dominant B/b ratio o f 3 to 5 . Thes e values may be representative

o f many rivers o f mode s t s ize in the UK too .

5.2.3 The criterion for choosing data for analysis was l argely the
ava i l ab i l ity in the pub l ished papers o f a set of stage-discharge resul t s .
However , information given was s e l dom as comprehensive as required .
Although a number o f researchers used artificial roughness on the flood

90
plain for at least some o f their tests , in some cases Manning ' s n values
quoted were rounded to two s ignifi cant figures , imp lying an uncertainty in
the value approaching 5 % , and it is not usual to find ful l detai l s - and in
many cases any details - of the basic calibration of the part icular form o f
roughening used ; nor tolerances o n the coe f fi cient values or indication o f
how accurately the Manning equat ion represented the resi stance function over
the range o f depths tested . In s everal cases , as indicated in the
Notes in Tab le 5 . 1 , resistance calibration s were avai lab le for
re - assessment , perhaps as a result of a persona l request for extra
informat ion . Considerab l e importance was attached to this matter , because
without a good knowledge of the basic res i stance functions for both main
channel and f lood p lains , the comparison of the s tage discharge data under
compound flow with the predictive methods is prob lemati cal . Both the As ano
and Prinos data sources suffer from l ack o f calibration detai l s .

5 .2.4 Another typical , though less s ignificant , gap in information is the

f luid viscos ity , which is temperature dependent . Much of the research was
in smooth sur faced flumes , so it i s perhaps surprising that the analyses in
the pub lished papers are dominated by the use of the Manning formula . This
is mos t appropriate for rough turbulent conditions , when hydraulic
resistance is independent o f viscos ity and Reynolds Number . The relevance
of viscosity in smooth turbulent flow not a lways being appreciated ,
temperature measurements were not usual ly ava ilable . In some cases ,
however , Author s were able to respond to a request for temperatures observed
at the t ime of the stage dis charge measurements , and from these the
viscosity for individua l tests could b e calculate d . I t i s shown in Appendix
2 that the Manning equation can represent smooth conditions in a laboratory
sized system at a given gradient and viscosity reasonab ly we l l , though
Manning ' s n will then depend on hydraulic gradient and viscos i ty . I n the

re-analy�is o f the research on smooth channe ls , thought was given to the

best equation to represent the particular form o f cons truction : in s ome
cases the smooth- turbul ent equation was used i . e . Co lebrook-White with k
s
= 0 ; in other cases appropr iate k values were ob tained from the calibration
s
data , perhaps with the wide channel convers ion of the trans ition formula ;
and in some cases there was l ittle scope for other than the assumption of
Manning with the researcher ' s own coefficient values . On the whole , the
results were not found to be sensitive to the cho i ce · of resis tance func tion ,
provided it accurately represented the basic frict ional resistance under

91
s imple ( non-compound) open-channel conditions over the depth range o f
intere st .

5.2.5 There have been many research proj ects on compound channel s , and not
a l l data from past research have been reviewed . Only if the pub l i shed paper
contained s tage di s charge tables and also provided other basic information
required , or there appeared reas onab le prospect of obtaining i t , was a
particular s eries o f tests considered . In one or two cases , the research
was at too sma l l s c a l e or at geometries too far removed from practical
conditions to be cons idered worth fo l l owing up . Only if more data becomes
available at main channel aspect ratios o f s ay 4 or more , including
roughened f lood p l ains , supported by ful l cal ibration information , would i t
be worth re-opening this study .

Het�odology

5.2.6 The analys i s o f these other sets of data proceeded from the
c a l culation o f the basic zonal dis charges , us ing the best fric tion formula
and coeffic ient value s ava i l ible from the cal ibration dat a , then adj usting
the sum of tho se dis charges for interaction using the predictive function�
derived in Chapter 3 from the FCF research . The adjustments were made as i f
the flow could b e i n any o f the four defined flow regions , using the
equations appropriate to each region in turn . Then the logic .of choice
between regions as given in e quations 3 � 34 to 3 . 36 was fo llowed , so
identi fying a l s o the correct predicted flow from the four alternatives .
These predi c tions were compared with the recorded observations , test by
test , and the dif ferences determined . Thes e differences were fina l ly
expres sed as percentages o f the predicted flow . For any given test series -
usua l ly a particular geometry and r oughnes s comb ination from one data s ource
- the grqup of results could be expressed statistically , as mean
di�crepancies and their standard deviations , again as percentages . The mean
discrepancy o f cours e shows the overall goodnes s of fit o f the predictive

metho d . The s tandard deviation has two components : the random scatter in
the data due to exper imenta l tolerances and also any dif ference in the
trends o f the theory c ompared with the data . Usual ly the as sumption is made
that a confidence band at 9 5 % leve l wil l be twice the s tandard devi ation .
Thi s assume s a normal distribution o f error s , which is probably rea sonab l e
for the experiment al tolerances but may not represent any di f ference in

92
trends proper ly . In summarising the resul ts o f the re-analys is , the mean
errors and standard deviation were assessed : both are important , the lat ter
especially so as it inc ludes any di f ference in trends between theory and
experiment . Any error in the bas i c knowledge of resistance woul d appear
primarily in the mean error .

Prel iminary analysis

5.2.7 As an i l lustration o f the task of achieving agreement between any

method o f prediction and the range of experimental results , Figure 5 . 1 shows
data from several s our ces where the main channel and flood pl ain were o f
equal roughnes s , in the form D I SADF against relat ive depth , H * . The
s elected tests are for very s imilar geomet r ies , with general ly smooth main
channe l s and flood pl ains , and they are shown in comparison with the large
scale data from the SERC-FCF , at about the same width ratio ( B/b of between
4 and 5 ) . The sma l l scale work shown was a l l at main channe l width/depth
ratios of about 2 to 3 , compared with ten at Wal l ingford . The FCF resul ts
p lotted are the running averages o f three : because o f this , and the large
number of results and the ir accuracy , they show very little s catter and
dist inct trends through flow regions 1 , 2 and 3 . The other data are more
scattered , over a band of the order o f 5% in D ISADF . There is broad
simi larity of trend of reducing inter ference e ffects as depths increas e ,
with s ome evidence o f a kink between H * = 0 . 3 6 and 0 . 44 that would be
consist ent with a Region 3 transi tion zone . Most results appear to be in
region 2 : reduc ing interference with depth . There i s no evidence of Region
1 ( increas ing interaction e f fects with depth at sha llow depths ) in the small
scale work , but this may result from the exc lusion o f data a t sha l low f l ood
p lain depths from the recorded information , because of Reynolds Number
l imitat ions .

5.2.8 Not a l l the Myers resul t s for different gradients were inc luded in
Figure 5 . 1 because they would overload the figure . As there are very many
results from this s ource , i t was decided to consol idate them for the range

o f gradients used into one set , to place them in depth order , and then take
running averages o f three . The resulting plot is in Figure 5 . 2 , and
although there remains scatter over a band o f several percent , the
indi cation is o f per formanc e proceeding through region 2 via region 3 into
region 4 . Re sults for roughened flood pl ains are shown in Figure 5 . 3 . The

93
Pr inos and Worml eaton cases are for a flood p l a in Manning ' s n o f 0 . 17 ,
whi lst the FCF results are for surface pierc ing rods , which result in
increasing n values with depth . The trends o f the two sets o f data are
diametrica l ly oppos ite , with the rod roughness giving progre ssively
increasing interference e f fects , but the other two data sets showing
diminishing interference with depth . So al though the l arge scale data came
ent irely in Region 1 , it appears that none o f these smal ler scale rough
f lood pl ain results were in that region . The Prinos and Wormleaton results
shown are , however , consistent with one another .

5.2.9 Figures 5 . 4 and 5 . 5 show the predicted interference e f fect as D I SADF

using the formulae derived in Chapter 3 . On Figure 5 . 4 , a lternat ive values
for the aspect ratio factor , ARF , discus s ed ear l ier are shown : unity and , in
a broken l ine , channel aspect ratio / 10 ( a l s o unity for the FCF of course ) .
Alt�rnative functions for Region 3 are shown , the two origina l s in terms o f
COH and us ing a constant D ISADF value o f 0 . 9 5 , wit� a third related to H * .
This last was intended to give a good representaion of the group of results
in Figure 5 . 1 that are probably straddl ing the trans it ion o f Region 3 .
Comparing with the selected data in Figure 5 . 1 for equal main channe l and
f lood p la in roughne s s , c l early the predictive equations shown in Figure 5 4 . •

work quite wel l in that they give results through the middle of the s catter
band , but in order to avoid region 1 conditions a value o f ARF above aspect
ratio/ 10 seems des irab l e . In fact the original value o f 1 would achieve
thi� , but is not proven because there is no data in this region , nor indeed
any evidence that Region 1 type of flow occurred in these sma l l scale rather
narrow channels . The FCF results are wel l forecast , not surpisingly because
these data were inc luded in the set upon which the predictive functions were
based . For this particular Wal l ingford geometry , with s e 0 , Region 3
=

seems to indicate a constant D I �ADF ( the other values o f s e tended to

support the alternative function depending on COH in Region 3 ) . Note that
the Region 2 predictions , where the equation is based on channel coherence ,
nicely fo l l ow the di fferent trends for the two aspect ratios covered .

5 � 2 . 10 . F igure 5 . 5 shows s imi l ar predictions for rough f lood plains , using

the rod roughnes s functions in the one case and the Manning formula in the
other . Surpris ing ly good agreement with the very disparate trends of data
in Figure 5 . 3 is obtained , with no adjustment to the FCF based functions .
It should be remembered that f * ' f / f ' for the rod roughnes s increases
F C

94
progress ively with depth o f flow . This i s not so with boundary roughness ,
which is why the trends di ffer radically. The ab i l ity o f the predictive
functions to cover a quite different aspect ratio and radically di fferent
type of r oughening is notab le . It is not proven that ARF = 1 , however , as a
range of values above aspect ratio / 1 0 would suffice in terms o f this' sample
o f data . To provide a good fit , ARF merely has to be high enough to avoid
any Region 1 predictions over the range o f depths for whi ch the sma l l scale
data are available .

5 . 2. 1 1 At this stage it might be thought that the predi ctive functions

derived in Chapter 3 are ful ly va lidate d ; that no modi fi cation is required
for aspect ratios di ffering from 1 0 . The equations · adapt well in this
sample of results to dif ferent geometr ies , and even accommodate · boundary
roughne s s though originally based on flood p lains roughened by surface
piercing rods . Detailed analys is of the ful l range of data avai lab l e show
that not a l l results fall so readi ly into this neat pattern of agreement ,
however .

Difficul ties

5 . 2 . 12 As il lus tration o f the prob lem posed by some ot the publ i shed data ,
the test resul ts from Prinos and Townsend for 3 0 5mm and 406mm channel widths
are shown on Figure 5 . 6 , for the four flood pl ain roughnesses they
considered . The only difference between these test conditions was channe l
width , changing the aspect ratio from 3 to 4 . At aspect rat io 3 , there was
cons iderab le interference e f fect , with up to 3 0% loss of conveyance . There
was s trong dependence on roughnes s too . Yet with aspect ratio 4 , the
interference effect has dropped to under 4% , with a few pos itive effects
rather than the expected reduction of discharge - and no ident i fiab l e
influence from f l ood p lain roughnes s . The natura l conc lus ion. was that no
predictive method could reas onab ly be expected to reconcile such resul ts .
This prob lem was resolved to some extent on re ferring the initial analysi s
o f thes e results to the originator of the data , who then advi sed that there
was a discrepancy between his Thes i s and the s tage-di s charge data pub l i shed
in 1 9 83 . The experiments had been repeated with extra care in sett ing
uniform flow , and so a revis ed and radic a l ly di fferent set of results was
provided . These revised data have been used in the .final analysis , of
course . The Prinos and Townsend procedures and results wi l l be

95
re ferred to again later . In essence , they confirmed that ARF should not be
less than aspect rat io/ 1 0 , preferab ly rather higher than that . However , the
agreement between the predictive method and these experiments became less
and less satis factory as the roughness o f the f lood p lains was increased .
The basic roughnes s of the meshes used i s open to que stion , however : no
calibration data were avai lable but it is known from other work with mesh
roughne s s that the Manning equat ion with constant coefficient does not
accurately represent the hydraulic resistance over a range of depths .

Asano e t al ( 1985)

5 . 2 . 13 Turning attention now to the widest aspect ratios covered in any o f

the pub l i shed res earch , of about 3 0 , as studied b y Asano and coll eague s :
the problem with the se results is that the flow was non-uni form in almost
a l l the tests reported . There s eems to have been l i ttle attempt to ensure
uni formity , and the common s i tuation was with water sur face s l ope exceeding
the channe l s l ope , by up to 50% for the deepest cases , ( in one case 70% ! ) .

To avoid including test results where a large proportion o f the flow was
being ' s queezed out ' of the f lood plain , the re-ana lys is exc luded water
surface s l opes more than 20% in excess o f the channel gradient , but even
with this l imitation the pattern o f secondary c ircu lations and shear zone
ef fects must have been- affected .

5 . 2 . 14 Several different as sumptions were made in carrying out the analysis

of the Asano et al data . A range of aspe ct ratio adj ustment factors were
considered , with ARF going from the basic value of unity to aspect ratio/ 10 ,
i.e. 3. As wel l as us ing the Authors ' Manning ' s n values , the wide channel
resistance function was appl ied with the in-bank tests to assess suitab l e k
s
value s . Al though the construction method was the s ame throughout the
serie s , the Authors quote a range of n value s from 0 . 009 1 to 0 . 0 1 1 4 , so the
ac�ual resistance is rather quest ionab le . However , the conclus ion was that ,
t aking the overall average , none o f the a lternatives tr ied proves any

advantage over the use of the Manning formula with the Authors ' quoted
coeffic ients for each test series , coup l ed with ARF = 1 . Some individua l
s erie s tended to support ARF = 3 , i . e . aspect ratio/ 1 0 . Many of the test
series have residual overa l l discrepanc ies between prediction and
observation , but this may we l l arise because of unc ertainty over the bas ic -

96
and s trangely variable - main channe l roughness coe fficients recommended ,
which coul d give sys tematic errors . The average agreement overall i s within
3% with a scatter of 2 . 3% . so within the t o lerance that might reasonably be
expected , e speci a l ly bearing in mind the pos s ible influence of non-uni form
flow in many tests . Mos t of these results come into Region 1 . so they are
indeed providing a test of the de finition of Q * in this zone . Being the
2
only avai lab l e resul ts at aspect ratio > 1 0 . they are potential ly very
valuab l e ; but as flow was non-uni form and no deta iled calibration data was
pub l ished , this potent ial value has not been ful ly rea l i s ed .

US WES ( 1 956)

5 . 2 . 15 As shown in Tabl e 3 . 1 , the tes t s carried out at the US Waterways

Experiment S tation were at large scale , though rather few o f them concerned
straight channe l s : the main thrust of the research concerned meandering
channe l s ( s ee Chapter 8 ) . The necess ary minimum calibrat ion data were
avai lable in the published report for both the main channel and the flood
plains , both smooth and with mesh roughening , though it was found necessary
to reanalyse these and e stablish the variation o f Manning ' s n with depth o f
f low t o avoid too much approximation i n determining the basic resistance s .
A 1 ft wide channe l was tested with a 3 0 ft total width across the f lood
p l ains . but this gave no usefu l results , as the origina l res earchers
themse lves concluded . This is becaus e the main channe l carried such a sma l l
proportion o f the total discharge , the inter ference e f fect proved
unmeasurable in terms of stage discharge . This led to the testing o f a 2ft
wide channel , with width/depth ratio o f 4 and re l ative width , B/b , = 8. In
terms o f large s ca le , the form o f roughnes s used (pane l s o f wire mesh) , and
geometry these tests are of considerable interest , coming much c loser to
reality in terms o f scale and geometry than much of the l aboratory work .

5 . 2 . 16 Serious errors o f prediction were obtained with ARF = 1 as imp l icit

in the original set o f predictive formulae . Region 1 was being properly
tested , in that results in that region were included in a l l 9 runs ( 3 at
each o f 3 f l ood p lain roughnes s e s ) and so the rather firm conc lusion could
be drawn from these tests that ARF should equal aspect ratio / 1 0 , i . e . that
the definit ion of Q* should be modified to include channe l width rather
2
than depth . For the 2 ft channe l , this gave a mean error o f 5% with 5 %
variabi l i ty using the calibration values o f k i n the wide channe l
S

97
trans ition function as the basic resistance . Some adjus tment to those
value s reduced average error to 0 . 7% and variab i lity to under 4% . I t was
this early but positive finding from the US WES res earch that led to much
e f fort in seeking to optimise ARF as a funct i on o f B/b and 2b/h , though in
the final analys is there does not appear to be any complex funct ion
invo lved .

Kiely (1991)

5 . 2 . 17 Information about very recent res earch was obtained privately from

Dr Kie ly , of Univers ity Col lege , Cork , E ire . As with the US WES , his main
interest was in meandered channe l s , but he had also resea rched a part icular

straight channe l geometry , with aspect ratio 3 . 7 and B/b = 6 , so in terms o f

geometry coming c loser to the U S WES geometry than any other . He tes ted
wit� both smooth and roughened f lood plains and provided good calibration
data for both conditions . There were five tests of compound f low for smooth
and for rough flood p lains . The conclusion here was again that the
a l l owance for the inf lence o f the main channe l width/depth ratio on the
interaction e f fects in Region 1 was best achieved with ARF set at aspect
ratio/ 1 0 , though with the smooth flood pl ains any higher value was equa l ly
va lid as no Region 1 results remained in the data set for ARF > aspect
ratio/ 1 0 . The fiflding that ARF = aspect ratio/ 10 was fairly positive with
rough flood p l ains .

Knigh t e t a l ( 1984)

5 . 2 . 18 The research described by Knight , Demetriou and Hamed ( 19 8 4 ) on

smooth compound channel s is characterised by particular care to achieve
uni form flow , and so the doubts and po ssible cr iticism o f s ome of the other
research pn that s core do not apply here . Also , calibrat ion data for the
flume were avai lable from within bank tests . They tes ted 3 width rat ios ,
B/b = 2 , 3 and 4 , but the one aspect ratio o f 2 : a typical narrow l aboratory
set-up , rather far from the geometry of two-stage channel s of hydraul i c
engineering and a l luvial rivers . 1 8 pairs o f s tage dis charge data are
ava i lable , with smooth flood plains and main channe l . (Knight and Hamed
( 19 8 4 ) a l so reported an extension of the test series to cover roughened
flood plains , but the data were not pub li shed in de tail and have not been
analysed here ) . Their results for B/b = 4 are included in Figure 5 . 1 , and

98
compar ing with Figure 5 . 4 they are seen to agree we l l with the prediction .
The result at minimum depth l ie s exactly on the predict ion o f Region 1 in
Figure 5 . 4 for ARF = aspect ratio/ 10 - though one point can hardly be said
to provide proo f . In fact values of ARF from 0 to 1 we re cons idered , and
for the B/b ratio 4 the prediction was exa c t ly correct on average , with 2%
variabi lity about that mean , for ARF = 0 . 2 . The best f it for B/b = 3 was
when ARF = 0 . 4 , but at B/b = 2 , ARF = 0 was best - in other words , the
effect o f any interference was neg l igib l e in terms of s tage dis charge
function . Re su lts were thus s omewhat variabl e in respect of the a l l owance
for channel aspect ratio , though good accuracy was achievabl e be treating
ARF as a variable to be optimi s ed .

Myers ( 1984)

5 . 2 . 19 Myers ' pre-85 research concerned smooth conditions in a relative ly

sma l l l aboratory f lume . Some tests were with only one flood plain ; others
were s ymme tric with two f lood p l a ins , and a lthough a l l the data were
re-analysed , only the symme tric cases wi l l be referred to here . These were
with aspect ratio 2 and B/b ratios of 3 . 2 and 4 . 7 ( see Tab le 5 . 1 for det a i l s
o f test geometries ) . One feature of Myers ' research programme was the
coverage o f a range o f s lopes , so that in a l l 1 5 3 pairs o f stage di s charge
result s were obtained , over 40% o f the tota l data set from other laboratory
research . At the minimum flume s l ope o f 0 . 22/ 1000 , one might ant i c ipate
cons iderab le di f ficulty in measuring the water sur face gradient and
therefo re in achieving uni form f low . Typical t o l e rances on setting and
measuring hydraulic gradient , and hence the equivalent tolerance on
discharge , wi l l be referred to again later . Based on the ful l f lume length ,
there may in this case have been a potential tolerance o f 10% when expressed
as equivalent discharge at minimum gradient , dropping to perhaps 2% at
maximum gradient . O f cours e , this source o f random experimental error doe s
not apply only to the research b y Myers : i t app l ies to most of the others as
wel l , where the flume length or gauging length wa s restricted by the
available fac i l ities , and espe cially where gradients under 1/ 1000 were
app l ied .

5 . 2 . 20 In re-analysing Myer s ' data , a range of pos s ib l e adj ustments for

width/depth ratio were tested , with ARF from 0 . 2 to 1 . As might have been
anticipated from the samp le of results examined earlier in Figures 5 . 1 and

99
5 . 2 , there were no Region 1 flows within the range of depths studied , so
that the only conc lus ion to be dr aw.m about opt imum ARF values from this
re search is that ARF could take any value from aspect r atio/ 1 0 to unity with
little effect on the stat istics of goodnes s of fit between prediction and
experiment . With B/b = 4 . 7 , 2b/h = 2 , the mean disc repancy from the
predicted flows was 1 . 6% , with variab i l iy 2 . 9% ; with B/b = 3 . 2 , 2b/h = 2 ,
mean discrepancy under 1 % , var iability under 6% . Thus the predi ctive
funct ions may be regarded as validated to much the same order o f accuracy
that could app ly to the data , though as Region 1 was not covered by the
res earch results , the val idation is confined to the higher Regions of
behaviour .

Wormleaton ( 1982)

5 . 2 . 21 The research by Wormleaton and c o l l eagues detailed in the 1982 paper

was for one channe l geometry but it covered a wide range of flume s lopes and
a l s o 3 sets o f art i ficial roughnes s in addition to the series with smooth
flood p l ains . The art ificial roughness took the form of 10mm dia
hemispheres , at different dens ities , and this form of roughnes s was
calibr ated , i . e . its bas i c resistance func tion determined , in separate flume
studies . Al though not included in the origina l pub lication , this
cal ibration data was made ava i l ab le for re-analys i s in the present s tudy , so
that an independent check was made of the roughness coe fficients for each
dens ity , and o f the ac curacy of fit o f the Manning formula to the
observations . Dr Wormleaton , in a personal c ommunicat i on , commented a s
fo l l ows o n the tolerances app l icable t o laboratory s ca l e reseach : "As suming

the multi-manometer s cale can be read to O . Smm ac curacy , then the

corresponding depth accuracy wi l l be 0 . 2mm. Although the scale of the
vernier point gauge can be read eas i ly to 0 . 1mm , its a ccuracy in setting to
a moving water surface is probably no better than 0 . 2mm . C learly over a
2 . 5m test length , this can lead to an error in the water surface s lope o f
0 . 0 00 1 6 . Howeve r , errors o f this magnitude are unlikely since s ix tappings
are used over the 2 . 5m l ength • • • As a general p o int , it i s c lear that in
any sma l l - scale laboratory work , asses sment of hydraul i c gradient is far
more prone to error than measurement o f depth or dis charge . " I f the error in
head di f ference was , say , 0 . 00005 , this represents a little over 1 0% for the

maj ority of the tests in this research programme , which would feed back to a
5% or so error in bas ic discharge assessment , and this is probably

1 00
typical of the tolerances to be expected in sma l l scale laboratory tests in
this fie l d .

5 . 2 . 22 I n these tests the comp ound channe l was formed from precast concre te
blocks , p l aced either s ide of the base of a Perspex flume . The roughnes s o f
the concrete was we l l e stab l ished from the calibration data . The main
channel wal l s were then l ined with Perspex so as to have the same surface
roughne ss as the bed .

5 . 2 . 23 Several values for the factor a l l owing for any influence o f aspect
ratio on region 1 flows were tested agains t the available stage discharge
results . The aspect ratio o f this channel was 2 . 4 , - and ARF values both
s ides o f 0 . 24 were considered , and the detailed s tatistics o f the goodne s s
o f fit were derived . The data for a fl ood plain Manning ' s n o f 0 . 0 1 7 are
inc luded in the plot of Figure 5 . 1 : with data for a common gradient l inked
together . The more comprehensive o f these sets , compared with the other
data in Figure 5 . 1 , suggests that there may be a positive systematic
disc repancy , but there is no evidence of region 1 flow . This i s indeed what
emerges from the s tatistics o f the comparison with the predictive method:
genera l ly speaking , the Wormleaton et al results are we l l represented by �
high enough value o f ARF to ensure avoiding any Region 1 predictions over
the ranges of depths studied . There i s no evidence from these results o f
requiring any change to ARF = 1 implicit i n the bas ic method , but again no
proof that ARF = 1 .

Prinos and Townsend

5 . 2 . 24 The Prinos and Townsend results have already been referred to , and
because there are 80 pairs o f s tage discharge results over a range o f
differen� aspect ratios and re lative widths , with 4 dif ferent flood p lain
ro�ghnesses s , they are an important source of informat ion . Bas ic
cal ibrations were obvious ly carried out to assess the wire-mesh roughness
uti lised , but the details were not included in the publ ished paper , and

could not be provided by the Authors , so their quoted n values have not been
subj ect to any checking . Detai led stage di s charge information was
publ ished , that in 1983 for the 406 and 508mm channe l s , and separately in
1 9 8 4 for the 203 and 305mm channe l s , but the l at ter set of data was
apparently unreliable and the first Author provided a set of corrected

101
stage /discharge data . The Authors also examined procedures for asses sing
the s tage discharge function , using the concept of inter facia l shear stress
to produce predictive methods .

5 . 2 . 25 Broadly speaking , the smoother of the f l ood pl ain conditions tested

by Prinos and Townsend show interference e f fects reas onably cons i stent with
ARF = aspect ratio/ 1 0 , though a somewhat higher value would improve
agreement in many o f the series of tests . Less satis factory results were
obtained from the data for the two roughes t flood p lain conditions , however :
there was greater variabil ity as we l l as greater res idual errors , so that no
firm conc lusions could be drawn .

5.3. Summary o f information from other laborato ry research

5 . 3 . _1 The re-:analys is of data from other sources contains too much detail
to include here . I t was g iven in greater detail in Technical Report number
5 o f the proj ect , which is unpub l ished but was avai lab l e to most of the
research workers directly concerned for their comments . Some further
detai l s , in the form of summaries of the statistical analyses , wi l l be found
in Appendix 7 .

Region 1 : influence of width/depth ra tio of main channel

5.3.2 The information relevant to any modificat ion of the original

predictive functions to accommodate a ful l range of main channel width/depth
ratios is summarised on Figure 5 . 7 . This provides the evidence in terms of
opt imum , or acceptable , ARF values in the matrix o f aspect ratio , 2b/h ,· and
relative width , B/b . Coverage o f the field is r ather sparse and irregular ,
with so much o f the avai lab le data being for rather narrow laboratory scale
systems .

5.3.3 The original formulation of Q*2 used in the predictor for Region 1
was der ived on the as sumption that the channe l was wide in relat ion to the
zone of interference from the flood plain , so it would not be surpris ing to
find the obvious ly narrow channels used in many of the Univers ity studies
departing from that assumpt ion , and approaching a width rather than depth
based discharge deficit , i . e . tending towards ARF = 2b / 1 0h . Conclusions
from the US WES and Kiely research p l otted in fig 5 . 7 show this trend , as

102
does one o f the geometries tested by Knight . Myers results are not
incons istent with this evaluation of ARF , and this app lies also t o the
smoother of the Prinos and Townsend flood p lain conditions . It was
original l y hoped that Figure 5 . 7 would lead to a contouring of suitab l e ARF
values but the coverage i s spars e and no realistic contouring could
reconci l e a l l the results from narrow channel s . Thes e include some showing
a lmost no interference e ffect , which is consistent with ARF = 0 , e f fectively
making region 1 dominat e the picture but with the discharge adjustment
factor set to 1 ; others show no evidence of Region 1 flow ( increasing
interaction e ffect s with f low depth) and this would be consis tent with
confining Region 1 to depths below those for which data is available , by
setting ARF = 1. Tes t series sugges t ing either ARF = 0 or
ARF > > 2b/ 1 0h are rather negative , in that they are not real ly providing
information on how Region 1 f lows may be in fluenced by main channe l aspect
ratio ; though they may a t the same time be quite positive in their
confirmation of the predictions for Regions 2 , 3 and 4 .

5.3.4 There is no reason to doubt the experimental ski l l s and c ar e fu l

measurements of the researchers whose results may be out o f l ine with others
or which do not fit we l l the empirical functions derived from the large
flood channel faci l ity at Wal l ingfor d . It i s known from turbulence
model l ing ( a s descr ibed in Chapter 6 ) that the momentum trans fer is just as
much a result of s econdary circulations a s it i s of l ateral variation of
mean depth velocity giving rise t o additional shear across the inter facial
p l ane . Might some of these research proj ects have been conducted in
fac i l ities that were not long enough to generate a representative system of
s econdary c ir culations as would be expected in very l ong channe l s ? Might
thes e secondary circu l ations a l s o be sensitive to inlet conditions , or to
any non-uni formity of f low? And is Region 1 particular ly sensitive to the
momentum exchange via such secondary circul ations?

5.3.5 It is worth stressing , however , that on the whol e the predictive

method has proved robust in that it trans ferred we l l to other very different
geometries and roughnesses . It i s only region 1 that obvious ly required
modi fication to fit some series o f tests to al low for the influence of main
channe l width/depth ratio , and this r equirement had been anticipated for
narrow systems . In this context , it is interest ing to look again at some o f
the pub l ished information at sha l low f lood p l ain depths . Some research

103
suggests that a radical step would occur in the stage discharge function as
the flow went above bank . Prinos and Townsend ( 1984) give a figure that
shows thei r own results for aspect ratios of 2 and 3 with comparab le data
from Worml eaton ' s and Myers ' work . This is reproduced as Figure 5 . 8 , and i f
one extrapo lates these results towards H * = 0 ( a s suming no Region 1 f l ows in
e f fect ) , then at just above bank ful l , the dis charge could drop to 7 0% ­
even as low as 40% - o f i ts bank ful l value . Certainly the tests in the FCF
at Wal l ing ford did not show anything l ike that degree of obstruction to
discharge soon after the flow went over bank ; and there is no evidence from
the field either that the "kink" in s tage dis charge function at bank ful l i s
ever o f such a magnitude ( see a l so section 5 . 5 fol lowing ) . I t is for this
reason that pre ference is given in the final interpretation to those results
which suggest a re latively gent le increase of interference e ffects . as to
flow goes over-bank as g iven by ARF = aspect ratio/ 1 0 .

5.3.6 What about wider channe l s than the FCF with its aspect ratio o f 10?
One would expect that the wide channel as sumption for the definit ion of Q*
2
in Region 1 would become increasing ly valid as the aspect ratio increases .
Unfortunately there is only the Asano e t al data at aspect ratios above 10 ,
and this does not provide the degree o f rel iab i l ity that one woul d hope for
(non-uni form f l ow ; absence of c a l ibration detai l s ; uncertain basic
res istance ) . The evidence from that research , however , is that its aspect
ratio might be above the lower l imit for being e f fectively wide , in that the
results divide between those best represented by ARF = aspect r atio/ 1 0 = 3
and a l ower value , al though i t seems c lear that the FCF aspect ratio of ten
was below the l imiting value . The provisional conclus ion is that we might

as sume a channel aspect ratio o f 20 as the l imit between wide , when

ARF in the predictive functions takes the value at that l imit , i . e . 2 , and
narrow when the appropriate va lue becomes aspect ratio/ 10 : thus

Fo.r aspect ratios > 2 0 , ARF = 2.0

For aspect ratios ' 2 0 , ARF = 2b/ 1 0h

104
Region 3

5.3.7 Another question in mind as the analysis of data from other sources
was carried out was the best function to de scribe region 3 . The SERC-FCF
results had shown preference for equ . 3 . 3 1 , though the s impler formula 3 . 3 2
was a lmost as good . A third funct ion is shown on Figure 5 . 4 :

D ISADF = 1 . 23 3 - 0 . 66 7 H * • • • 5.5

I t transpired , however , that few dat a sources provided a real test o f Region
3. Some did not extend to sufficient depth ; some were at a few wide-spread
depths that mis sed out Region 3 ; in fact Region 3 proved rather e lus ive so
that most data sets gave no basis for making any recommendation . The
conso l idated and averaged Myers ' results in F igure 5 . 2 appear to support a
constant D ISADF o f 0 . 9 5 in Region 3 , though individual results p lotted in
5 . 1 seem more cons istent with equation 3 . 3 1 or 5 . 5 . Myers was the only data
set that c learly included Region 3 and contained sufficient points to make
an ana lysi s o f the a lternatives worth whi l e . The statistical detai l s o f how
we l l the three a lternatives fit thes e data , both cons idering Region 3
r�sults by themse lves , and a l s o taking a l l r egions together , are as
fol lows :

Mean errors and s tandard deviations : %

A l l data : Region 3 only:

Equation 3 . 3 1 ( COH) : + 0 . 47 ± 4 . 1 9 + 1 . 7 2 ± 4 . 43

Equation 3 . 3 2 ( constant ) : + 0 . 49 ± 4 . 1 5 + 1 . 12 ± 3 . 39
Equation 5 . 5 ( H* ) : - 0 . 39 ± 4 . 00 + 1 . 86 ± 3 . 3 2

5.3.7 There i s n o clear cut conc lusi on to b e drawn regarding Region 3 .

There i s no firm evidence t o suggest changing from the formula that best
representing the SERC-FCF result s , namely equ 3 . 3 1 in terms o f channel
coherence , COH .

105
5.4. Skew channel s

5.4. 1 Ervine and Jasem , ( 19 9 1 ) tested two-stage channel s with skewed flow .
These are the only independent results featuring skew channe ls , but they
a l s o take on extra signi ficance in the definition o f the aspect ratio
adj ustment factor , ARF . The test flume at Glasgow was small compared with
the FCF at Wal l ingford , and might be regarded as providing a ver tica l ly
exaggerated mode l , about 1 / 1 0 scale on plan and 1 / 2 . 5 scale vertically , so
providing 4 t imes vertical exaggeration , with width/depth ratio about 2 . 5
rather than 1 0 as in the FCF . Having surface piercing rods as its
arti ficial f lood plain roughnes s in some tests , i t a l so provides results
with high ratios of f lood p lain to main channel friction factor .

5.4.2 The Glasgow f lume in which thes e tests were conducted i s 8 . 5m in

length , 0 . 7 6 4m wide , and the skew channel was 1 50mm wide by 6 1mm deep . The
ang l e o f skew was 5 . 84 ° , and it differred from the FCF in that the flood
p l ain was al igned with the gradient axis rather than the main channel being
so a l i gned . However , at such low angles of skew this is unlikely to matter .
The s lope was 1 / 1000 , which imp l ies 0 . 9948/ 1000 a long the channel axis ,
which would influence discharge by about 1 / 4% . This e f fect has been
neglected in the ana lysi s which fol l ows . The 1 0mm dia . rods used for flood
p lain roughening were in staggered rows , the rows being at 1 00mm centres
longitudinal ly and the transverse spacing a l so 1 00mm .

5 . 4. 3 Comprehensive cal ibration tests were made to establ ish the basic
resistance o f the main channe l , and o f the fl ood p lains when both smooth and
rough . Modified versions o f the smooth turbulent equation were derived as
basic res istance formulae for main channe l and smooth f lood p lain . The
calibration o f the rod roughnes s was analysed in the same way as the simi l ar
rougheni�g in the FCF at Wal l ingford , as described in Appendix 2 . The range
o( ratios of f l ow depth to rod diameter di f fered from the FCF range , and a
somewhat modi fied form o f drag coe fficient function was derived , with c0

being proportional to the -0 . 4 power o f the ratio z/d . The avai lability o f

good c a l ibration data with very modest experimenta l tolerances gives more
than average confidence in the analysis of the results from this research .

5 .4.4 The way in which the factor ARF works is very s imi lar to the
operation of the al lowance for skew deduced in Chapter 4 : they are both

1 06
multipliers o f a calculated dis charge de ficit . De a l ing first with the
roughened fl ood p lain case , s everal comb inations o f as sumed values for ARF
and o f the a l l owance for skew were considered . All these results lie in
Region 1 , except with the highest value of ARF tested when a small
proportion passed into Region 2 . The dominance o f Region 1 is to be
expected with high friction factor ratios , f / f ranging from 5 . 4 to 1 4 . 4 in
F C
these test s . The relative depth , H* , rose to 0 . 7 , a higher coverage o f H*
than in any other test series examined . These results are therefore
important in providing a good test of the ARF concept , though if comparab le
aligned channel tests had been carried out the results would have been o f
doub le value . The a l l owance for skew deduced for smooth flood p l ains as
given in Chapter 4 is in the form :

*
D I S DE F = D I SDE FAL G ( 1 . 03 + 0 . 074�) 5.6
SKEW I NED

and this yields a factor of 1 . 46 for � = 5 . 84 ° . However , the conclusion was

also drawn from the FCF tests with rough flood p l a ins that i f the friction
factor ratio exceeded 5 ( a s in this case ) then the allowance should be less .
An intermediate value of 1 . 17 was tes ted , based on the FCF rough f lood plain
value , as we l l as 1 . 00 i . e . no a l l owance for skewne s s . These three skewnes s
a l l owances were combined with several values of ARF , from 0 . 1 5 t o 0 . 30 but
the agreement between prediction and observation proved quite sens itive to
the product skewne s s factor x ARF . This is shown by the fol lowing tab le o f
results :

107
TABLE 5 . 2 . ANALYS I S OF ERVINE AND JASEM SKEW CHANNE L RESULTS :
rod roughened flood p l ains .

The upper figure i s the mean disc repancy between experimental and predicted
discharge s , % ; the lower figure is the variabil ity ( S . D . ) , ±% .

Skewne s s factor Aspect ratio factor , ARF

used
0 . 15 0 . 18 0 . 20 0 . 22 0 . 246 @ 0 . 30

1 . 00 -7 . 9 3 -4. 7 5 -1 . 32 +3 . 5 4
3 . 83 3 . 43 3 . 10 2 . 95
1 . 17 - 2 . 96 + 1 . 23 +5 . 83 +12 . 5 3 +24 . 15 <1!
3 . 25 2 . 97 3 . 05 3 . 96 3 . 45
1 . 46 - 1 . 50 +6 . 96 32 . 5 1
3 . 12 3 . 13 9 . 66

Notes : @ this is aspect ratio / 1 0

<I! includes region 2 flows for H*> 0 . 52

5.4.5 I t is cl ear from the above that ARF does not exceed aspect ratio/ 1 0 ;
and nor can it be much l e s s than this value . Bearing in mind that the
observed dis charge var ies from 85% o f the basic zonal calculat ion at the
sha l l owes t depth tested down to only 65% at the maximum depth tested , the

accuracy achievable o f 2% with variab i l ity not much above that aris ing from
the experiment s themse lves is heartening . In fact , with ARF x skew
a l l owance = 0 . 227 , the average di s crepancy reduces to zero and the standard
deviation is 3 . 0% . There i s evidence o f s l i ght curvature in the detai l ed
results . The 3% deviation is thus partly from the use o f a l inear functions
when a l l owing for relat ive depth and the e f fect of the relative roughnes ses
of main channel and fl ood p l a in .

5 .4.6 The overal l conc lus ion from these rough fl ood p l a in tests i s to

confirm once again that the appropriate value of ARF is close to aspect
ratio/ 1 0 , perhaps rather less than aspect ratio/ 1 0 when the main channel

width/depth ratio is as low as 2 . 5 . Also confirmed is the indication from

the FCF results that the a l l owance for skewnes s when· the friction factor
ratio exceeds 5 is rela tively sma l l . The comb ination of ARF = aspe c t
ratio / 1 0 and the "smooth" skewness factor would be unduly conservative . A
1 08
sma l l amount o f skewness appears to have very l ittle additional interaction
e ffect when the flood p lains are very rough .

5 .4.7 Various as sumpt ions were a l so te sted against the results with smooth
flood plains . The combination of ARF = 0 . 246 and the skewne s s factor o f
1 . 46 based o n the FCF formulation gave a mean dis crepancy o f -3 . 8% with
standard deviat ion 2 . 7% , a reasonab l e accuracy of predict ion , though the
analysis does not in fact test the ARF value as none of the data turn out to
be in Region 1 . The results progress through Regions 2 and 3 to 4 , with a
l arge proportion being in reg ion 4 , hal f the results being for H* values
above 0 . 3 5 . The overall conclus ion from these smooth fl ood p lain skew
channe l tests is that the predictive method based on the FCF data with ARF =

aspect ratio/ 10 is about 4% optimistic when compared with the conditions in

the Glasgow research . This i s within the combined tolerances to be expected
in this extens ion o f the proposed design procedures and those that might
have arisen in the experiments thems e lves .

5.5. Field information

5�5 . 1 Ramsbottom ( 1989) considered a l ternative methods of asses s ing fl ood

dis charges , making cons iderab l e use of field data from British rivers ,
supp l ied from the water data records of ten Water Author ites . Several o f
the rivers h e s tudied are two-s tage channe l s o f reasonably c l assical
cross-section : there is a main river channe l flanked on one or both s ides by
f lood p lains . There is obvious ly much more stage/dis charge information
avai l able for within bank f l ows , but the reason for choosing the rivers to
study was the avai lab i l ity of a proportion of above-bank data . The
within-bank data provides the basic calibration of main channe l roughnes s
that any method o f assessing fl ood conditions requires , but i n general there
is much less prospect of obtaining corresponding calibration
information about f lood p l ain resis tance . Several of the river sect ions
considered by Ramsbottom were also suitab l e for analys is in the context o f
the predictive methods now proposed , though not a l l were suf ficiently akin
to a conventional compound channe l sect ion , and some had too few stage
dis charge data for above bank conditions to provide a worth-while test o f
the method .

109
5.5.2 Another source o f information is the field research carr ied out by
Myers ( 19 9 0 ) on the River Main in Northern Ire l and . He has coup led together
laboratory res earch on a model of an improved reach of the river with
extensive field measurements , determining the discharge by many measurement s
across the width using a current meter reading at 0 . 6 depth as the
depth-average veloc ity . There are two study sections in this reach o f the
River Main , which are of classic two-stage channel shape though with
appreciably s loping flood pl ains . Again there are measurements at
discharges be low bank ful l to provide a basic resistance function for the

main channe l .

5.5.3 In the study by Ramsbottom , the within-bank data were analysed to

provide bas i c values of Manning ' s n , which were compared with those that
might be deduced from published values for s imi lar r iver s ( see Section 7 . 4
and Appendix 5 ) . Also , from inspection o f the particular r iver reache s , as
we l l as from the flow measurements on the f l ood p l a in in one or two cases , ·
suitable f l ood plain values for Manning ' s n were assessed. The same
discharge data , plus those obtained by courtesy o f Myer s for the River Main ,
were a l s o used by Wark under a Case Studentship with H R , Wal l ingford to
test the l atest vers ion of the lateral distribution turbulence method und�r
field conditions . As he had analysed the within-bank flows to determine
suitable Manning ' s n values - including their variation with depth - there
was s cope for cross-checking with the analyses for this proj ect , though on
the who le the near-bank-ful l values o f n used in what fol lows are the same
as those derived by Wark . Five river sections and associated stage/
discharge data were s e lected for analysis , and brief des criptions now
fo l low :

5.5.4 Ri ver Hain , Northern Ireland : This reconstrusted BOOm long

reach is .des cribed by Myers ( 19 9 0 ) and by Higginson , Johnston and Myers

( 19 9 0 ) . I t s purpose was to provide adequate freeboard for agricultural l and
in the upper reaches o f the catchment , whi l s t improving the channel
conveyance past industrial premises in the lower reaches . Strong fisheries
interests led to the adoption of a compound cross s ect ion through the reach
adj acent to the factory development . The main channe l is 1 2m wide , with
depth between 0 . 9 and 1m , and each berm s lopes towards the channel at 1 / 2 5 .
The gradient o f the study reach is 1 / 5 20 , and the water sur face s lope was
shown to match this we l l over a wide range o f flows . The r iver bed is
coarse grave l , with
1 10
o between 100 and 200mm . The s ide s l opes o f the main channe l consi s t of
50
quarried s tone up to 0 . 5 tonne weight , and the berms are usua l ly covered
with heavy weed growth . Two o f the surveyed sec tions , no s 6 and 1 4 , are
used in the analyses , with the flow data having been col lected from a bridge
at the up s tream end of the experimental reach .

5 . 5 . 4 River Severn , Hont ford Bridge . This is a much studied reach of

river alr eady mentioned in Chapter 2 . This is a natural river section with
a cableway extending over the ful l width inc luding the fl ood p lains , and a
l arge body o f accurate current metering data provides perhaps the best
ava i lable information about a natura l channel with flood plains . The se are
grass covered , and the gauged section is on a straight part of the river .
The width is about 40m , with a bank ful l depth o f 6 . 3m . Cross-section
parameters for this site are shown on figure 2 . 2 .

5 . 5 . 5 River Torridge , Torringt on , Devon . This is a natural river

sect ion with one f lood p l ain on the l e ft . Flow measurement is from a cable

way which spans the river channel itse l f and also the inner flood p lain up
to a flood bank . In fact a l l the data availab l e refer to conditions with
flow confined within the span of the cab l eway : higher floods can overtop _
the bank and gain access to the remaining width of flood plain , but such
conditions ar� not included in the avai lab le data . The channel bed is o f
sma l l stones up t o 0 . 3m boulders . The flood p lain i s pasture , with trees at
the river bank and on the flood bank . The river itse l f is 29m wide and
nearly 3m deep and is fairly straight , with gradient 1 . 39/ 1000 .

5 . 5 . 6 River Tren t at North Huskham . This is real ly a three-stage

channel , in that it has a narrow berm at an el evation of some 5m above river
bed , with an extens ive flood plain at a s l ightly higher elevation . The flow
gauging is confined to the main channel and narrow berm by the l imit s o f the
ca� leway , but visua l assessments of f l ood p l ain flows are made at higher
depths to add to the gauged discharge . There is extensive data at this site
for the two lower stages , rather less for the third s tage when the ungauged
wide fl ood plain might contribute up to 25% of the total discharge (result
o f model study at HR , Wal l ing ford , as we l l as from computation) . The
channel bed here i s of fine gravel and al luvial s i l t s , so woul d be expected
to become mob ile at high stages . The main channel itself is 7 2m wide , and
the gradient is 0 . 3 5 / 1000 . The f lood p l ain vegetation is mos t ly grass with

111
some sma l l trees and bushe s . Al though this s ite is real ly three stage , the
lower two stages are we l l within the scope o f the procedures developed in
the manual . Extension to the third stage i s rather problematica l , both
because of the geometric approximations that then have to be made and a l so
because the stage/discharge data inc lude an unmeasured proportion o f fl ood
p l ain f l ow .

5.5.7 Further detai l s of a l l these s ites wi l l b e found in Table 5 . 3 , and

their cross sections are inc luded in figures 5 . 9 to 5 . 1 3 . · The actual
cro s s - sections had to be s omewhat s imp l i fied , as wil l be expl ained in
Chapter 7 , sect ion 7 . 1 , to provide the various paramet ers that enter into
the predi ctive funct ions : bed width , 2b ; top width of channe l , 2 w ; mean
e
channe l depth , h ; width across flood p lains , 2B , inc luding the reduction to
this when f l ow only part ial ly inundates s l oping f lood p lains ; and average
b �� slope , s . These s imp l i fied shapes are also shown on the cros s-section
e
drawings : it should be noted that the method is intrins ically suited to
us ing the actua l cro s s-sections in the basic flow cal cul ations , though
computer program l imitations for this study meant that the simp l ified
sections were used throughout .

5 . 5 .8 I t having been estab l ished at this stage that the a l l owance for width
to depth for Region 1 flow should as sume ARF = 2b/ 1 0h , the appropr iate ARF
value was fixed for each geometry . Also , with one exception to be mentioned
l ater , the Manning ' s n value estab l ished from stage discharge approaching
but not exceeding b ank- ful l depth was appl ied to the range o f increased
depths in fl ood conditions . There was s ome fl exibi l ity in the choice of
Manning ' s n for the berms , however , as there is no cal ibration value
available in the usual field s i tuation . The predicted flows were compared
with the observed above b ank data for each river , and mean discrepancies and
st andard . deviations were calculated . The values with the range o f
a ssumpt ions tested are l isted i n Tabl e 5 . 3 . The predic ted stage dis charge
funct ions using the preferred comb ination of roughnes s coefficients et c .
are shown together with the ful l range o f obs ervations in figures 5 . 9 to
5 . 13 .

5 .5 .9 Figure 5 . 9 shows the River Main at section 6 and the comparison of

predict ion with measurement . The ful l l ine is for ne ·= 0 . 0 28 and n =
F

1 12
0 . 040 . The extrapolat ion to above bank conditions is quite good , though on
average the predict ions are 3 . 5% high with a standard deviation o f 8 . 6% .
I t seems quite pos s ible that the variability come s a s much from the field
data as from any imperfection in the predict ion method . The observations
appear to fo l low two trends , one set lying above the theoretical curve and
another set below , as might occur due to seasonal e f fects in terms of bank
vegetat ion or a change in actual channe l roughne s s . This possibil ity is
examined by taking n = 0 . 0 3 0 and n = 0 . 0 5 0 , shown as a broken l ine . Th is
e F
agree s rather better with the bulk o f the field data , but l eaves some
dis tinc tly o ff- l ine , and hence the mean di scharge disc repancy is +6 . 0% ,

with standard deviation 9 . 1% .

5 . 5 . 10 Figure 5 . 1 0 shows s imi l ar results for the River Main at s ection 1 4 ,

with n = 0 . 0 247 , the best-fit within-bank value at high stage , and n =
e F
0 . 045 . It should be mentioned here that the predictive procedure required
an a ssumption to be made to extend the formulae based on a horizontal flood
plain to cope with a slop ing one . C learly the flow is unaware of the ful l
width of berm when it is only partly submerged , and so the e f fective width ,
2B , i s taken to be the water surface width ( see a l so Figure 7 . 1 o f Chapter

7) . Also , with slop ing berms it i s far more likely that the normal Region 1
formula wi l l change sign at very shal low depths . This point was covered in
paragraph 3 . 4 . 1 1 but in analysing thes e results it seemed des irab le to a l low
for some res idual minimum interference under this sha l l ow partial
inundation conditions . Hence , a minimum value o f Q*
o f 0 . 5 was app l ied .
2C
This becomes rel evant at shallow flood p la in depths with low B/b ratio s ,
such as occur on s l oping berms . S everal of the data sets brought this l imit
into p l ay over the lower range of flood p la in depths . Agreement overall is
quite good , and to the s ame order as the field tolerances imp l icit in using
s ingle measurements in each vertica l : an average discrepancy of +3 . 5% with
standard . deviation 6 . 5% .

5 .5 . 11 The particular feature about the River Severn stage di scharge data
shown in figure 5 . 1 1 is that it is very ext ensive , with 39 values above

bank level . The theoretical curves shown are for n = 0 . 0307 and n 0 . 03 3 8 ,
e F
this n value being j us t 1% lower than the best fit value for within-bank
e
flows and the n value be ing an average figure for the two flood plains .
F
With so many field measurements available , it was considered justi fiab le to
take running averages of three to minimi se the ef fect of random errors , a

113
practice that was adopted for mo st o f the data from the FCF . The average
di s crepancy between theory and measurement i s +0 . 3% , with standard deviation
of 2 . 7% . This is perhaps the best set o f field data , and agreement with it
could hardly be bettered , using the now standard value of ARF = aspect
ratio/ 1 0 , and a we l l documented n value . It is worth mentioning that over
e
much of the range of fl ood flows , the discharge adjustment factor i s betwen
0 . 9 5 and 0 . 9 0 , and a l l flows are in Region 1 .

5 . 5 . 12 The data for the River Torridge shown on Figure 5 . 1 2 inc lude only 5
results c learly above bank , with one margina l ly above bank-full . The
predicted curve ( fu l l l ine ) i s for n = 0 . 026 , the best fit value for the
e
higher within bank condition , and n = 0 . 0 3 0 . Agreement i s less
F
satis factory , with a mean dif ference of 8 . 7% and standard deviation 6 . 5% .
The main discrepancy can be e l iminated by sett ing n = 0 . 02 4 and n = 0 . 026 .
e f
I t is possib l e that there are seasona l differences here that increase the
conveyance at t imes o f f lood . Th is river has an a l luvia l bed , so it is
conceivab le that the roughness chang es a t high stages when some o f the
sediment wi l l become mob i l e , there could be a de l ivery o f di fferent size

material into the reach from upst ream , or there might be local scour that
has not been t aken into account .

5 . 5 . 13 The final exampl e is the River Trent at North Muskham , shown in

Figure 5 . 1 3 . Here again there is qu ite a lot o f data , but the sect ion is
compl icated by having a berm and then a wider flood pla in at higher
e l evation , g iving rise to a dua l "bankfu l l " e l evation shown on the figure .
The detai led informat ion on this s ite refers to the po ssib i l ity o f this
reach of the river being affected by its conf luence with a t idal reach
downstream , so some o f the data may relate to a hydrau lic gradient
di f fering from the constant figure as sumed . The influence of f low over the
first berm can b e asses sed without much prob l em , as the s e ct ion up to main
f lood p lain leve l c lose to a conventional asymme tric compound channel . For
thi s case , n = 0 . 03 2 based on the analysis of within b ank flows (mean error
e
0 . 3 % ; standard devi ation 2 . 9 % average taking running averages of threes) .

The same value has been taken for n to give the theoretical stage/discharge
F
funct ion between BF and BF • This agrees reasonably we l l with the data :
1 2
average error 1 . 0% , st andard deviation 3 . 3% . (The observations have been
averaged in running threes ) .

114
5 . 5 . 14 In e ffect the third l evel o f this River Trent s ite provides another
test case , though to apply the normal theory the section has to be
approximated somewhat as shown on Figure 5 . 1 3 . Also the upper range o f
flow incorporates a n unmeasured f l ood pl ain component , s o the data is not
idea l . In order to achieve reasonab le agreement it has been necessary to
adj ust the main channel Manning ' s n value with stage , and the theoretical
l ine shown uses ne varying l inear ly from 0 . 033 at flood pl ain stage to
0 . 0266 at 0 . 5m greater depth . Other treatments of these data have also
been obl iged to bring in a reduct ion of main channel roughnes s at very high
stage s , which could we l l be justifiable and real istic in view of the
al luvial nature of the river bed , and the possibi l ity of a change in
e f fective gradient in large f loods .

5 . 5 . 15 The use o f field data to val idate a predictive method i s less

rigorous that using good laboratory dat a because there is usua l ly no
calibration data t o provide. the basi c f lood p lain resistance , and also the
roughnes s coeffic ient o f the main channel is known to be variab le in many
cases - and suspected o f varying in other cases too at high stages when an
al luvial bed may be in motion . However , from the engineering point o f
view , val idation b y comparison with large s c a l e information from natural
and improved r ivers is particular ly s igni ficant . Such val idation has been
accomp l ished , without need for any further cons iderat ion of the
necessary adjustment for width/depth rat io , to the expected level o f
accuracy . In fac t , with some o f the better data s ets , agreement has been
to within the tolerance band of the measurements thems e lves .

5 . 5 . 16 It is interest ing to note that these practical cases are for the
most part dominated by Region 1 flows : only the River Main progresses to
Region 2 over the depths considere d . Yet Region 1 behaviour was virtua l ly
absent from nearly a l l the sma l l s ca l e l aboratory work considered ear l ier
in. this Chapter . This i s amp l e justi fication for the programme o f research
in the FCF at Wal l ing ford , undertaken by s everal groups of dedicated
investigators from Universitie s , supported by funding from SERC .

5 .6. Conclusions from other data sources

5 .6. 1 The predictive functions based on the results from the l arge SERC­
supported FCF at Wal ling ford proved robust in that they trans ferred we l l to

1 15
mo s t other sources o f s tage discharge data , both from laboratory and fie ld ,
covering a discharge range from 5 1 / s to over 5 00 m 3 / s . Regions 2 , 3 and 4
were wel l va lidated by independent laboratory s c a l e information to within
the l ikely tol erances o f the data , though not a l l sources of data with
rough flood p l ains could be satis factorily re conci led . Re gion 1 was also
we l l validated by independent data involving some skew , and also by the
field data .

5.6.2 Regions 2 , 3 and 4 are predi c ted by functions depending on channe l

coherenc e , COH , and so the satis factory trans fer to radical ly different
geometries and roughnes s types demonstrates that the channel coherence is a
useful measure of the way in which inter ference between main channe l and
flood pl ain flows a f fects the stage discharge funct ion .

5.6.3 Re gion 3 is not wel l covered by the independent test series , but the
one series which covered that region adequate ly did not provide a strong
differentiat ion between thr ee al ternat ive functions for that region . Thus
the function derived as the best repre sentation o f large scale tests , in

terms o f COH , shou ld be retained .

5.6.4 Region 1 presents a signi ficant prob lem in small channels o f narrow
main channe l aspec t ratio . Data from di f ferent s ources can present
somewhat di fferent pictures , either implying the absence of region 1 and
inter ference e ffects caus ing maj or reduction in conveyance at shallow
depths over the flood p l ain , or the virtual absence of interference e ffects
on the s tage discharge curve at any depth . No such doubts arise with large
scale or field data : Region 1 c learly exists and there is a gent le increase
in interaction e f fects as the flood p lain become s inundated , as observed in
the FCF .

5 .p.5 There may be some doubt as to whether al l the l aboratory testing used
suffic ient ly long approach sections to achieve secondary circul ation
patterns that were representative of uniform flow in very long channe l s ,
though c learly the larger scale tests were ade quate in that regard . The
interpretat ion o f s ome o f the l aboratory research has to remain somewhat
open . In any event , its relevance to large systems with greater
width/depth ratios is dub ious .

1 16
5 .6.6 Some f l exibil ity was introduced into the Region 1 formulation to
al low for an anticipated effect from main channe l aspect ratio . This takes
the form of an additional aspect ratio factor , ARF , and with this
flexib i l ity bui l t in the great maj ority o f test results from other sources
could be predicted within the order o f accuracy imp l icit in the laboratory
procedures and field observations .

5 .6.7 The conclusions drawn from the re-ana lysi s o f l aboratory scale
research on the best value o f ARF to use for Region 1 were not very robus t ,
in that many tes t series did not - and could not because o f low flood p lain
Reyno lds Numbers - include flows sha l low enough to provide a reasonable
test for region 1 . However , other test data included this region and it
was suggested that for main channel aspect ratios exceeding 20 , a cons tant
value o f ARF = 2 should be taken (a wide channe l condition) , whereas for
narrower conditions ARF = aspect ratio/ 1 0 , making the discharge deficit
proportional to channe l width rather than depth . This last point was then
ful ly confirmed by fie ld data , though no very wide rivers were include d .

5.6.8 Much research e f fort went into l aboratory studies o f compound

channe l s with sma l l main channe l aspect ratios , but it now appears that
their per formance may not have a very c lose relat ionship with the more
practical range of geometries and channel s izes . Of course this practice
stemmed from the l imited size of typical university fac i l ities , and the
inappl icabil ity of much o f the smal l scale research to engineering des ign
shows a l so the wisdom of those who fought s trongly for the provision of
large scal e faci l it ies .

5 .6.9 Particular importance is attached to the satisfactory prediction o f

flood discharges in the real r ivers for which data were availab l e , with
agreemen� being achieved to within the probabl e tolerance s of the data and
wel l within the obj ective in mind at the s tart o f the proj ect in terms o f
accuracy o f computation metho d .

1 17
 TABLE 5 . 1

OTHER SOURCES OF STAGE D I SCHARGE DATA ( Laboratory research)

NAME Date Max 2B . Max h Range Range Range of se Total Flood plain roughne s s e s , NOTES
mm mm o f B/b of 2b/h flume sl . hor/ no o f Manning ' s n values
/ 1 000 vert results

ASANO 1985 3000 121 1 . 25 - 10 - 0 . 94 - 0 44 0 . 0098 Ho detailed

3 . 33 30 1 . 07 calibration .
Very non-un .
flow
ERVINE 1992 764 61 5 . 10 2.5 1 . 00 0 14 Varies with depth 5 . 84° skew
Rod roughness
10mm dia in
100mm grid
1-' KIELY 199 1 1200 54 6 3.7 1.0 0 10 0 . 0 1 0 , 0 . 0 157 Plotted
<»
1-'
(with bed shift) cal ibrations
KNIGHT 1983 610 76 1 . 00 - 2 0 . 97 0 18 0 . 0 10 Special care
1984 4.01 re uniformity
Calibration
ava il able
MYERS 1978 760 121 3 .21 - 1 . 32 - 0 . 22 - 0 153 0 . 0098 5 cal ibration
1983 4 . 74 1 . 99 2 . 28 runs , 1977
1984 Some asymm .
data
· PRINOS 1983 1 2 70 102 2 . 70 - 2.0 - 0 . 29 7 0.5 80 O . Ol l , 0 . 0 14 , Ho calibration
1984 5 . 26 5.0 0 . 0 1 8 , 0 . 02 2 data availabl e
Anomoly between
narrow &: wide
US WES 1956 9 1 00 152 8.0 - 2.0 - 1.0 0.5 9 0 . 0 1 2 , 0 . 025 . Ca libration
30 . 0 4.0 0 . 03 5 approx. avail able .
Large scale .
WORMLEATON 1 2 10 120 4 . 17 2 . 42 0 . 43 - 0 39 .0 . 01 0 7 , 0 . 0 1 3 5 , Many calibration
1982 1.8 0 . 0 1 7 , 0 . 02 1 • runs for all
roughnesses
TABLE 5 , 3 ,

ANALYS I S OF FIELD DATA FROM RIVERS FOR WHICH STAGE/DISCHARGE DATA ARE
AVAILABLE .

River Location Number Width/ Disch . G radient n n Mean S.D

e F
of depth range / 1 00 0 dis c .
obs . ratio m 3 /s % %

Main , N . Ireland 14 12 . 3 14. 8 - 1 . 90 6 0 . 03 2 0 . 040 +8 . 0 10 . 5

section 57 . 8 0 . 03 0 0 . 040 -8 . 7 9.2
6 0 . 028 0 . 040 -3 . 5 8.6
0 . 03 0 0 . 05 0 +6 0 •. 9.1

Main_, N . Ireland 11 11.0 18 . 5 - 1 . 906 0 . 0247 0 . 040 + 10 . 6 11.5

section 57. 8 0 . 0247 0 . 03 0 +6 . 7 7 .9
14 0 . 0247 0 . 025 +3 . 5 6.5
0 . 0247 0 . 020 0 5.8

Severn Mont ford 36 @ 3.0 170 - 0 . 195 0 . 03 1 0 . 03 5 +1. 2 2.8

Bridge , 3 13 0 . 03 2 0 . 03 5 +4. 1 2.8
Engl and 0 . 03 0 0 . 03 5 -1.7 2.9
0 . 03 0 0 . 033 -2 . 4 2.7
0 . 03 1 0 . 0 37 +1. 8 3.1
0 . 0 3 07 0 . 0338 0 2.7

Torridge Torrington 6 8.3 208 - 1 . 45 0 . 026 0 . 060 +11 . 5 6.9

Devon , 314 0 . 026 0 . 03 0 + 8.7 6.5
Engl and 0 . 024 0 . 026 + 0. 1 6.0

Trent N Muskham , 25 @ 7.4 395 - 0 . 3 20 0 . 03 2 0 . 03 2 +1 . 0 3.3

Eng land 530

9 3.3 595 - 0 . 03 0 - 0 . 040 + 10 . 0 2.3

857 0 . 02 5
0 . 027
0 . 0 23 -1 . 0 2.0

@ denotes running aver ages o f three were taken .

1 19
Fig 5.1 Selected data chan ne ls with equal main channel and flood plain
roug h ness: d ischarge adjustment factor - DISADF, versus
_
_

0.5 .-
----r-+::---.
t-
+
+
+

+
0.4 - +
++ +
+ +
+
+
+
++ + +
+
+
+
0.3 - +
+

..
:I:
+
+ +
0.2 r- + +
+
+

0.1 r-

I I I
0
0.80 0.85 0.90 0.95 1 .00 1 .05
DISADF JBW/30/1 0-91 /3D

Fig 5.2 Myers' resu lts consolidated with ru nning averages of three
 0.5

0
0 I
0.4 -
0 '\.
0 SERC - FCF 1
V
0 >�'
. ...... -v,.,·
Prinos
- ....
/ .. .
0.3 X Wormleaton .
1 0/
�- /
.. .

/.� .I
:I:

0.2 r- .
.
) 0
.// /
y .
>(.
I.
•

0
0.1 -
*
0

I I I I
0
0.5 0.6 0.7 0.8 0.9 1 .00
DISADF

Fig 5.3 Selected data from channels with flood plai n Manning's n
approx. 50°/o above main ch �nnel value: discharge adjustment
_

0.4
.......
'I- ,
I
I
I
0.3 I

0.2

0.1
Aspect ratio = 2

0
0.70 0.75 0.80 0.85 0.90 0.95 1 .0
DISADF JBW/31110-91130

Fig 5.4 Pred ictive fu nctions for two geometries : SERC-FCF, aspect ratio 1 0,
8/b = 4, Se = 0; Knight, aspect ratio 2, 8/b = 4, Se = 0; equal main
channel and flood plain roughness
 0.4 1�­
�?

Rod /
roughness
0.3
' ....
\a AR = 2
•

:::t: \ i' ARF - 0.2

\0:
\
0.2
\
\
\
\
0.1
�

0
0.5 0.6 0.7 0.8 0.9 1 .00
DISADF

Fig 5.5 Pred ictive functions for two geometries and roug hness types:
aspect ratio 1 0, rod rough ness on flood plain; aspect ratio 2,
boundary rough ness with flood plain Manni ng's n approx. 50o/o
above mai n channel n

0.5 r------,--,

Flood plain 2b = 305mm

roughness: - - - - - - - · 2b = 406mm

0.4
0 0.01 1
X 0 .0 1 4
+ 0.01 8
!::. 0.022

0.3

:::t:

0.2

0.1

0
0.70 0.80 0.90 1 .00
DISADF

JBW/32/10-91130

Fig 5.6 Prinos and Townsend's resu lts for channel aspect ratios of 3 and 4:
DISADF versus relative depth, fou r d ifferent flood plain roughnesses
 50 I I I I I I I I I I I

20 r- -

10
..a 1-
m l:.U
. -

1- -

.2 1-
I

tW
-

li! 1- !:. Ky
.t: p -

""0
-
0 E
�
5 -

oM !:.
Wo
-

�w
-
Kt !:. <> o P -

p
0 OA
8M
oP
-
Kt -

��
)A
t:. A

2 - + Kt oA -

!:. A
OA
OA
1 I I I I I I I I I I I

1 2 5 10 20 50
Main channel aspect ratio - 2b/h

• Definitive SERC-FRC results with ARF 1 .0 =

!:. Results suggesting ARF 2b/10h =

<> Results suggesting normal basic value of ARF 1 .0 •

+ Results suggesting little interference, i.e. ARF < 2b/1 Oh

0 Results consistent with any value, 1 � ARF � 2bl1 Oh

JBW/33/10-911:10

Fig 5.7 Scope of geometries covered by review of stage d ischarge

i nformation, i n terms of width ratio - B/b, and mai n channel
aspect ratio - 2b/h : also showi ng i nfluence of aspect ratio on
predictive fu nction.
Sources: A - Asano et al, Kt - Knight et al, Ky - Kiely, M - Myers,
P - Prinos and Townsend, U - US WES, W - SERC FCF,
Wo - Wormleaton, E - Ervine and Jasem (skew)
 'Separate-channels' method (M2)
T Wormleaton et al (nr "' 1 .9)
li Authors (nr = 2.0, We = 20.3cm)
0 >> (nr 2.0, Wc = 30.5cm) •

X >> (nr = 1 .3, Wc = 20.3cm)

0 Myers (nr 1 .0) Q

0.5 r------,

,.

0.4 1-
I
., 0
I

T �
'
0
I I
Cl
"'
�· ){ '
'
; , \ I

�
"' , \ I
0.3 - "'
, \ ,
r , \ I
, , 0 ){ '
H. 11
I
I , p
l I
l't I
I
11 I
I I I
I �
p. � I
� I
){ I
I
I
I I
0.2 - ,�
I Q
11 I
I
, I
I
J)
' I

,
'
� p ¥
I
I I
y I I
I
0
I
I , I
I I
I I
I
I I

, I
,
0.1 1- I
I
A 6 X 0

I I I I

0 0.2 0.4 0.6 0.8 1 .0

DISADF

JBW/34110-91130

Fig 5.8 Prinos and Townsend plot of DISADF agai nst relative depth for
their own narrow channels and other data sources : nr is ratio of
flood plai n to n1ain chan nel Manning's n
 River Main Section 6
Cross-section geometry

--+- Channel bed

2.0 -

1 .5 -

0
(!)
� 1 .0 -

!I)
(I)
....
-
(I)
E
-
0.5

0.0 -
I I I I I I I
0 5 10 15 20 25 30
metres

Stage discharge

2.50

2.25
--
2.00 - --
...-�- -
0 1 .75 --
(!)
<(
!I) 1 .50
(I)
-=
(I)
1 .00
s
(I)

�
(/)
1 .25
+
0.75 . ++ BF
*-H-
0.50 + + Measured
+
0.25

0.00
0 10 20 30 40 50 60
Discharge (cumecs)

JBW/35110-91130

Fig 5.9 Field i nformation: River Mai n, N.l., section 6. Predicted stage
discharge cu rve for ne = 0.028, nF = 0.040, full line; ne = 0.030,
n F = 0.050, broken line
 River Main Section 14
Cross-section geometry

- Channel bed
2.0 -

1 .5 -

0
(!)
� 1 .0 -
(/)
Gl
.::.
Gl
E

0.5 -

0.0 -
I I I ' i I I
0 5 10 15 20 25 30
metres

Stage discharge

2.00

1 .75

1 .50
0
(!)
<( 1 .00
(/)
Gl
....
-
Gl
1 .25
.§.
Ql
Cl
.s 0.75
m

0.50
+ Measured
0.25 +

0.00
0 10 20 30 40 50 60
Discharge (cumecs)

JBW/36/1 0·91/30

Fig 5.1 0 Field i nformation: River Main, N.l., section 1 4. Predicted stage
d is charge cu rve for ne = 0.0247, n F = 0.025
 River Severn at Montford
Cross-section geometry

_..,_ Channel bed

6.0 -

5.0 -

4.0 -

3.0 -
G)

:
E
•
2.0 -

1 .0 -

0.0 -

-1 .0 -

-2.0 - I I I I I I
0 25 50 75 1 00 1 25

metres

Stage discharge

6.0

5.0
0
�G)
4.0
•
.::.
• 3.0
.§.
•
m
.s 2.0
.(/)
+ Measured
1 .0

0.0
0 50 1 00 1 50 200 250 300 350
Discharge (cumecs)

Jf!J'N/3711 0-111/30

Fig 5.1 1 Field information : River Severn at Montford Bridge. Predicted

stage discharge cu rve for ne = 0.0307, n F = 0.0338
 . River Torridge at Torrington
Cross-section geometry

20 -

_....,_ Channel bed

19 -

CO
18 -
�
•
E
17 -

16 -

15 -

14 I I I I I
0 20 40 60 80 1 00 1 20
metres

Stage discharge

1 9.0

1 8.0 +

0
0
�
!I) 1 7.0
f
1ii
.§.
•

l
en
1 6.0

1 5.0 + Measured

1 4.0
0 50 1 00 1 50 200 250 300 350
Discharge (cumecs)

JBW/38110-91/30

Fig 5.1 2 Field inforn1ation: River Torridge, Devon. Predicted stage

discharge cu rve for ne = 0.026, n F = 0.030
 River Trent at North Muskham
Cross-section geometry

11 -
_...,_ Channel bed
10 -

9 -

8 -
ut

g
ID
7 -
E
6 -

5 - e........�i.
I
rt
4 -

3 -
.r
fit --

2 -
1 - I I I I I

0 50 1 00 1 50 200 250 300 350 400

metres

Stage discharge

9.0

8.5

0 8.0
0
et:
(/J
ID 7.5
-=
ID
.§.
ID 7.0
0>
s
(/)
6.5

+ Measured
6.0

5.5
0 1 00 200 300 400 500 600 700 800 900
Discharge {cumecs)

JBW/30/1 0-01130

Fig 5. 1 3 Field i nformation: River Trent at North Muskham. Predicted

stage discharge cu rve for ne = 0.032, n F = 0.032 at first flood
berm ( B F1 to B F2); ne variable above flood plain (BF2), n F = 0.036
6. TURBULENCE METHODS

6. 1 Resume o f turbul ence theory a s current ly app l ied to compound

cross-se ctions

6.1.1 The method o f p redicting the s tage-discharge function derived by

empirical analysi s o f the FCF results has its l imitations . Although based
o n dimensional analysis and wel l supported by independent sources of
information , there remains a question about its generality . One problem is
its transfer to compound channel s that do not have such a simple geometric
shape as the ones typical ly tested , with their symme try and horizontal flood
p lains . A method is described in Chapter 7 , section 7 . 1 , for extending the
empirical equations to more natural sections , which does not stretGh the
procedures too far b eyond the s cope of the supporting research , but it would
have been more satis fying to have a method which does not depend s o much on
the geometry ' s conformity to s ome near-ideal norm. Turbulence methods have
that degree o f generality and so deserve s erious consideration . It wil l be
demonstrated , however , that at their present stage of development there
remain uncertainties over their appl ication in the general design of
compound channel s .

6.1.2 Turbulence theory was briefly reviewed in Chapter 2 , 2 . 3 . 1 3 to

2 . 3 . 16 . I t is based on the most fundamental equations of fluid motion ,
which include terms describing the mechanism whereby turbul ence generates
energy diss ipation within the body of f luid . It differs from the hydraulic
equations familiar to engineers , such a s Manning , Chezy and even the more
compl�x formulae for non-uni form flow ( the Bernou l l i equation) and for
non- steady f low ( the St Venant equations ) in that it is based on the
internal mechanics of the fluid , rather than on a knowledge of the external
forces o� the whol e cross-s ection . Turbulence equations bring in the local
internal syst em o f stresses ( the Reyno lds stre s s e s ) which arise from the
gradients across the section in the three dimensional ve locity structure .
Full s o lution o f the resulting three-dimensional equations is pos s ib l e for
steady uni form f l ow in open channel s of a lmost any cro ss- section , but is
expens ive in computer time even with present-day fac i l ities and is by no
means straight forward : it remains very much a research area . It is more
usual there fore to s imp l i fy the turbul ence equations to g ive a quas i
two-dimens iona l approach whi ch considers any cross - section geometry but

120
takes the l ateral distribution of verti c a l ly averaged vel ocity ( or dis charge
intensity) as a sufficient measure of the flow s tructure arising from the
irregular cross-sectional shape .

6.2. Turbulence methods and comparison with FCF data

6. 2 . 1 Shiono and Knight ( 1990 a , b ) have reviewed the more practical methods
in the l iterature , drawing attention to the importance of momentum exchange
arising from s econdary current s as wel l as from the Reynol ds stres ses . They
develop the depth averaged momentum equation in the longitudinal direction
for uni form f l ow at depth H for any point in the cro s s section and c l early
distinguish between the Reynol ds stress term , which depends on the mean
value of the product o f local ins tantaneous forward and transverse .
velocities , and the secondary current terms , which depends on the product o f
average forward current s and average secondary current s . ( See Appendix 4
for detailed theory) . With certain as sumptions , these s impl i fy to the
fol lowing partial differential equation :

a du
pgdS + - [ d(£ + £ ) - ] - 1: = 0 • • • 6.1
ay s t dy b

where d is the local flow depth , S the channel gradient , p the f luid
density , g i s the gravitational accelerati on , y i s a pos ition across the
flow sect ion , U i s the depth average ve locity at position y , 1: i s the
d b
local bed shear stres s , and E and E are the equivalent eddy viscosities
s t
aris ing from s econdary currents and turbulence respectivel y .

6.2.2 A s Shiono and Knight ( 1990 a , b ) point out , whenever such l ateral
distribution turbulence mode l s come t o be used , they immediate ly pose the
prob lem of what values to use for the eddy viscosity terms . Some
researche�s have adopted a s ing l e constant value for non-dimens ional eddy
viscosity , NEV = e/U*H , acros s the who l e sect ion , say 0 . 16 which is a
typical value for non-compound open channe l s , equivalent to setting :

. • • 6.2

where U* is the local value of the shear velocity at the boundary and H is
e
local depth .

121
Other have varied the NEV value across the section according to some
relationship based on l aboratory resul ts or field data ( Shiono and Knight ,
1990 a , b ; Knight , Shiono and Pirt , 1989 ; Knight , Samuel s and Shiono , 1990) .
Wormleaton ( 1 988) used a two-component eddy viscosity , the second component
being based on the l ength scale and velocity scale of the shear layer , to
account for the turbulence that is clearly generated by the shear zone at
the inter fa ce . NEV i s commonly treated as a c atch- a l l parameter covering
the various turbulence e ffects , inc luding both the shear layer and the
influence of secondary currents . However , a lthough there is now some
understanding o f the balance between these turbulence sources , there i s a s
yet n o general method f o r determining the appropriate value o f turbulence
coefH cient for various geometries and roughnes s conditions . It is this
l ack that makes even the s imp l i fied turbulence model s di f ficult to .justify
in a hydrai l i c design context unl e s s local cal ibration data i s avai lable
over a good range o f conditions .

6.2.3 The e quation used by Wark , Samuel s and Ervine ( 1 990) is the discharge
intens i ty form of l ateral distr ibution e quation , rather than the depth
averaged velocity ver s ion used by Shiono and Knight ( 1990 a , b ) , the two
forms giving signi ficant ly different results where the depth varies
strong ly , i . e . at the bank l ine .

L!_g l a �
8 dl
pgdS - [e - 0
_
6.3
+ ay t ay ]
• • •

where � is a factor re lating stress on an incl ined sur face to that in the
horizontal plane . In the above , f is the local friction factor arising from
boundary shear , as determined from the appropriate resistance equation for
the local value s of depth , d , and discharge intensity , q . The terms in this
equat ion express the balance between gravity , and bed shear plus l ateral
shear . Of course i f e i s used as a catch-al l term , the third term a l so
t
in�ludes the shears generated by s econdary current s . These Authors found

that the mos t appropriate form o f finite difference solut ion to equ . 6 . 3 is
one that computes the lateral shear term at mid-node positions . The
equation i s non-l inear and Newton ' s method o f iteration is app l ied , with an
initial ' s eed ' so lution obtained by setting the term to zero , in other words
us ing the bas ic zona l c a l culat ion of main channel and f lood plain discharge
intens itie s . Convergence usual ly occurs within f ive or s ix iterations ,

122
though the final result may be 1 5 -20% different from the intial seed value .
The question is , how wel l does it per form?

6.2.4 This had been examined in broad terms in the 1990 paper by Wark ,
Samuels and Ervine for a selection of data from the FCF at Wal l ingford and
e l s ewhere . Their figures reproduced a s Figures 6 . 1 and 6 . 2 i llustrate that
reasonabl e agreement i s pos sib l e , though it has to be remembered that what
i s being sought is a correction to a basic calculation that is already
approximately correct . Close inspection o f Figures 6 . 1 and 6 . 2 is required
to judge how accurate the s imulation might be in reproducing the observed
s tage discharge curve . The plotted observations tend to traverse the
sequence o f theoretical curves for di fferent values o f NEV , requiring an
increas ing value as depth increases for s ome distance , then reversing the
trend to cross the sequence back to l ower value s .

6.2.5 Resear ch funding for a more detailed comparison between this

turbulence method and the ful l set o f FCF data became avail able through the
Mini stry o f Agriculture , Fisheries and Food research prog ramme at HR ,
Wal l ing ford . The first approach was to test the method with a range of
values o f NEV from 0 . 16 to 0 . 29 , to examine how wel l each value represent .
the observed data for all geometries , with both smooth and rough f lood
pla 1ns . The goodness of fit was represented by the mean discrepancy and the
s tandard deviat ion about that average . This was done with both the
individual s tage discharge readings and also with running averages of threes
to e liminate s ome o f the experimental t o lerances . ( These procedures had
previously been fol lowed in the empirical analys is of the results described
in Chapter 3 ) . Only the averaged data result s are summa rised here in Tab le
6 . 1 appended . Taking all the smooth flood p lain cases together , it wi l l be
s een that NEV = 0 . 2 7 gave the best fit of the values tested , with mean error
about 0 . 4% and var iab i lity under 4% . This was not the optimum NEV value for
al� the geometries however : some required higher and some lower values o f
NEV for best prediction . (Geometrical information for the numbered test
series is given in Tab le 3 . 1 ) . With roughened flood plains , taken a l l
together , NEV = 0 . 2 2 gave best resul t s , though thi s was not true of the
rough series taken individual ly .

6.2.6 Just three examples o f the results in graphical form are shown , as
Figures 6 . 3 , 6 . 4 and 6 . 5 . The first two are for NEV = 0 . 2 7 , the best

123
overall value with smooth flood plains , and the third is for NEV =0 . 2 2 , the
best for rough flood p lains . They show the residual discrepancy ratio
between the turbul ence prediction and the observations , p lotted against the
ratio o f f l ood p lain flow depth to main channel depth , ( H-h) /h . ( Re lative
depth , H * = ( H-h) /H , is shown as an additional non-l inear scale) . This is a
much more sensitive way o f p lotting than that o f Figures 6 . 1 and 6 . 2 and
i l lustrates how the error varies systematical ly as depth increases .
Figures 6 . 3 concerns the sequence o f tests with varying width ratio s , B/b .
The variab i lity expres sed a s standard deviation about the mean error of
under 4% typifies all the smooth flood plain series , but here the p lotted
observations cover an actual error band from -5% to + 1 0% though of course
the many point s with discrepancies below 4% dominate the picture .
( Statistical ly about 2 / 3 rds of the data shoul d lie within the band .defined
by SD . ) What is interesting about thi s p lot is the clear indication that a
fixed value o f NEV , even when chosen as best fit to a set of data , does not
provide a prediction method of high accuracy: its per formance varies with
relative depth . Also , there are indications that the turbulence procedures
somehow fail to account for the different regions of flow that are observed
as depth increases . Figure 6 . 4 provides corresponding results where the
channel s ide s lope is varied from series to series , with much the same
overall spread of dis crepancies but with a rather different pattern with
depth . Figure 6 . 5 i s for the rod roughened f lood plain tests , which a l so
varied series to s eries in their main channe l s ide s lope . There is a
s omewhat wider spread of residual discrepancy , 10% : the s tatistical finding
was that the standard deviation for the whol e of the rough f lood plain data
was under 6% with thi s value o f NEV.

6.2.7 The next form o f analysis was to assess the optimum value of NEV for
each individual stage/discharge result in all the FCF test series . This
involved �ulti-stage iteration : the solution of the lateral shear turbulence
e q�ation is itse l f an iterative finite difference solution , but it had to be
appl ied repetitive ly to converge on the value o f non-dimens ional eddy
vis cosity , NEV , that gave agreement with the measured discharge . Tolerances
and a l s o maximum numbers of iterations were specified . This approach was
also app l ied to the in- bank calibration data for the main channe l , and the
mean value o f NEV obtained for the non-compound trapezoidal section was
0 . 125 ± 0 . 1 0 , appreciably below the average values found appropriate for
compound cross-sections .

124
The figure for a l l the smooth resul t s was 0 . 29 0 ± 0 . 13 , and for a l l the
rough results 0 . 2 1 7 ± 0 . 08 . The average best fit values of NEV and their
variabi lity are given in the following table :

Test series Av . NEV S D

. 1 0 . 3 14 0 . 120
2 0 . 259 0 . 103
3 0 . 195 0 . 05 2
5 0 . 240 0 . 05 7
6 0 . 207 0 . 07 7
7 0 . 20 1 0 . 045 rough f lood p lains
7a 0 . 152 0 . 029 rough f lood p lains
8 0 . 35 6 0 . 150
9 0 . 13 7 0 . 03 0 rough flood p lains
10 0 . 429 0 . 109
11 0 . 300 0 . 07 1 rough fl ood p lains

6.2.8 It is thus apparent from the variation o f the average values o f NEV
from series to series that the geometry and roughnes s condition a f fects tne
value that best fits the data . The s t andard deviation expresses the
variabil ity from depth to depth in any one series using the optimum value o f
NEV ; the variability is c l ear ly appreciable . In the main , this variation is
not a random s catter : there i s sytemat ic var iation with depth as shown
vividly by Figures 6 . 6 t o 6 . 8 . In thes e , individual values o f NEV are
plotted against f lood p l ain depth ratio , ( H-h ) /h , with H* as an additional
scale . (Note that the test data are averages of three , to reduce
experimental s catter ) . Figure 6 . 6 r e fers to the test series at varying
width ratios ( see Tab le 3 . 1 for geometric detai l s ) , and c l early as the f l ood
p lains f�r st become inundated NEV is, c lose to the s ingle channe l value of
0 . � 25 , increas ing through Region 1 to between about 0 . 27 and 0 . 45 depending
on width ratio , B/b . Then through Region 2 as depth increases the value o f
NEV reduces again , with some hint o f a di f ferent t rend through Region 3 and
beyond . At g reater depths , where the channel coherence is above about 0 . 95 ,
the optimum va lue o f NEV seems to be returning towards a bas ic value of the
orde r of 0 . 1 5 again , (which is typ ical of the values usua l ly quoted for
smooth l aboratory f lumes ) , for the narrowest flood pl ains , test 03 .

125
6.2.9 Figure 6 . 6 referred to tests with channel side s lopes o f 1 : 1 . In
Figure 6 . 7 this bank s lope is represented by test 0 2 , whi lst 08 is for s
e =
0 and 1 0 i s for s e = 2 . Series 0 8 , with rectangular main channel , appears
to have a sudden switch from the basic low va lue of NEV to a much higher
va lue at relative depth , "* ' about 0 . 23 , ( H-h) /h = 0 . 3 . Does thi s indicate
a reluctance for secondary currents to cross the bank l ine when banks are
vertica l , unti l some triggering depth is reached? The three series do not
form a coherent sequence of resul t s , however : s e = 1 does not l ie between
the results for 0 and 2 , although a l l show a characteristic S-shape as depth
increases . There is no evidence o f a return to a basic low value of NEV at
greater depths in this p lot .

6 . 2 . 10 Figure 6 . 8 gives the results for the test series with rod-�oughened
f l oo d p lains . Again at first inundation o f the flood p lains the optimum NEV
value is close to the basic simp l e channel value . It increases with depth ,
with quite s trong dependence on the bank s l ope , s . Here the sequence is
e
logical , with the vertical bank condition ( series 0 9 ) requiring lower NEV
values than the other series with s l oping banks . The vertical bank
condit ion r e quires l ittle increase over the basic value o f NEV , which
sugges t s a l imited influence o f secondary current exchange between main
channe l and f lood p lain .

6.2.11 I t i s apparent that to achieve accurate p r edictions from a

s imp l i fied l ateral distribution turbulence model with an a l l -embracing
non-dimensional eddy viscosity or turbul ence coefficient , a considerabl e
degree o f empiricism woul d s ti l l b e required t o accommodate the apparent
dependence on relative depth , width ratio , channel s ide-slope and f l ood
plain roughnes s . I t appears that the different regions of f low identi fied
from the or iginal empirical analysis of the F e F results are confirmed by the
variation of NEV , and would therefore have to be represented separat e ly in
any empirical functions for NEV .

6.3. App l i cation , generality and confirmation o f turbul ence methods

6.3 . 1 The application of this type o f relat ively s imp l e turbulence model
holds great promi se and wi l l probab ly form the basis of a next-generat ion
des ign procedure a lthough it requires a s igni ficant computation e ffort to
so lve the governing partial di fferential equation . The engineer does not

126
normal ly have those procedures at his finger tips , though c learly the
software could be made available so that he woul d j ust have to supply the
geometric and roughness information , p lus the value of NEV to use . It is
this last which provides the problem , because at present the method of
determining reliable values of the turbulence coe fficient is not
estab lished . Also the evidence from the analysi s of the FCF results does
not support the concept of using field observations within bank - nor even
with sha l low f l oo d p lain flows - to provide a s ite-speci fic turbulence
coefficient . The evidence is that it wil l not be constant for a given
geometry and roughness condition , but wil l vary appreciably with depth . So
a i though the basic e quations are general , there is insuffic ient knowledge at
present about the NEV function . In this function i s hidden the complexity
o f compound channel performance .

6.3.2 This review of the current application of the turbul ence e quation to
compound channel s has been l imited to a cons ideration of how wel l a
particular ver sion o f the method fits the large scale laboratory data from
the FCF at Wal l ingford . This l eaves considerabl e uncertainty over the NEV
function even in those ideal geometries , and it was fel t that further
research e f fort was needed , beyond the s cope of the present exercise , into
the best method of assess ing the turbulence coef ficient before looking to
other sources of in formation . It should be realis ed , however , that the
search for a s ing le value of NEV to apply to the who l e cross-section i s not
the only approach that might be fol lowe d . Knight , Shiono and Pirt ( 1989)
al lowed it to vary across the section when analysing results from a natural
river , the Severn at Montford Bridg e . They devel oped analytical solutions
for constant depth and for l inear depth variation , and were then ab l e to use
gauged data at four depths to so lve for the required NEV values in the seven
zones forming the cros s-section ( deep part o f main channel , slop ing sides o f
main channe l , two f l ood p lains , two s loping edges to f lood p lains ) . They l
1\
t

deduced values o f 0 . 07 for the deep part of the main channe l , 3 . 0 for the
f lood p lains , and 0 . 2 for each of the four slop ing boundaries . The degree t
o f agreement achieved with some observed stages and discharges was as
fol l ows , for the Knight , Shiono and Pirt ( 1 989 ) method with varying NEV :
also shown are the results for an overal l section value of 0 . 1 6 :

1 27
 Variab l e NEV Constant NEV of 0 . 16

S tage Observed Cal cu lated Error , % Cal culated Error , %

dis charge discharge discharge
.m ml/s m l /s m 3 /s

6 . 09 3 30 . 8 3 34 . 1 1.0 346 . 0 4.5

5 . 20 220 . 6 229 . 8 4.2 235 . 2 6.6
4 . 73 1 88 . 8 195 . 6 3.6 197 . 3 4.6

So reasonab le accuracy of prediction is feas ib le provided there is

sufficient accurate field data to cal ibrate the particular reach o f river -
though with more field information it might emerge that the zonal values of
NEV derived for the Montford Bridge section are o f more general val idity .

6.3.3 One of the benefits o f turbulence model l ing is that it also provides
an estimate of the distribution o f shear stress and a l s o o f the l ateral
variation of discharge intensity. Its potential is i l lustrated in
Figure 6 . 9 from Knight , Shiono and P irt for these same data sets . The
observed lateral variat ion o f depth mean ve loc ity is shown in comparison
with the calculated variation . There are prob lems adjacent to the bank l ine
where di screpancies are quit e s igni ficant , but agreement within the main
channel and on the flood plain away from the bank is good . Other approaches
to turbulence model l ing a l s o suf fer from probl ems in the r iver bank zone , so
there remains some deficiency in a l l the simpler two-dimens ional approaches .
The potential benefit of knowing the distribution of shear stress around the
perimeter o f the channe l and flood p lain wil l be recogni sed by those
invo lved with .sediment transport and river morpho logy.

6.4 Comparison of turbulence method with empirica l method of prediction .

6 . 4. 1 Turbulence methods even with the restriction o f vertical averaging

provide more detailed information on flow distribution than can be obtained
from the empirical procedures of stage/discharge prediction . However , where
they would be of particular interest near the main channe l bank , the
turbulence methods suf fer from some inaccuracy.

128
6.4.2 In terms o f providing the stage dis charge funct ion for the FCF tests ,
the turbul ence method which uses a sing le valued cross-sectional figure for
NEV proved somewhat less accurate than the empirical method based on the
same data . Tab l e 6 . 1 indicates a best per formance for a l l the smooth fl ood
plain cas e s of -0 . 3 8% average error (which could obvious ly be reduced by
using a value of NEV intermediate between those actual ly tes ted) with
vari ab i l ity of 3 . 7% . The corresponding rough flood p l ain figures are -0 . 3 5%
with variab i l ity 5 . 4% . Thes e standard deviations are not primarily
experimental error s , but represent mos t ly the imper fection in the trend o f
the theory compared with the trend o f the data , and s o indicate a need for
further consideration and refinement . The empirical method achieved a lmo s t
exact agreement with the data used i n i t s calibration o n average , smooth and
rough together without any dif ferentiation in treatment , with a residual
variab i l i ty o f 0 . 8% , part of which i s the tolerance in experimenta l
measurement . The empirical method trans fers satis factori ly to other
geometries and roughnesses , in mos t cases with very satis factory agreement
as shown in Chapter 5 .

6.4.3 The recommended empirical method s eems justified as the firs t choice
at the present time for engineers to app ly , but i t i s clear that the ful l
potential o f the two-dimens iona l l ateral distribution turbul ence model s has
not yet been rea l is ed . I t appears that e ither the value of non-dimensional
eddy viscos ity for use in the method has to be obtained from a s et o f
empirical funct ions bringing i n width ratio , roughness ratio , relative depth
and bank s l ope , or else some di fferent approach to its evaluation has to be
developed from the res earch data . I t is hoped that such deve lopment wi l l
proceed t o a satis factory out come , given the excel l ent and detai l ed data
base now ava i l ab l e .

129
TABLE 6 . 1 : LATERAL DI STRIBUTION TURBULENCE METHOD : ANALYS IS OF FCF DATA

Values of NEV us ed : 0 . 16 0 . 22 0 . 24 0 . 27 0 . 29

Tes t Number
series of tests

SMOOTH FLOOD PLAINS :

01 24 Mean +4 . 5 4 +2 . 1 6 +1 . 37 +0 . 89
S D 6 . 21 3 . 82 3 . 22 2 . 86
02 27 Mean +2 . 9 5 +0 . 04 -0 . 9 3 - 1 . 52
S D 4 . 53 3 . 63 3 . 98 4 . 29 .
03 20 Mean + 1 . 27 - 1 . 81 -2 . 82 -3 . 45
S D 2 . 58 3 . 50 4 . 34 4 . 94
05 6 Mean +2 . 82 -0 . 20 - 1 . 10 - 1 . 67
S D 3 . 46 1 .95 2 . 30 2 . 78
06 18 Mean +2 . 3 0 - 1 . 65 -2 . 9 1 -3 . 7 0
S D 4 . 58 4.34 5 . 18 5 . 79
08 23 Mean +4 . 3 4 +1 . 15 +0 . 1 2 -0 . 5 3
S D 5 . 96 5 . 41 5 . 75 6 . 08
10 17 Mean +5 . 9 4 +3 . 87 +3 . 1 7 +2 . 70
S D 6 . 73 4 . 72 4. 11 3 . 75

Al l smooth : Mean +3 . 49 +0 . 5 8 -0 . 3 8 -0 . 9 8

135 S D 3 . 14 3 . 53 3 . 70 3 . 82

ROUGH FLOOD PLAINS :

07a 1.6 Mean +4 . 25 -0 . 7 0 -1.95

S D 5 .97 3 . 30 3 . 56
07b 4* Mean - 0 . 87 -6 . 59 - 8 . 80
S D 3 . 18 8 . 16 1 0 . 20

09 8 Mean - 2 . 05 -7 . 5 0 -9 . 45
S D 3 . 67 9 . 24 1 0 . 90
11 14 Mean +8 . 26 +5 . 03 +3 . 9 3
S D 1 0 . 3'6 2 . 29 5 .94

130
TABLE 6 . 1 ( cant)

All rough :

*
40 Mean +5 . 5 4 -0. 35 - 1 . 70
S D 5 . 07 5 . 44 5 . 62

Note s : Mean discrepancies and standard deviations ( S D ) a r e percentages .

*
denotes too few data points for averages o f three to be meaningful though
the ' Al l rough ' results did incorporate averaging for series 0 7 a .

131
 • • Measured
0.30
-- Comp NEV=[Link]

0.25 --- Comp NEV=0.080

- - - • Comp NEV=0.1 60

-· ·- Comp NEV=0.240

Bankfull - Comp NEV=0.500

0.10

0.05

..______._ __,___.___.._____.__...L__.___J______J._.....J
0.00 .____.__ __,__...._
.._ .

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 . 7 0.8 0.9 1 .0 1 .1 1 .2 1 .3 1 .4

Discharge (cumeos)

Fig 6.1 Stage-discharge relationship for FCF at Walli ngford ; narrow

flood plai ns, 8/b = 2.2; series
. . .
03 .observations compared with
. .
n r.�n

• .Measured

- Comp NEV=[Link]

- · - Comp NEV=0.080

- - - Comp NEV=0.1 60

-·· Comp NEV=0.240

- Comp NEV=0.500

0.00 L-----1----'-------'---'----l...- .-... --'---..L___----1.. __ _.____.

0.0 0.2 0.4 0.6 0.8 1 .0 1 .2 1 .4 1 .6 1 .8

Discharge (cumeos) JBW/4011 0-111 /30

Fig 6.2 Stage-discharge relationship for FCF at Wallingford ; medium

flood plains, 8/b = 4.2; series 02 observations compared with
p red ictions fro m lateral distribution tu rbu lence method
 1 .1

1 .0 - 8 - 0.50
0

0.9 -
6 0
6
- 0.45
0.8 '- 6 0

�
��
o .P�
0.7 -
- 0.40 ;£
.s::.

�:
.s::.
0.6 - •

:2- - 0.35 �
I

;;. 0.5 - X6 G .s::.

ii.
6 + - 0 •30
Gl

6 + >€>+
-c
Gl
0.4 -
+=I
>
6+ 0 as
X -
0.25 �
0.3 - 0
+ }6. 0

0.2 ,...
++ �6 6
0
xO O
-

-
0.20

-!1: 0
+ + 0 - 0.10
0.1 ,... '6 0
+ 66 -
I I I I I I I I I
6 I I I
0.0
0.90 0.92 0.94 0.96 0.98 1 .00 1 .02 1 .04 1 .06 1 .08 1 .10 1 .1 2 1 .1 4
Residual discrepancy factor
Averaged
data + Series 0 1 6 Series 02 0 Series 03 X Series 05

Fig 6.3 Residual discrepa ncy facto r between o bservatio n and predicti on
u � i ng l ateral distri bution turbulence method; va riou � relative

1 .1

1 .0 0.50
�
0.9
X 6
X 6
0 0.45
0.8
X 6
0
X 6
0.7 X �
;£
X 0 � 0.40

.s::. 0.6
X 6 I
;:::... X 0 6
•

.s::. 0.35 �
X
'
.s::.
;;. 0.5 6 . ii.
X 0
()f. 0 6 0•30
Cl)
-c

0.4 >0 6 Gl
>

)($) 6 �
6 0.25 �
0.3 >45 0
X
0.2
0
0
�6 6 X
X 0.20

0
0.1 0 0.1 0
'6 66
0 X
0.0
0.90 0.92 0.94 0.96 0.98 1 .00 1 .02 1 .04 1 .06 1 .08 1 .1 0 1 .1 2 1 .1 4
Residual discrepancy factor
Averaged
6 Series 02 X Series 08 0 Series 1 0 JBW/41/1 0.D113D
data

Fig 6.4 Residual d iscrepa ncy factor between observation and predicti on
using lateral dist ribution turbu lence method ; vari ous main channel
side slopes; averaged data ; NEV = 0.27; smooth fl ood plai ns
 1 .1

1 .0 f- -
0.50

0.9 f- +
0
0.8 f- X +
-
0.45
D.
+ 0 �
0.7 f- X ::2'
-
0.40 ;.
:§. 0.6 i- X 0
+ •
.

.s::
X +
-
0.35 :X:
•

e. 0.5 i- 0 .s::
ii
X + Q)
X -
0.30 "0
0.4 i- X + �
X + 0 1ii
X +
-
0.25 ;!
0.3 i-
X 0
X + -
0.20
0.2 -

X +
++ 0
-

X
X + + 0.1 0
-
0.1
0
-

+ -

I
0.0 . I I I I I I I I I I I I I I
0.86 0.88 0.90 0.92 0.94 0.96 0.98 1 .00 1 .02 1 .04 1 .06 1 .08 1 .1 0 1 .1 2 1 .1 4
Residual discrepancy factor
Averaged
data + Series 07 D. Series 71 0 Series 09 X Series 1 1
!

Fig 6.5 Residual discrepa ncy factor between observation and prediction
using lateral distribution turbulence method; various main channel
·

ain s
1 .1

1 .0 &
0
0.9 0
0.8
X D.
0.7 �
;.
k i+ er0 0.40
.s:: 0.6 + D. X I
+
•
OD., J:
:2 0.35 .r:
•
e. 0.5 + X
OD. ii
+ + D.
0.30
Q)
"0

0.4 d< + �
D.
�
X 0 D. :i
0.25 &!
0
00
0.3
D. D.4.
0x D. +
0.2 0 D.
�
0 0 If- � + +
0.1 0 D. + +
D.
L-
O.O
0.05 0.10 0.1 5 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 ---L--�--�L---�--�
0.60
NEV
Averaged
data + Series 01 A Series 02 0 Series 03 x Series 05 J!fW/4211 [Link]

Fig 6.6 Optin1um values of NEV for individual stage-discharge

observations as fu nction of depth : variou s relative widths ;
smooth flood plains
 1.1

1 .0 -
- 0.50

1::!. X
0.9 -
1::!.
X
0 -
0.45
0.8 1- 1::!. X
0
0.7 1- 1::!.
X
X
�
�
g. 0,40
-

0 X
.c 0.6 1- 1::!. X I
;::... X •

.c 1::!. 0 -
0.35 :.:.
� 0.5 1-
.c
I

X
a.
-

0 X CD
1::!. <X 0
-
0 30 -o
CD
•

0.4 1- 1::!. X 0 >

1::!. :;:I
X 0 as
1::!. X 0
0 -
0.25 �
0.3 - X
I::!. I::!.'A
0
X
-
0.20
1::!. 0
0.2 - X 1::!. 0 -

X 0
0.1 -
X
1::!.
If i 0 - 0.1 0
X
A X
:!. f::l.
11:--
0
I --�
-

o . o �.�--L-
� --� --� I --�I�---L
I ----L-
I --�I ----�1----�1-----L
0.05 0.10 0.1 5 0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
Averaged NEV
data 1::!. Series 02 X Series 08 0 Series 1 0

Fig 6.7 Optimum val ues of NEV for individual stage-discharge

o bservations as function of depth : various main channel side
_

1 .0 0.50
1::!. +
0.9
0
X 0.45
+
0.8 6.
0
0.7 +
X 0.40
�
�
5. 0.6 0
+ X I
:I:
•

.c X 0.35
� 0.5 0 +
t
X CD
+ -o
X 0 30
•
CD
+ X >
:;:I
0 X as
+
X 0.25 �
0.3 + X
0 + 0.20
X
0.2 + X
+
0 + X
0.1 + 0.10
0 + X
+

0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60

NEV
Averaged
1::!. 0
JBW/4311 0-91fJO
+ Series 07 Series 71 Series 09 X Series 1 1

Fig 6.8 O ptimum values of NEV for ind ividual stage-discharge

o bservations as function of depth : various side slopes; rough
flood plains
 --- Analytical solution X data Stage (m) Q(m3/s)
H = 7.81 0 = 330.80 . ... 6.09 334.1

Q = 330.80 m3s·1

0.00
- 1 00 -80 -60 -20 0 20 40 60
lateral distance V (m)

-- Analytical solution X data Stage (m) .O(m3/s)

2.00
H ... 6.92 Q ... 220.60 .A 5.20 229.8
(ij'
...... .
E
� 1 .50

f Q = 220.60 m3s· 1
� 1 .00
c
eel
G)
E

-[
� 0.50

0.00
-1 00 -80 -60 -40 -20 0 20 40 60
lateral distance V (m)
-- Analytical solution X data 3
Stage (m) O(m /s)
2.00
H = 6.45 0 = 1 88.80 .A 4.73 1 95.6

�
� 1 .50
� Q = 1 88.80 m3 s·1
]G)
>
c
1 .00
· eel
G)
E

i
� 0.50

0.00
-1 00 -80 -60 -40 -20 0 20 40 60
lateral distance V (m)

JBW/M/10-91130

Fig 6.9 Lateral distribution of depth mean velocity: comparison between field
observation and calibrated model of Knig ht, Shiono and Pi rt, 1 989, River
Severn at Montford Bridge (from Knight, Sh iono and Pirt, 1 989)
 HR WallingJ_ord

HYDRAULIC DES IGN OF

STRAIGHT COMPOUND CHANNELS

VOLUME 2

by

P Ackers , Hydraulics Consultant

Contents o f Volume 2

Detai led development o f design method ,

- Part 2
Appendices .

Report SR 2 8 1
OCTOBER 1 9 9 1

Address: Hydraulics Research Ltd, Wallingford, Oxfordshire OX 1 0 8 B A, United Kingdom.

Telephone: 0491 3 5 3 8 1 International + 44 491 353 8 1 Telex: 848552 HRSWAL G.
Facsimile: 0491 32233 International + 44 491 32233 Registered in Engl and No. 1 622 1 74 ·
A l l r i gh t s re s e rved . No part o f t h i s publ i c a t i on may be reproduc ed ,
s t o red i n a re t r i eva l sy s t em , or t r ansmi t t ed , in any f orm or by
any means , e l ec t r on i c , mechani c a l , pho t ocopying , re cording or
o t he rwi s e wi t hout t he p r i or permi s s i on o f H R Wa l l ing f ord .

© HR Wal l ingford, 1991

Thi s r e p o r t d e s c r ib e s the deve l opment o f new and imp roved de s i gn p r o c edur e s
f o r two - s t a g e ( compound) f l o o d channe l s . Thi s work was c a r r i e d out by P e t e r
Acke r s a s c on s u l tant t o H R Wa l l ing f o r d , with funding made ava i l ab l e b y the
Reg io na l Wat e r Author i t e s in 1 9 6 6 , p r i o r t o the i r demi s e when the i r
r e spons ib i l i t i e s in t h i s c on text pas s e d t o the National Rive r s Author i t y .
The s e funds wer e p rovided f o r the b e t t e r d i s s emina t i on o f r e s e a r ch r e su l t s
on t h i s subj e c t i n t o engine e r in g p r a c t i c e .

The r ep o r t i s in two vo lume s . The f i r s t b e g in s with a Summary and D e s ign

Method wh i ch e f f e c t ive ly provides a Manu� l f o r the hydrau l i c d e s i g n o f
two- s ta g e channe l s . The d e t a i l e d r ev i ew s upp o r t ing the s e new p r o c edure s
f o l l ow s , c o n t inuing i n t o vo l ume 2 , which a l s o c o n t ains s ev e r a l App e nd i c e s .

The hyd r au l i c engineer wi l l f ind the e s s ent i a l i n f o rma t ion in the f i r s t

s e c t ion , Swnmary and D e s ign M e tho d , but wi l l p r ob a b l y wish t o r e f e r t o s ome
o f the d e ta i l s g iven in the ma in b o dy o f the r e p o r t and in the App endi c e s t o
ext end h i s unde r s t anding o f the c omp l ex b ehav iour o f two - s t a g e f l o o d
chann e l s .

Appendix 7 p ro v i d e s a d e s ign examp l e o f the c omputat ion p r oc edur e s ,

inc l ud ing t ab l e s ind i c a t ing how ob s e rve d s t a g e - d i s ch a r g e data might be us e d
t o extend t h e s t a g e - d i s ch a r g e func t i on . The s e t ab l e s wi l l a l s o provide u
c r o s s - check f o r any c omput e r p r o g r amme deve l o p e d t o s o lve the re c ommended
hydrau l i c e qu a t i o n s and l o g i c p r o c e dure s .

I t i s s t r e s s � d that the e qua t i on s g iven in t h i s Manua l a r e f o r the hydr a�. l i c

das ign o f s tr a ight par a l l e l two - s t a g e c onveyance s , a l though informat io n
wi l l b e found extending t h e app l ic at i on t o sma l l ang l e s o f skew ( no t
exce ed ing 10 ° ) , I n f o rmat ion g iven o n meand e r in g channe l s in Chapter 8 o f
the main t ex t ( s e e volume 2 ) shows that they b ehave qui t e d i f f e r e n t l y .
Improvement s i n the hydraul i c c a l c u l a t ions f o r meand e r e d and i r re g u l a r
channe l s mus t awa i t �urther work .
 DETAI LED DEVELOPMENT OF DES IGN METHOD

CONTENTS

PART 1 ( See VOLUME 1 )

Page

1. INTRODUCTION

1.1 The importan c e o f compound channe l s and over -bank f l ows

1.2 Scope o f treatment : s t ra ig h t channe l s , the addit ional
p r ob l ems p o s e d by skew channe l s , meand e r i n g and curvature
1.3 Appr oach to d e s i gn : tradit iona l c a l cu l a t ion methods f o r
s t a g e / d i s cha r g e in compound channe l s

2. FLOW RE S I STANCE IN CHANNELS OF COMPLEX CROSS-SECTION

2. 1 Re sume o f r e s i s tance for s imp l e open channe l s :

ava il a b l e formu l ae and the i r r e l evance and l imitations
2.2 Compound c r o s s s e c t ions : var i at ions in hydrau l i c
parame ter s , f i e l d and l ab o ratory c a s e s ; inapprop r i ate
to treat as s ing l e c r o s s - s e c t ion ( examp l e s ) ; the
summat ion method and var iant s in the l i te ratur e ; cho i c e
o f ve r t i c a l divis ion
2.3 Al l owing for the e f f e c t s of interact ion : r e sume o f
approaches in l it e r a ture and their l imitations ; for ce
ba lance and inter f a c i a l shear ; exper imental s tud i e s
and typ i c a l r e s u l t s ; m o r e fundament a l method s b a s e d o n
turbul ence theory a n d p r e s ent l imitat ions .
2.4 Features in f l uenc ing the d e g r e e o f intera c t ion : app roach
from dimens i on a l ana l ys i s ; dependence on r e l a t ive
roughn e s s e s of f l ood p l ain and main channe l ; on width
r a t io ; o n dep th r a t i o ; on b ank s lope ; c ro s s - s e ct ion
coheren c e .

3. HYDRAULIC DES I GN BASED ON EXPERI MENTAL ADJUSTMENT FACTORS

3.1 Research at Wa l l ing ford ; s cope o f exp e r iments and

measurement s made .
3.2 Other s ourc e s o f exp e r imental data
3.3 Rec ommended bas ic method : s eparate c a l cul ations for
channel and f lo o d p l a in , then summe d , and corrected for
interact ion e f fe c t s
3.4 Ana lys i s o f exp e r imental r e s u l t s : f l ow reg ions ; channe l
coherenc e ; inf luence o f : f l ow depth f l ood p l ain width
channel s ide s l ope asymmet ry with support ing p l o t s and
emp i r i c a l r e l a t ions
3.5 S eparation o f main chann el and f l oo d p l ain e f fe c t s
3.6 I n f luence o f f l oo d p l ain roughne s s : form o f roughne s s used
3.7 Hydrau l i c de s i gn formu l ae : formu l ae for inter ference
e f f e c t s in d i f f er ent f l ow r e g i on s ; cho ice of reg ion ;
goodne s s o f f i t to exp e r imental data
CONTENTS ( CONT ' d)

4. SKEW CHANNELS

4. 1 The importance o f momentum trans fer with non-al igned flow

4.2 Research on skew compound channel s ; increased e f fect of
interference due to skewnes s
4.3 Extension o f design method to skew channe ls ; l imitations

5. OTHER SOURCE S OF STAGE-D I SCHARGE DATA FROM COMPOUND CHANNEL

EXPERIMENTS

5.1 Allowance for width/depth ratio in general ised predictive

functions
5 .2 Other sources of res earch data ; s cope and l imitations ;
methodology ; preliminary analysis ; difficulties ; data
sources and review
5.3 Summary o f information from other laboratory reseach
5.4 Skew channels
5.5 Field information ; rivers for which data were available ;
comparison o f calculated s tage-dis charge above-bank with
f i l ed observations .
5.6 Conc lus ions from other data sources

6. TURBULENCE METHODS

6. 1 Resume o f turbulence theory as current ly applied to

compound cro s s-sections . Internal f luid mechanics .
6.2 Turbul ence methods and comparison with FCF data .
Review o f two-dimensional methods , non-dimensional
eddy viscosity and its evaluation , comparison with
experiment
6.3 Appl ication , genera l ity and confirmation of turbulence
methods . Approaches to asses sing NEV , need for
calibration
6.4 Comparison of turbul ence method with empirical method of
prediction . Present l imitat ions and potential

PART 2 Page

7. ANCI LLARY TOP ICS

7.1 Appl ication to more complex cros s sections ; parameter 1

de finition
7.2 Shear stres s ; experimental evidence of variation around 3
perimeter ; method for assessment o f typical values
in channel
7.3 Critical flow , energy and water l eve l s ; Froude number 6
CONTENTS ( CONT ' d )

Page

7.4 S ources o f bas i c informat ion on roughne s s ; 8

L ined channe l s
Natural r ivers
Grav e l b e d channe l s
Sand b e d channe l s
Ve g e t a t i on

7.5 Need for and ut i l i s at i on o f f i e l d data ; ext r apo l at ion o f 16

s t a g e / d i s charge func t ion
7.6 Incorporation into num e r i c a l mode l s : one-dimens iona l ; 18
two-dimens i o na l : s ep a r a t e zone p r o c e dure vi s - a-vis
lump e d channel pro cedure

8. IRREGULAR P LAN FORM 21

8. 1 Features o f meander ing f l ows in-bank . Characte r i s t i c 21

g e ometry o f natural chann e l s , s inuos it y , p l anform l o s s e s ,
r e duct ion in conveyance there from , s e c ondary currents
8.2 Above-bank f l ows in meander ing channe l s . Review o f 25
exis t ing informat ion , reve r s a l o f se condary current s ,
momentum and f l ow exchange , inter ference and e f fe c t
on conveyance
8.3 F low mode l s for s inuou s , meandering and i r r eg u l ar channe l s . 32
Hydrome chani cs app roach , momentum appr oach , f ine g r i d
mode l l ing

9. SED IMENT TRANSPORT 37

9. 1 Gener a l aspect s o f s e d i ment tran sp o r t . Mode s o f transport 37

9.2 Transport pro c e s s and theory . D imens ional anal ys i s , 38
emp i r i ca l fun c t i ons , inc luding Acker s and Whi t e ; suspended
load d i s t r ibution
9.3 The i n f luence of compound f l ow on bed mat e r i a l t ranspo r t . 42
Typ i c a l r iver sec t ion , method of c a l cu l at i on for sand and
gr ave l , l o s s of t ranspo r t capa c i ty above bank fu l l ,
p e r formance o f equiva l ent s imp l e channe l , imp l i cations f o r
f l uv i a l morpho l ogy .
9.4 Suspended s o l i d s in compound channe l s 47

10 . CONCLUD ING REMARKS 48

1 0 . 1 Summary of hydrau l i c de s ign f o rmulae f o r the conveyance of 48

s t r a i ght compound channe l s ; app l i c ation l o g i c ; t o l e r ance on
a s s e s sment ; l imitat ions
1 0 . 2 The advantages o f compound channe l s : envi ronment a l ; 55
hydrau l i c ; ma int enance
10 . 3 S tate o f know l e dg e and need for further r e s earch 57
CONTENTS ( CONT ' d )

Page

11. ACKNOWLEDGEMENTS 60

12 . REFERENCES 61

13 . NOMENCLATURE 72

APPENDICES

1. Dimens iona l analys is app lied to compound channels

2. Res istance funct ions for the SERC-FCF at Wa l l ingford
3. Coherence
4. Turbulence method: so lut ion for general cross section
shapes
5. Data on channe l roughnes s
6. Examp le of channe l geometry conversion and stage dis charge
comput ation
7. Ana lyses o f other sources o f laboratory data ; tabular
summaries
7. ANCILLARY TOP ICS

7.1 Appl ication to more complex sect ions .

7.1. 1 Natural river cross-sections and also many artificial or " engineered"
two-s tage channel s differ in shape from the c lassic compound trapezoid for
which mos t of the research evidence is available . Their berms , or f lood
pl ains , are l ikely to have a cross fal l and the main channels of natural
rivers are seldom of simple trapezoidal shape . Their beds may irregular ,
deeper on one s ide than the other ; and the ir banks may not be trimmed to an
even gradient . Despite these complexities o f form , the hydraul ic engineer
has traditiona l ly handled rea l cross-sections us ing the basic parameters of
cross-sectional area and wetted periemter , whi ch j oint ly provide a measure
of hydraulic mean depth , R = A/P . What is required is an extens ion o f the
basic methods of handling comp l ex cross-sections so that the methods derived
from research on " classic" sections can be app l ied in practice .

7.1.2 As the recommended method s tarts from the basic computation o f flows
in the lower-stage main channel and the upper-stage f lood p lain s eparated by
vertical divisions , using conventina l friction formulae , there is no prob lem
in terms of the basic computat ion : the "real " cros s section can be used ,
with appropriate areas , wetted perimeters and hydraulic mean depths of the
zones of flow . The prob lem arises sole ly from the need to simp l i fy the
section geometry to deduce the va lues o f several o f the independent
variab les contained in the adj us tment equations , particularly for Region 1

flow , the sha l lower range of depths o f flood p lain inundat ion . The re levant
geometric variables to be defined are :

h - main channel mean depth

H - depth of flow rel ative to mean bed l eve l , hence H * = ( H-h) /h
we - hal f top width of channe l
B effective hal f width of val ley floor at f lood p lain level
-

s e - bank s lope
N - number of flood p lains
F

7. 1.3 Reasoning that the interact ion effect is mainly dependent on

condi tion adj acent to the bank l ine o f river , H * has to be defined so that

1
( H-h) i s the flow depth on the flood pla in at the r iver edge , not an average
depth assuming the flood plain to have a cross fal l . we is probably the
mos t obvious of these geometric variables : the tops of the river banks
define the vertical divisions between main channel and berm flows , and the
distance between is obvious ly 2w . The bank s l ope i s less readily defined
c
as the bank itse l f may be formed of a compound s lope or curve . From the
engineering point of view , what is required i s a representative value and i t
i s suggested that the way to achieve thi s i s to plot the actual
cross-sect ion and " eye-in" an average bank s lope at each s ide matching the
upper two thirds , say , of the actual bank profil e . This is i l lustrated on
Figure 7 . 1 . s e i s then the average value of the left and r ight bank
figures . Having identi fied wc and s e • the mean bed leve l i s a l s o fixed , by
the requirement that the area of the trapezoid so defined i s the same as the
true channel cross-section .

7.1.4 The number o f flood p lains or berms wil l usua l ly be s e l f-evident , and
s o this leaves only B to be defined. For horizontal flood plains , for the
analysi s of the experimental data , B was hal f the total width between the
outer l imits o f the berms . Where they are s loped , thi s is c learly the mos t
appropriate de finition when the flood p l ains are inundated over their ful l
width . However , with partial inundation o f the flood plains , the flow
"knows nothing" o f the dry part o f the cross-section , so that for partial
inundation the value o f B i s hal f the effe ct ive width of the above berm
flow , i . e . hal f the actual water surface width . This can be de fined from
the " real " geometry at any flow stage . These procedures for defining the
geometric parameters are i l lustrated in Figure 7 . 1 . (The use of b as the
semi-channel bed width and B as the semi above-berm width stems from the
terminology adopted as s tandard by the teams o f researcher s us ing the FCF at
Wal l ingford . I t was considered preferab l e to r etain these de finitions in
the present pub l i cation , whil s t str e s s ing their special nature in the
engineering context . w is a l s o a semi-dimens ion . )
c

7.1.5 The discharge adj ustments in flow Regions 2 , 3 and 4 are based on the
channel coherence , COH , which is explained and defined in Chapter 3 , paras
3 . 3 . 4 to 3 . 3 . 6 . These def initions and the formulations o f equations 3 . 1 ,
3 . 2 and 3 . 3 are general and can be appl ied to the real section , however
comp l ex , or to a s imp l i fied section following the der ivation of the previous
paragraphs . The value o f coherence der ived wil l not be very

2
sensitive to the method used , which can therefore be chosen for convenience
of cal culation .

7.2 Shear Stress

7.2. 1 The variation o f shear stress around the boundary of a compound

channe l was i l lus trated in a qual itative way in Figure 2 . 9 . Shiono and
Knight , ( 1990b) , provided a valuabl e picture o f the various proces s e s at
work , including the boundary shear stress distribution , reproduced as
Figure 7 . 2 . In the absence of l ateral shear and secondary flows , the
distribution o f horizontal shear in the vertical is l inear , varying from
zero at the water surface to pgyS at the bed . However , Figure 7 . 2 shows
that momentum trans fer at the interface and also secondary circulations may
modify the basic depth-rel ated distribution of s tress on the solid boundary
by br inging to it some higher - or indeed lower - ve locitie s . Hence the
shear stres s distribution is complicated by severa l processes aris ing from
the interaction between main channel and f l ood plain zones .

7.2.2 Knight , Samuel s and Shiono ( 1 990) analysed some ear ly results from
the research on the FCF showing the vertical distributions of shear stress
at positions acros s the channe l , for a particular f l ow depth , see
Figure 7 . 3 . There is reasonab l e approximation to the "norma l " l inear
var iation with depth at the centre l ine (Y = 0 , where Y is the distance from
the centre l ine ) , and towards the edge of the flood plain (Y = 1 . 5 ) , but
there are maj or departures over much of the width , e specially in the region
of the s l oping banks . Clearly the conventional formula for the shear stre s s
o n any horizontal surface , " = pg ( y z ) S does not apply (y = flow depth , z
H
= vertical distance from bed of point o f interes t ) . Shiono and Knight
( 1990b ) cont inuing analysis o f the same source of data plotted the boundary
shear stres s , � B ' in the form of the difference from what might be
considered a s tandard value , • = pgyS :
0

Re lative change in shear stre s s , 0�* = (� - � Bo) /� • • • 7.1

0 0

where o = {( l + 1/s2 ) • . . 7.2

s be ing the l ocal cross-s lope o f the bed . o is thus an a l lowance for the
fact that where the boundary has a cross s l ope its hor izontal component o f
l ength de fines the shear act ion on the column o f water above . Shiono and
Knight ' s results are i llustrated in Figure 7 . 4 , for three flood plain widths
3
and a range of re lative depths , H* .

7.2.3 Although the se p lots are at too small a scale to be used directly in
design , the information therein is very significant . Within the main
channel , o � * is positive indicating a reduction of shear stress from its
"norma l " value , and with o � * = approx 0 . 1 5 to 0 . 3 5 , the reduction is
important , for examp le in the context o f sediment movement . Over the flood
pl ains , o�* is negative and s o indicates an increase in shear stress over
the normal va lue , pgyS , again by a signi ficant proportion even remote from
the channe l bank line with relative wide flood plains .

7.2.4 For the particular geometry upon which Figure 7 . 4 is based , the
channel bed extends to Y = 0 . 75m , and the bank top is at Y = 0 . 9m . The bank
top shows a considerab l e increase of stress over the normal value , with 0 � *
ranging up to and even beyond 4 . Thi s s igni fies that the shear stress
loca l ly at the edge of the channel bank with shal low flood p lain depths , H*
= 0 . 1 approx . is five times its normal depth-based value . Thi s arises
because the high ve locity within the main channel spi l l s on to the berm , and
this spi l l age effect extends some distance across the f lood p lain , perhaps
to Y 1 . 3m , i.e. up to 3 t imes the channel depth o f O . l Sm beyond the bank
l ine . At the base o f the s l oping channe l bank , Y = 0 . 75m , the positive
value of 0 � * is rather above that at the centre l ine , indicating a rather
l ower actua l shear stre ss . This is characteristic of shear stress
di stribution in trapezoidal channe l s , it diminishes towards the re-entrant
corner , and in theory would drop to zero if the corner was truly sharp and
there were no secondary current s . So over the depth of the s l oping main
channel bank , the shear stress distribution passes from a "below normal "
value t o an "above normal " value , very much above normal a t sha l l ow overbank
flows .

7.2.5 In broad engineering terms , the reason for the s igni ficant reduction
in bed shear stress in the main channe l below the value given by pgyS is
that the component of weight down the stream gradient is only partly
ba lanced by the boundary shear stress . With a two-stage channel , the
interact ion between the flow zones gives additional stress on the interface
between main channe l and flood plain , and also the secondary circulations
and the turbulence arising from momentum exchange change the flow structure
from that in a s imp le channe l . As a first attempt to quant ify the magnitude

4
o f the e ffect , it might be reasoned that the mean bed shear s tress wil l
approximate t o that which would occur with the same mean ve locity . The
discharge , as we have a lready seen , is reduced b e l ow the bas ic calcul ated
f igure for the main channe l considered separately by a factor , DISADF , that
C
depends on the f l ow geometry and roughnes s e s of the zones , but which is
calculab l e . The mean velocity reduces by the same factor , o f course , and
with a s quare l aw of boundary drag , as in the Manning and rough-turbulent
:I
e quations , the resultant mean boundary shear stress is proportional to V •

Hence , to a first order o f approximation , one might expect that the mean
shear stress on the main channel bed wou l d be given to a sufficient
approximation for engineering purposes by:

• • • 7.3

or by:

• • • 7 .4

depending upon whether the channel may be considered wide or not .

7.2.6 From the detailed measurements o f shear stress (using a Preston tube)
in the FCF program of research , the ave rage bed shear stresses were
establ ished for the range of test conditions , though here only the results
for varying flood plain width are cons idered , with channe l bank s lope , s e =

1 , and smooth channel and f l ood p l a ins . For thes e smooth conditions , the
s quare law o f rough turbul ence doe s not Strictly apply , but in Appendix 2 ,
eq . 3 . 6 , it was shown that a power law o f 1 . 8 woul d be appropriate . The two
equations above can therefore be modified by providing D ISADF with the
C
exponent 1 . 8 as an a l ternative . Thus , us ing the procedures for calcu l at ing
the discharge adj ustment factor , with the logic of select ing regions and the
approach to the separation of the zonal adjustments for Regions other than 1
as exp lained in Chapter 3 , para 3 . 5 . 10 , theoretical values o f • can be
BAV
c a l culated for comparison with experiment . F igure 7 . 5 shows this , with the
upper diagram for test series 02 ( see Tabl e 3 . 1 for the geometry) . Both
methods of calculation , using the hydrau l i c mean depth of the main channe l ,
R ' and the water depth , H , were used , coupled both with the s quare l aw
C
exponent of 2 and the smooth law va lue o f 1 . 8 . One would expect the data to ·

l ie between the two theoretical graphs for exponent 1 . 8 ( shown as ful l

5
l ine s ) and indeed they do . The observed data T l ie fairly close to but
BAV
above the plot based on h . m . d . , R ' and as the p lotted function is rea l ly an
C
indication of mean shear stress around the who le solid perimeter , it is to
be expe cted that the mean value on the bed wil l exceed this . Test series 0 1
and 03 at different B/b ratios are shown in the l ower part o f Figures 7 . 5 ,
and the picture remains much the same , the observed mean bed shear stress
lies between the values cal cu lated on the basis o f f l ow depth and on the
basis of h . m . d . , lying nearer to the latter . It appears that the s imp le
procedure incorporated in equations 7 . 3 and 7 . 4 above straddle the true
value of mean bed shear stres s , whi l s t expl aining the bulk of the departure
This calculated adjustment , D ISADF l • e ,
from the "norma l " value , pgyS .
c
accounts for a reduction o f up to 3 0% in this particular test serie s .

7.3 Critical f l ow , energy and water l eve l s

7.3.1 Critical f l ow i s usua l ly defined i n s tandard hydraul i c s textbooks as

the flow condition in an open channel when the specific energy for a given
discharge is at a minimum , and for which maximum discharge occurs for a
g iven energy leve l . It a l so indicates a change in flow s tate , in that sma l l
surface disturbances wi l l trave l upstream with sub-critical conditions but
cannot do so with super-critical conditions . It i s this l atter criterion
that makes the concept o f critical flow of particu lar s igni ficance in
numerical calculations of non-uni form or non-steady f l ows . The theory o f
critical f l ow i s deal t with a t some l ength b y J aeger ( 19 5 6 ) including the
proo f that whether energy or momentum i s considered the same conventional
definition of critical f low in an open channel o f general cro s s-sectional
shape app l ie s provided it may be a ssumed that the velocity
distribution is uni form . This l eads t o the conventional de finition o f
Froude number , F r = V/ ( gA/W) , where V i s the mean velocity o f flow , A the
cross-sect ion area , W the water surface width and g the gravitationa l
acceleration . Critical f l ow i s when Fr = 1 .

7.3.2 The assumption o f uniform velocity distribution may not be an

unreasonab l e approximation for s imple cross-sections but it i s clearly
inadmis sab l e with compound channel s . The ve locity variat ion across the
section can be described by a or � . depending on whether one is concerned
with energy or momentum , and the incorporation of these factors into the
energy and momentum equations then gives differing formulations for the
Froude Number , Fr :

6
Energy basi s :

r, aw Q2 ]
I �A
da
-
Fr dd) 2gA2 • • • 7 .5

Momentum basis :

,- tlli - � � 1'
Fr I l( A dd) 2gAl ] • • • 7.6

whi ch revert o f course to the conventional definition for a = � = 1. The

appearance o f the water surface width , W , in the above functions indicates
that in a channel with hor izontal berms , there wi l l be a dis continuity in
the calculated Froude number/ s tage function for a g iven channel gradient ,
and there could be dual ity in the critica l condition in more general case s .

7.3.3 Knight and Yuen ( 1 9 9 0 ) carried out experiments to examine and compare
aspects of critical depth in a compound channel with b = h = 7 5 mm , B = 225mm
and s e = s = 1 , with variab l e s l ope , and for a range o f relative depths ,
F
0 . 05 < H* < 0 . 5 . They were concerned not only with the concept o f an
overa l l value for the Froude Number but a l so with its local variation across
the channe l . With depths and ve locities being measured at many verticals
across the width , they were ab le to assess the loca l values o f Froude
number , U/{ ( gy) , and specific energy , E = y + U 2 / 2 g wher e U is the depth
mean velocity at any vertical . I t i s worth mentioning at this s tage that
the water level i s the same at each point across the s ection , no doubt
because with an a ligned system o f f low there i s hydrostatic pres sure
throughout . The lateral variation o f Froude number when the overal l flow i s
critical i s i l lustrated i n Figure 7 . 6 . This confirms that there can be
local zones o f super-critical ve locity on the berms near the channel bank
l ine , induced by the increase o f dis charge intensity due to lateral shear ,
a lthough on the berms away from the bank l ine the flow i s sub-critica l , as
i t is within the central deep s e ction .

7.3.4 Petryk and Grant ( 19 7 8 ) examined methods o f c a l culating the Froude

numbe r in compound channe l s , referring to field observations of surface
disturbances that c lear ly indicated a variation o f Froude number across the
section . They were seeking exp l anation for the observation o f a pattern o f

7
surface waves in the main channe l s of flooded rivers , when overal l the flow
might be expected to be sub-crit ical . With cross- sections more typical o f
natural rivers than those tested by Knight and Yuen , there can obvious ly b e
condi tions where the high ve locity in the main channe l can yie l d locally
high Froude numbers when the sha l low depth and roughnes s o f the flood p lain
render the flow there sub-crit ical . There is a l s o the condition a lready
ment ioned when the penetration o f fas t main channe l flow on to the edge o f
the flood p lain can generate a pattern o f surface waves on the berm its e l f
because there the depth is shal low , so increasing the Froude number above
the main channel value . Viewed in the context of surface wave patterns ,
there are c learly different possib l e combinations according to the local
va lues o f Froude number acro s s the channe l width .

7.3.5 Samuels ( 1 989) includes a review o f the influence o f Froude number on

numerical mode l l ing , and how it might proper ly be calculated incorporating
va lues of the momentum coe fficient , � . The subj ect is a compl ex one and
further research is required for a ful l understanding . For hydraul i c
engineering purposes , the important point is perhaps that the simple "text
book" definition o f Froude number no longer app l ies to compound channel s ,
and that with a knowledge of the separate f l ows as calculated by the method
given earlier in the Manual , approximate values for the main channe l and for
the f lood p lain zones could be calculate d . They wil l not be the same as the
overal l section value but are probably more relevant for engineering
purposes .

7.4 Sources o f basic information on roughnes s

General

7 . 4. 1 The main resistance functions used in open channel design are the
Manning equation and the Colebrook-White equation . The former is for rough
turbulent flow and so should not be used for relative ly smooth construction
materia l s , such as good qual ity concrete l ining ; the latter is for turbulent
conditions embrac ing a l l surface conditions from smooth to rough , so is more
general . However , the choice of e quat ion can not be separated from the data
base available on the roughne ss coefficient . The Manning equation has been
so widely used in engineering practice that extens ive l i sting s o f the
coefficient value , Manning ' s n , are availab l e in the l iterature , based on

8
the body o f experience in the use o f that e quation in hydraul ic des i gn .
Ven T e Chow ( 19 5 9 ) g ives such information for a who le range o f construction
material s , through metal , wood , brick , masonry and concrete , to channel s
excavated in earth , grave l and even rock ; natural streams in the mountains
or p l ains , weedy reaches and a l s o various ly des c r ibed f l ood p lains . Thes e
values are listed i n Appendix 5 , Tab l e A5 . 1 .

7.4.2 The roughness coef ficient in the C o lebrook-White funct ion is l e s s

empirical i n that it has a phys ical bas is , namel y the textural roughness of
the surface referred to as an " e quiva l ent sand roughne s s , k " • the d iameter
s
of grains forming a p l ane granular surface that would provide the same
resistance . This fundamental concept has been extended over the years to
incorporate empirical information from a wide range o f surfaces and
construction materials , including typical values for rivers and gravel bed
streams etc . Tab le A5 . 2 in Appendix 5 gives values for concrete and some
other materia l s .

7.4. 3 The Manning equation wi l l normal ly be used for natural channel s , and
for rivers in an " engineered" condition . The methods g iven above apply only
to straight , or very g ently skewed or curved channel s ( limiting de f l ections
s ay 1 0° ) , and wherever poss ib l e the roughnes s coe fficient used should be
based on actual measurement of the river under study . It is usua l ly found
that the Manning ' s n value varies in a systematic way with stage and
Figure 7 . 7 i l lustrates the resu l t s o f field measurements in five Scottish
rivers , by Sargent ( 1 9 7 9 ) . The coe f ficient value reduces with increase o f
stage , possibly because there are typical ly features in the bed , such as
bars , shoa l s and scour holes , that have a greater influence at sha l low
depths than at stages approaching bank ful l . The value o f main channe l n to
use for above-bank f l ows woul d b e the value obtained when f low is j ust below
bank-ful l . However a coe fficient variation o f the form shown in Figure 7 . 7 .
can a l so arise where the relative roughness is somewhat over-severe for the
Manning equation to app ly , when the C o l ebrook-White equation might prove
more robust .

7 .4.4 A convers ion between the Manning coefficient and the value o f k ( in
8
metres ) is availab l e through the formula :

. . . 7. 7

9
but the Manning equation is only theoretical ly correct where 7 < R/k s < 130 ,
so Manning might not be expected to provide a good fit to measured data
when :

• • • 7.8

This suggests restrictions at sha l l ow flow depths in typical rivers .

Gravel bed ri vers

7.4.5 The dominant size of sediment found in the beds o f alluvia l rivers is
related to their gradient , so th,at s teep rivers in mountainous terrain have
beds of boulders and coarse gravel , those in the sub-montane region wi l l
have grave l beds with some s and , and in the p lains beyond wil l have sand and
s i l t beds . A particular feature o f coarse bed streams is the wide range of
sediment s izes found in them - and being transported through them .
Considerab l e sorting i s observed between different parts o f the stream bed
as we l l as in depth . Armouring frequently occurs , where a one or two grain
thick l ayer of coarse material overlies the bulk o f the bed with its mix o f
a wide range of s izes . This armour layer i s l e ft by decreasing flows after
a f lood event , by the winnowing out of finer material when the flow is no
l onger competent to move the coarsest fraction . This layer then protects
the underlying material with smaller n s ize , until a flow l arge enough to
50
initiate motion in the armour l ayer occurs , so triggering rapid transport of
the sediment forming the bulk o f the bed .

7 . 4. 6 The resistance of boulder and gravel bed rivers is associated with

the texture o f the bed arising from the coarser fractions of material there ,
and s o current methods use various modi fi cations o f the rough turbulent
equat ion , {which is one of the limiting reg ions of the Colebrook-White
function ) , relating the l inear roughnes s of that equation , k5 , to the ·bed
material size . A variant o f that is to use the S trickler form o f the
Manning equation , with its l inear measure o f roughnes s determined from bed
grading ( Strickler , ( 1924) ) . The Limerinos { 19 7 0 ) equation was based on
Cali fornian data , and e ffective ly incorporates the convers ion into Manning ' s
n o f the k s va lue that woul d be used in the rough turbul ent function :

n = 0 . 1 1 3 R l ; & / [ 1 . 16 + 2 l og ( R/D )J • • • 7.9

m 84

10
o is the grain s ize for which 84 percent o f the bed materi a l i s fine r .
84

7 . 4.7 Bray ( 19 8 2 ) reviewed the r e s i s tance o f gravel bed r ivers , general ly

confirming the Limerinos function . Hey ( 19 7 9 ) e f fectively used a modified
form of this rough turbulent equation which included a cross-section shape
p arameter . There have been two international conferences dealing
specifica l ly with g ravel bed r iver s , from the morphologica l ,
s edimentological and hydraul i c points o f view , and the proceedings o f thes e
conferences provide an excel lent s tate-o f-the-art summa ry : Hey , Bathurs t and
Thorne ( eds ) ( 19 8 2 ) ; Thorne , Bathurs t and Hey ( eds ) ( 1987 ) .

Sand bed channels

7.4.8 In laboratory expe r iments starting with a p lain sand bed , once the
flow conditions are abl e to generate sediment movement , r ipples or dunes
wi l l form . The normal condition in nature i s a l s o for s imilar features to
form on the bed : a p l ane bed i s an unusual condition and i s more l ike ly to
o ccur at high transport rates when the stream velocity is high enough to
wash out the p re-exis ting features . The pres ence of bed features means that
the overa l l resistance of the bed wil l comp r i s e both the drag due to the
obs t ruction of ripples or dunes ( form drag) and the r e s i stance o f the
granular texture itse l f ( grain r e s i s tance ) .

•
o
= T form + • •

g ra�n
• • . 7 . 10

where

•
0 p gyS . . . 7 . 11
y flow depth
S hydraul i c gradient

7.4.9 The grain r e s i stance for coarse material can be e s t imated from the
rough-turbulent e quation , as was noted for g rave l bed river s , but for s and
bed r iver s the subj e ct i s considerably comp l i cated by the exis tence of bed
features . Ripples and dunes and combinations of them are known as " lower
regime" and the high transport p l ane-bed region of rapid f low , together with
the anti-dune condition that can arise in s teep channe l s at high Froude
numbers , forms the "upper regime " . The dis t inction between lower and upper

11
regime is not c l ear cut : there is a trans ition between them as ve locities
increase and it is possible for different parts of the bed to be in one
regime or the other , or to be somewhere between , when flow conditions are
not clearly one side or the other of the dividing criter ion . Features are
dependent also on sediment size : g rave l bed rivers do not have r ipples and
have shoal s rather than dunes .

7 . 4 . 10 White , Pari s and Bettess ( 1980 ) used the same parameters as appear
in the Ackers and White ( 1973 ) sediment transport cal culation method , for
assessing the resistance of rippled and duned sand bed rivers , and it is
poss ib l e to combine these functions in g iven circumstance s to assess
suitable va lues of Manning ' s n , as i l lustrated by Acker s ( 1980) for
irrigation canal s . Although the method has been shown to be r e l iab le and
forms the basis of modern de sign procedures for sand bed irrigat ion cana l s ,
it i s too complex to cover in detail here . Bettess and Wang Shiqiang ( 19 8 7 )
a l s o used the same sediment parameters to study upper regime bed form
resistance , and the transition between upper and lower bed forms , but again
it would be inappropriate to detail their procedures here . Suf fice it to
say that typical Manning ' s n value s for straight sand bed channe ls are in
the range 0 . 0 2 2 to 0 . 040 depending on s ize of channe l and s ize of sediment ,
but maj or sand bed rivers can show considerabl e variation in time s of severe
f lood if the bed of main channe l goes through the trans ition from r ipples
and dunes to plane bed . This was i l l ustrated from the river Indus by Hogg ,
Gugenasheraj ah , Gunn and Acker s ( 1988) ) , us ing flood data for 1976 and 1986 ,
showing a reduction in n to about 0 . 0 1 1 as the dunes are washed out and the
m
bed becomes p lane , later reve rting to a duned bed with n about 0 . 03 . The
different bed forms possible in sand bed r ivers are thus of s igni ficance to
hydraulicians , though within UK few r ivers would come into this category ,
many having e ffectively r igid beds .

Vege ta tion

7 . 4. 11 River vegetation fal l s into three categories : mid-channe l aquatic

weed ; channe l edge growth (gras s , reeds , wil lows etc) ; and banks ide/flood
berm vegetation ( pasture , growing crops , orchards , trees , shrub s , hedges
etc ) . This rich variety is environmental ly desirable but it inevitab ly has
an influence on the hydraulic performance of the system. Moreover it varies
seasona l ly , and so asses sment of the roughne s s coeffic ient can not be

12
cons idered an accurate s cience . Clearly , past experience based on
measurements at the site of interest wi l l provide the best guide , though o f
cour se any seasonal changes mus t b e borne in mind . Research has also
provided important sources of information , though again caution i s required
in trans ferring results from one geographi c zone to another , which may
support different flora .

7 . 4 . 12 The mos t extensive work on grass comes from America , and i s

described b y Kouwen , Li and S imons ( 19 8 1 ) . The method i s to identi fy a
retardance c lass based on a US Dept o f Agr iculture c l a s s i fication , as shown
in Tab le 7 . 1 , and then to use a s imp l e formula invo lving the product o f mean
velocity and hydraulic mean depth to assess the Manning ' s n value . There
are dua l functions depending on VR : at very low values long grasses wi l l
remain erect and increas ing depth and velocity wi l l increase the n value due
to greater depth o f immersion . Above a l imiting value , they wil l deflect so
that Manning ' s n reduces with increasing depth and ve locity. For shorter
stands of grass , n diminishes progre s s ively with increas ing VR , though not
very s trong l y . The governing equations are given in Tab l e 7 . 2 .

13
TABLE 7 . 1 . GRASS COVER RETARDANCE CLASS I FICATION .

Stand : GOOD FAIR

Average length C lass Class

Longer than 0 . 76m A B

0 . 28 - 0 . 60m B c
0 . 1 5 - 0 . 28m c D
0 . 05 - 0 . 1 5m D D
Less than 0 . 05m E E

TABLE 7 . 2 . MANNING ' S N VALUES FOR GRASS SURFACES .

The coef ficients p and q apply to the equation

n = p + q/ (VR) . . . 7 . 12

Retardance Coe fficients in equation Limit s of VR

c las s : p q m2/s

A 0 . 44 0 -1. 617 < 0 . 154

0 . 046 +0 . 022 > 0 . 154

B 0 . 403 -3 . 33 6 < 0 . 05 3
0 . 04 6 +0 . 0 1 0 0 . 05 3 - 0 . 1 79
0 . 03 5 +0 . 0 1 2 > 0 . 1 79

c 0 . 03 4 +0 . 046 < 0 . 083

0 . 028 +0 . 005 > 0 . 083

D 0 . 038 +0 . 002 < 0 . 100

0 . 03 0 +0 . 003 > 0 . 1 00

E 0 . 029 +0 . 00 1 < 0 . 123

0 . 0225 +0 . 002 > 0 . 123

14
7 . 4 . 13 Regarding channe l s with aquatic weeds , research by HR Wal l ing ford
l ed to the fol lowing formula , depending on the extent of weed coverage :

n . + 0 . 02 �-!Fr . . . 7 . 13
e = �oasl.c --w

where �oasJ.c
. i s the Manning ' s n value for the channe l without weeds
l<w i s the fractional sur face area coverage o f weed growth
Fr i s the channe l Froude number , V/{ ( gA/W )
C

Larsen , Frier and Ves tergaard ( 19 9 0 ) des cr ibe both field work in a weed
affected reach of r iver and flume tests , and develop a s imi lar type o f
function a s those g iven for grasses i n Tab le 7 . 2 above . They re late
Manning ' s n to VR , with the dry weight of growth in g /ml forming a further
parameter . They suggest that there is a basic winter function for n in
terms of VR , and that the summer funct ion wi l l depart from this for VR < 0 4
.

m 2 / s , the presumption being that above this va lue the weeds wi l l lay flat or
be s coured away . One field measurement may then characterise the trend of
the summer roughness function . The inf luence of weed growth on an East
Ang lian r iver was investigated by Powe l l ( 19 7 8 ) c lear ly demonstrating the
strong seasona l i ty of the roughness coefficient , a l s o indicating large
tolerances on its assessment .

7 . 4 . 14 Regarding flood p la in roughness , Klaasen and Van der Zwaard ( 19 7 3 )

carried out laboratory research on mode l led vegetation , inc luding such
features as orchards and hedges , whi ch may be a he lp ful source of
information . So far as orchards and forests and forests are concerned ,
provided there is no undergrowth and the water sur face is be low the top
growth , the method of ana lys is used for the rod roughness in the FCF can
also be app l ied , uti l i s ing a knowledge of the typical diameter and spacing
of tree trunks . There is a l so the information in Appendix 5 . However ,
in the absence of actual measurements under above bank conditions , there
wi l l probab ly be greater to l erances on e s t imating the conveyance due to
uncertainty 1n the roughnes s coefficient than wi l l arise from the
computation o f the recommended adj ustment to these bas i c va lues to a l low for
the f lood p lain/channe l interaction .

15
7.5 Need for and utilisation o f field data

7.5.1 I f river engineers are to make best use of existing knowledge o f the
behaviour of compound channe ls when as sessing the flood conveyance o f thei r
river system , i t i s important that they not only acquire the best quali ty
field data over as wide a range o f depths as feas ible , but also that they
interpret them correctly in the framework o f what is now known about the
comp lexity of two-stage channel s . Understanding o f the proces ses at work
has been defic ient in the past , and so conventional methods o f treating
field data under over-bank condi tions has probably l ed to serious error s .
I t has been demonstrated quite positively that the main result o f
interact ion between main channel and f lood p lain flows i s the reduction o f
the main channel flow , yet the convent ional treatment o f above-bank stage
di scharge data has , in effect , been to a l l ow for any interact ion by
adj usting the flood plain roughness coef fic ient , the basic resistance
funct ion for the main channe l being assumed to correctly represent its
component discharge at above-bank flow. This has perhaps been inevitab le ,
given the previous s tate o f knowledge , but the net result may have been the
use of inf lated values of flood plain roughness . The combination o f this
with inaccurate methods of treating the compound sect ion must have led to
many errors - and in some cases large errors - in assessing the flood
conveyance o f rivers . Thus it i s firmly recommended that all future
analyses of stage- discharge data under above-bank conditions should make
fu ll use o f the new methodology .

7.5.2 The problem is i llustrated by s ome samp l e calculations for a small

river , bed width 15m , channel depth 1 . 5m , s ide slopes 1 / 1 , Manning ' s n =

0 . 03 , two flood plains o f width 20m , Manning ' s n = 0 . 06 , channel gradient

0 . 3 / 1000 , ( this i s the same cross-section as later used to typify a small
sand bed river in Chapter 9 ) . The stage discharge function for depths up to
3m is shown in Figure 7 . 8a on a log- log basis for a range of assmptions .
The basic calculation before making a l lowance for interaction is shown in
the upper part o f Figure 7 . 8 as a broken line , and shows the ful l depth f low
as 80 . 3 4 m3 / s . With a l l owance for interact ion , this reduces to 6 1 . 8 1 m3 / s
a s a t the terminal point o f the ful l line . The chain dotted line
i l lustrates the assumption that would have to be made about the flood p lain
roughness in order to achieve close agreement at the highest stages if no
a l l owance is made in the analys is for interaction e ffects : this is with n
0 . 60 , TEN times the true

16
value assumed in this example . This unrealistic value comes about because
what is being attempted is to get the correct discharge by adjusting the
f lood plain roughness coeffi cient when in real ity the " lo s s " of conveyance
under above-bank conditions occurs in the main channel , and not on the flood
plains .

7.5.3 It may be noted that the predicted s tage discharge curve shown by the
ful l line in the upper Figure 7 . 8 has a change in gradient at bank ful l ,
with a humped character over the l ower range o f flood p lain depths . This is
very simi lar to many fie l d observations o f real r ivers ( see for examp l e
Chapter 5 , section 5 . 5 . ) . Even with a fal s e increase in flood plain
roughness there is no way that thi s characteristic hump can be produced
without taking acc ount of interaction . = 0 . 60
The l ine drawn for n
F
forecasts signi ficantly higher di scharges in this range , even though it can
give approximately correct discharges at high s tage s . Note also that the
s tage-discharge function with a l l owance for interaction does not g ive a
straight l ine on this log-log plot and so methods o f interpreting and
extrapo lating from observed stage/discharge data that presume the existence
o f a power law i . e . a straight l ine on a l og-log plot , for above-bank flows
are l ikely to be inaccurate and could be s omewhat mis leading .

7 .5.4 The fal se picture o f the division o f flow that emerges i f one tries
to compensate for interaction e ffects in this way i s shown in the l ower
Figure 7 . 8 . There are two sets of curves corresponding to the a ssumptions
explained above , for both Q /Q and Q / Q . The ful l predict ive method shows
C T F C
Q /Q reducing from unity at bank ful l to 0 . 624 at depth 3m ( H * = 0 . 5 ) . The
C T
figure with no a l l owance for inter ference would be 0 . 7 19 , but us ing the
increase of n to achieve the correct maximum flow suggests that the main
F
channe l component o f the total is 0 . 96 2 . Turning to Q /Q ' with Q being
F C F
for both flood p lains , the correct prediction at maximum depth is 0 . 60 2 , the
basic calculation with the true n values gives 0 . 390 , whi lst the falsely
assessed n value would yie ld 0 . 040 . This is a gross distortion o f the
F
real ity of the flow divis ion and the consequent potent ial for serious error
using the traditional methods o f analys is mus t cause considerab l e concern .

7.5.5 It is not the purpose here to exp lain in detail the field procedures
for the measurement o f s tage and discharge . There are British and

17
Internat ional Standards on the subj ect as we l l as codes o f practice . The
subj ect is we l l described in the book Hydrometry , edited by Herschy ( 19 7 8 ) .

7.6 Incorporation into numerical models .

7.6. 1 One o f the mathematical pro cedures used for assessing f lood wave
propagation down a river system is channel routing o f the Muskingum-Cunge
type , see for exampl e Cunge ( 1969) . This takes account o f the speed o f
movement o f the flood wave and also its diss ipation , ut i lizing the flow
parameters sect ion by s ection along the r iver val ley . Garbrecht and Brunner
( 199 1 ) have recent ly pub lished a development of the method which
speci fica l ly aims to take account o f two-stage channel e f fects . They do
this by separately computing for main channe l and flood plains in a given
reach , and then j o ining the out flows from these zones together before
progressing to the next reach . However , they neglect the interaction e f fect
between the zones so that the ve locities used are the basic values whi ch we
have seen may be 1 5% or so different from the true values under over-bank
conditions . Clearly the methods o f al lowing for interaction deve loped in
this Manual could be incorpo rated into such a rout ing model , thereby
improving its abil ity to s imulate real rivers . In their recent paper ,
Garbrech and Brunner compare their hydrologic rout ing method with the U S
National Weather S ervices ful ly dynamic DAMBRK model ( Fread , 1984) , using
the latter as a bench mark . However , the bench mark method itself also has
the shortcoming of not making al lowance for the interaction e ffects o f
compound cro s s-sect ions .

7.6.2 One dimensional dynamic computational model s typical ly solve the St

Venant e quations o f energy ( or more strictly momentum) and continuity in a
time and space framework , uti l i s ing geometric information at many
cross-sections de fining the fluvial system. Some models may use the
cross-section data to de fine a unitary channel : this is no longer to be
recommended because by so doing the roughne s s coe f ficient i s also required
to take account of spurious changes due to the geometric anomalies
introduced by flow over the flood p lains , as we l l as real changes in
roughness with stage as the flood plains are inundated , and the extra
resistance due to interference e ffects . However , if the model requires the
sect ions to be treated as units , not divided into main channel and flood
plain zones , the predictive methods given above could be used as a

18
roughnes s/cross-section pre-processor , to deduce overa l l equival ent
conveyance functions in terms of f l ow depth , which could be incorporated
into the model data s tore as " look up" tab l e s .

7 .6.3 Other model s wil l use cross-section information in its more rational
form , with s eparate data for f lood p lain and main channe l . In thi s case
also it woul d seem appropriate to use the predictive methods g iven here in
the form of a pre-proces sor to provide the c onveyance/depth funct ion at each
sect ion in the model . Conveyance , K , is usual ly defined by :

K Q/{S • • • 7 . 14

where S i s the hydrau l i c gradient , s o can readily be assessed from the

predictive equations over the required range of depths . From a knowledge o f
the distribution o f flow between the main channel and flood p lain zones , i t
i s a l so poss ible to a s s e s s the momentum coefficient , to b e associated with
the conveyance as a funct ion of stage . Both are required for use in
one-dimensional mode l s .

7.6.4 The ques tion o f Froude Number , Fr , was dealt with in Section 7 . 3 .
and it was explained that in a two-stage channel the Froude Numbers in the
main channel and on the berms wi l l be different , and a l s o d i f ferent from a
who le cross- section va lue . These differences are real , o f cour s e , and as
the Froude Number i s a measure o f the speed o f propagation o f a sma l l
sur face disturbance , i t i s s ignificant in assess ing the stab i l ity o f
numerica l s chemes and their associated time s teps . It i s there fore
conceivabl e that the flood p l ain component could be computational ly stab l e
whi l s t the main channel component would be unstabl e - and that the s tabi l ity
status could not be obtainea from the whole-channel parameter s . C l ea r ly
care has to be exerc ised , with recent improvements in the understanding o f
compound channel f l ow providing s cope for a s igni f icant step forward , both
in the r e liab i l ity of s imulating real r ivers in 1-D model s and in a s s e s sing
the stab i l ity of computat ional s chemes .

7.6.5 This report deal s essential ly with straight r ivers in their f lood
p la ins , and so , in a model l ing context , it provides a one-dimens ional
treatment o f a one-dimens ional sys tem . It wi l l be c lear from Chapter 8 that
the methods deve loped her e are not app licable to systems with irregular plan

19
form , where the proce sses at work are significant ly di ffe rent . The methods
of a l l owing for int eract ion e f fe cts with straight aligned systems are not
adequate , therefore , for 1-D mode ls of highly irregular rivers , nor are they
appropr iate for mode ls incorporat ing two dimensions on plan . It wi l l be
apparent from Chapter 8 that there is much to be learned about how to
incorporate the exchanges o f flow and momentum into numerical mode ls of
meandering or very irregular r ivers . Such mode l s , even i f two-dimens ional ,
are current ly over-simp l i fied. Improvements corresponding to those that are
now possible in deal ing with 1-D systems wil l have to await the outcome of
detai led analys is and review of the findings from later phas es of research
in the FCF at Wa l l ingfo rd .

20
 1\
28 at full inundation
28 at artial inundation

W.L.

2b

28

I
2Wc

I
I

I
I

h
I
se = 112(8cL + scR)

I. q_
2b
.I
Fig 7.1 Il lustration of method of assessing the sectional goemetry
parameters for a natu ral river cross-section

Shear layer
Local velocities

Momentum transfer Depth-mean

velocities

Rood plain

Direction of flow
Secondary
flows

Boundary shear I

stress I
f JBW/45/10-91130

Fig 7.2 So me of the flow processes in two-stage chan nels, with their
influe nce on bou ndary shear stress (Sh iono and Kn ig ht, 1 99 1 )
 200

1 00

0
0.00 0.50 1 .00 1 .50 Rood plain

'
�� �-M.90m
'
'
,,
- ... 1ft.
.....
' '
'
'

0
0.00 0.50 1 .00 1 .50 Bank

200

1 00

0
(N!m2)
0.00 0.50 1 .00 1 .50
Shear stress Main channel

J8WI46/1 0-Il1130

Fig 7.3 Verti cal d istributi ons of shear stress, "'.tt : bed values based o n Preston tube
measurements, others using laser doppler anemometry. Y denote lateral
d i stan ce from centre-line. H* = 0.1 52. (Knight, Samuels and Shiono, 1 990)
 (a)
1

3 4 5
0

-1 Y(m)
H.
l::J.
0.100
-2
V 0. 1 47
+ 0.202
X 0.245
-3 D 0.304

0
0 0.395
(/) 0 0.500
:X:
0) -4
0.
-
-
-;!:'
-
"'
fll
..._

'"+
.....
-

� (b)
'o 1
(/)
:X:
0) 4 5
0.
-

I
0

�
!1i' -1
fll

i....
H •

as
(» -2 A 0.1 1 1
ii V 0.1 57
.E:
+ 0.1 97
(»
X 0.242
0) -3 D 0.298
c
Cd 0 0.396
-6 0 0.479
�
:;:I -4
Cd
(ii
a:

(c)

-1

-2 A 0.093
V 0.148
+ 0.196
X 0.245
-3 0 .299
8 0.400

-4

JFNi/4711 [Link]

Fig 7.4 Lateral variation of boundary s hear stress for d ifferent relative
depths, H•, for B/b = 2.2, 4.2 and 6.67, (series 01 , 02, 03)
(Shiono and Knight, 1 99 1 )
 0.4

0.3

0.2

0 Experimental data, 02

Broken lines - square law

Full lines - smooth law
( 1 .8 power)
0.1
.

\
.

\ .

0
0.5 1 .0 1 .5 2.0 2.5 3.0 3.5

0.4

0.3

0.2

Experimental data, 01

Experimental data, 03
0.1
+

0
0.5 1 .0 1 .5 2.0 2.5 3.0 3.5

JBW/48/10·91130

Fig 7.5 Comparison of observed average shear stress o n bed of main channel with
calculated val u es based on ·flow depth, H, and hydraulic mean depth, Re,
series 01 , 0 2 and 03 (see table 3.1 for details of geometry).
 2.00 ,--- ------,
sym. expt. H. Q(m3/s)
/),. 21 0.1 00 0.01 69
'0 V 23 0.153 0.01 83
;t 1 .50 + 27 0.198 0.0201
X 24 0.250 0.0220

0 sym. expt. H. Q(m3/s) Fr(av)

g
25 0.299 0.0274 0.977
26 0.353 0.031 6 0.951
0 27 0.400 0.0374 0.960
• 28 0.451 0.0446 0.948
t:J. 29 0.495 0.0548 0.974

0.50

0.00 L------��-
0.00 0.20 0.40 0.60
Lateral distance Y

Fig 7.6 lateral distribution of local Froude n u m ber when overall flow
condition is critical (Knight and Yuen, 1 990)

2.2
o River Tyne at East Linton
2.0 t:J. River Esk at Mussel burgh
x River Tyne at Spilm ersford
1 .8 • Water of Leith at Murrayfield
+ River Almond at Craug1ehall
1 .6
lii' 1 .4
�
1i)
e 1 .2
"0
....
1 .0
8
.D
Cl) 0.8
0>
::s
Ill
<!) 0.6

0.4

0.2

0
0.01 0.02 0.03 0.04 0.05 0.06
Manning's 'n'

JBW/49/1 0.11 1 130

Fig 7. 7 Manning's n val ues for with i n bank flows i n five Scottish rivers,
(Sarge nt, 1 979)
 3.0 r------:;;:r--.,...--....,

E 2.0
I

.c
15..
<l)
Cl --- Full prediction, nF = 0.06
Bas!c. nF - 0.60
Bas1c, nF - 0.06

1 .0 l..-----l--'---'---....l--..J__..I.---L-.1..._...1
10 3 1 00
Discharge -m /s

3.0
(b) I '
I '
I '
I \
I '
I \
I \
I \
I \
I \
2.5
I \
I \
I \
I \
E
. I \
.c
15.. I \
<l)
Cl

2.0

--- Full prediction

Bas!c, nF = 0.60
Bas1c, nF = 0.06

1 .5
0 1 .0
Discharge ratios JBW/5011(}.91130

Fig 7.8 (a) Stage d ischarge functio n fo r a hypothetical small river, comparing full
pred iction al lowing fo r interaction with traditional calculations.
(b) Ratios of d ischarges; channel/total , and flood plain/chan nel, with and
without allowance fo r interaction effects
8. IRREGULAR PLAN FORM

8. 1 Features o f meander ing flows in-bank

8. 1 . 1 Even when flowing below bank-full , a curved or meandered channe l

shows dist inc t ive flow features that make its hydrau l i c performance
s igni ficantly dif ferent from straight channel s . When a fluid flows round an
open channel bend , secondary currents are generated because the radial
pressure aris ing from the hor izontal curvature is not in balance at a l l
points i n the depth with the centripetal acceleration imposed by the mean
curvature in p lan . The faster moving upper l ayer s tend to move outwards ;
the slow near bed laye r s move towards the inner bank . This sets up a
secondary circulation which develops as flow proceeds round the bend . In a
meander ing system , the secondary current cel l set up in one bend decays as
flow passes through the cros s-over and is rep laced by one o f opposite hand
as flow passes through the subsequent bend .

8. 1.2 In a meander ing system , the length o f stream i s greater than the
straight l ine distance a l ong the val ley , o f course , and thus the availab l e
hydraulic gradient along the stream is l e s s than the val ley s l ope . The bank
ful l capacity is therefore reduced by two e f fects , this loss of ava il ab l e
gradient as a result o f channel sinuos ity and a l s o the additional head
losses arising from the success ion of bends . Thi s "bend loss" o c curs in the
secondary circulations , their deve lopment , decay and reversal in quick
succes sion , from redistributions of flow across the channel width and from
flow separation from the convex bank .

8. 1. 3 The system o f secondary currents , and the spec i a l form of those

currents in meandered channel s , also affects the morphology of those
channel s . There i s a familiar deepening o f mob i l e bed channel s on the
outsides o f bends , accompanied by shoal ing on the insides of bends . The
hydraulic engineer o ften makes good use o f these secondary currents in
siting intakes to avoid b lockage by bed mater ia l and to minimise the intake
of suspended sediments . In the context of hydrau l i c capacity , howeve r ,
these natural channel forms with almost triangular cross sections , switching
regular ly from deep on the l e ft to deep on the r ight and vice versa in a
meandered system , would be expected to per form somewhat diffe rent ly from an
artificial meandered channel with trapezoidal cross-section . The hydraulics

21
of channel s with irr egular plan form is c learly very complex , even without
the interactions with f lood plain flow when above bank .

8. 1.4 There have been many re search studies into the flow r ound open
channel bends , and several into meander ing channe l s , see for exampl e the
recent text by Chang ( 19 8 8 ) . I t is not the purpo se here to provide a
comprehensive review o f previous work but rather to point out some salient
features . Ervine and E l l i s ( 19 8 7 ) included in-bank meandering in their
review of flow in rivers with f lood plains , and they provided the following
summary o f some o f the typical geometric features , with the terminol ogy
de fined a l s o in figure 8 . 1 .

TABLE 8 . 1 Typical ge omet ri c details for meandered rivers.

S inuosity ( Sy) channel length along curved " thalweg"

straight line "va l ley" l ength
Description : Straight , Sy 1 . 00 to 1 . 05
Sinuous , Sy = 1 . 05 t o 1 . 5
Meandered , Sy > 1.5

Meander wavelength (between bends o f same hand ) , L = 10 tw approx .

M c
= 11
Average radius o f curvature in bends , 2 . 7 twc
�
= 11
Doub l e amp l itude o f meanders , for Sy 1.5, a 0.5 L
M M
for Sy 2, a = 0.8 L "
M M
for Sy 3. 1.4 L "
� M
for Sy 4, a 2.0 L "
M M
Meander belt width , W = a + tw c by definition
M M

The reader is re ferred to Leopo l d and Wo lman ( 19 5 7 ) and Jansen et a l ( 1 979 )

for a more comprehensive treatment o f fluvial morpho logy . However ,
according t o the above classi fi cation , many o f the research proj ects on
meandered channel s have actual ly concerned s inuous channel s , as the
sinuos ity was below 1 . 5 . A meandering channel with the cross-over sect ions
at 6 0 ° to the val ley axis wou ld typically have a sinuos ity of 1 . 4 or so .
Such a channe l is i l lustrated in plate 4 (Vo lume 1 , fol l owing summary
repor t ) .

8.1.5 The ba lance between the various components o f overall channel

resis tance when within bank can be broadly assessed from the large scale

22
research at the US WES S tation reported in 195 6 . The main series of tests
on meandering channels was with overbank f low , with a channe l o f bed width
0 . 6m . They were , however , p re ceded by within bank calibration runs , with a
straight channe l o f the same cross-sect ion and a l s o with bank full
conditions at each o f three sinuo s ities . Knowing the s inuos ity in each
case , the loss of bank full capacity because the gradient has been reduced
by the factor 1 / Sy is readi ly compute d . In fact , the observed bank full
flows are lower again , and this is because a proportion o f the gradient
along the thalweg is taken up by the form drag o f the succession o f bends on
p lan , with the balance overcoming the basic boundary frict ion . In the
s traight channel , o f course , the boundary friction accounts for the whol e o f
the energy dissipation . Th e fol l owing t ab le summa rises thos e result s :

TABLE 8 . 2 A l location of energy dissipation in US WES experiments.

S inuos ity , Bank ful l discharge Obs erved Reduction Proportion of

Sy a l lowing for the bankfu l l factor for thalwed s l ope
reducted gradient discharge planform used in planform
along thya lweg losses form losses

1/s 1/s

1 . 00 6 2 . 89 62 . 89 1 . 00 0
1 . 20 57 . 41 43 . 9 1 0 . 765 0 . 415
1 . 40 5 3 . 15 39 . 09 0 . 736 0 . 459
1 . 57 5 0 . 19 34. 56 0 . 6886 0. 526

8. 1.6 The above tests were made with a trapezoidal channel cross-section
rather than a natural ly shaped channe l , and al though artificial it does
provide a basic comparison . With the greatest s inuo sity of 1 . 5 7 tested , the
main channel bankful l capacity was reduced by an overa l l factor o f 0 . 5 5 .
0 . 80 o f this ( 1 / {Sy) comes from the greater path length o f the channe l , and
a further 0 . 69 ( { ( 1-0 . 5 2 6 ) ) from planform losses , g iving 0 . 80 x 0 . 69 = 0.55.
Thus depending on the sinuo s ity , up to hal f the total energy dis sipation can
be ascribed to planform losses in the se particular tests .

8. 1.7 S imi lar information is availab le from a prel iminary analys is o f

information from the early within bank tests i n the FCF at Wa lling ford with

23
a meandered channel of 60° cros s-over ang le and sinuosity 1 . 3 7 ( see
Plate 4 ) . The sect ion geometry differed from any of the straight channel s
tested (as described i n Chapter 3 ) but from the initial calibration of those
s traight channel s the basic resistance function for the cement mortar
construction was known ( see Appendix 2 ) . Hence the equivalent s traight
channel capac ity could be ca lculated with confidence . There was no attempt
to obtain a discharge measurement at precisely bankfull but a who le series
of s tage discharge tests were carried out over a range o f depths up to about
96% o f bank ful l . There were 18 such groups of data , and the running
averages of threes were taken , as explained in connection with the straight
channel tests , to minimise experimenta l s catter . It was then possib l e to
assess the reduction factor for p lanform losses for these within-bank flows ,
the loss of gradient due to s inuos ity being fixed by the plan geometry o f
course . The planform reduction factor varied from 0 . 95 a t shallow flows
(when the boundary drag would be more· s igni ficant ) to 0 . 86 at 88% of bank
ful l . Extrapolating to bankfu l l , the form drag reduction factor would
become about 0 . 82 , implying that about one-third o f the thalweg gradient was
used up in planform losses and two-thirds in boundary drag . The bankfull
capacity of the corresponding straight channel would have been 0 . 120 m3 / s ,
reduced to 0 . 1 0 1 m 3 / s by the greater channel l ength , and further to 0 . 082
m3/s by the planform losses . In thes e FCF tests , the p lanform losses were
rather less than in the nearest comparab l e US WES tests , though the rea son
for this is not yet estab lished . Perhaps the width to depth ratio o f the
channel has a signi ficant effect , as might the detai l s o f plan geometry .

8. 1.8 I f a resistance formula with an empirical coefficient e . g . the

Manning equat ion , is used to determine conditions in an irregular , sinuous
or meandering channe l , the use o f stage discharge observations to establish
the coef ficient value wil l automat ica l ly take account o f the form losses due
to plan irregularity as we l l as the boundary drag arising from the
composition of its bed and banks . It is to be anticipated that the
coeffic ient va lue s in such cases wi l l be much in excess of those for
straight channel s with otherwise s imilar boundary compositions and roughness
texture . Sources o f informat ion on channel roughness are mentioned in
Chapter 7 , section 7 . 4 , and some details are given in Appendix 5 . Cowan
( 19 5 6 ) proposed a system o f bui lding up Manning ' s n for a channel from
various elements of resistance and then applying a factor to al low for
meandering . For sinuosities below 1 . 2 he suggested no specific addition to

24
the n value ; for Sy � 1 . 2 to 1 . 5 , factor n by 1 . 1 5 ; for Sy above 1 . 5 factor
n by 1 . 3 0 . These last two n factors are equivalent to a l l owing 25% and 4 0%
o f the thalweg gradient to be used up by p l anform losses .

8. 1.9 The above r e fers to within bank flows , and demonstrates the
compexity o f the f low in channel s o f irregul ar or meander ing p lanform
compared with straight channel s . What e ffect the comb ination o f a meandered
main channel with a reasonably straight f lood p l a in wil l have on the channe l
processes is considered next .

8.2 Above-bank flows in meandering channel s

8.2. 1 Ervine and E l l i s ( 1987 ) r eviewed conditions where a deep channe l

meanders through a rel ative ly straight f lood p l a in , commenting that there
had been l ittle attention paid to the mechanics of overbank flow under such
conditions . It wi l l be obvious that with the main channe l flow no longer
conf ined within its banks , there wil l be exchange s of flow (with its
accompanying momentum) between the main channel and the f lood p la ins . With
fairly gentle meanders of modes t s inuos ity , the expectation is that flow
would leave the tapering f lood p la in to enter the main channe l , at the same
time s queezing f low out f rom the oppos ite bank of the channel on to the
oppos ite f lood p lain . In more tortuous systems , one might ant i c ipate that
maj or flood flows along the va l ley f l oor would a lmost ignore the main river
channel , except for its obstructing influence as the dominant f lood
flood p lain f l ows crossed and recrossed it as a transverse trough in the
val ley floor .

8.2.2 There has been much recent detailed work , both in the FCF and
e l s ewhere , in which the deta i l s o f this three-dimens ional flow structure
have been examined . Wi l letts ( 199 1 ) provides ear ly pointers to the results
o f that research , with figure 8 . 2 . Showing how the secondary current that
occurs with within-bank f l ows reverses with overbank f l ow . This a l s o
change s the direction o f b e d movement , and there have been many c a s e s o f
fie l d observations fol lowing maj or floods that confirm this p icture , bed
sediments having been l ifted out of the deep channel on to the flood p lain
beyond . S o far as the water f low is concerned , with a reasonab le depth of
f low over the f l ood p l a ins , the continuity of f low within the main channe l
is broken : no longer is i t basi c a lly the same body o f water proceeding down

25
the r iver channe l ; it is being exchanged continuous ly with the flood p l ain
waters , at l east within the meander belt width . This exchange involves
additional head loss , because of differences in the momentum vectors between
these continual ly mixing flows . Figure 8 . 3 ( E rvine and E l l i s , 1 9 8 7 )
i l lustrates these flow comp lexities .

8.2.3 There are very many geometric and roughnes s conditions involved in
any comprehensive study o f even the s imp lest aspect o f meander ing r iver
flood f l ows , i . e . their stage discharge function . The research in the FCF
at Wal lingford is not yet comp lete { Summer , 199 1 ) and it would be premature
to attempt an appraisal in sufficient deta i l to provide a ful l design method
for irregular channel s . However , some preliminary indications o f the order
of magnitude of the influence of channel irregularities on the hydraulics
compound channel s wil l not be ami s s .

8.2.4 The ear ly work at the US WES publ ished in 1 9 5 6 has been referred to
already in the context o f straight compound channel s . The main thrust of
that research was into meandered channel s , concentrating on the influence o f
a meandering main channel o n the f l ood capacity . The t e s t s wer e a t l arge
s ca l e and covered three s inuo sities , as wel l as three roughness conditions
on the f l ood p lains , created by laying down sheets of expanded metal . In
terms o f the detail in the publ ished results and the a c curacy with which the
bas i c roughnes ss e s were determined , the test s eries was not idea l . For
examp l e only three flow depths wer e tested the shal l owest of which was of
the same order of depth as the expanded mesh roughening , which photographs
in the original pub l ication show to have been somewhat i rregular ( expanded
metal is difficult to keep f l at and uncurled at the edges ) . However , the
scale of the tests and their s cope make them a use ful r eference source .

8.2.5 Perhaps the most useful o f the pres entations o f information in the
original pub l i cation is that r eproduced as figure 8 . 4 . It is in
non-dimensional form , showing the reduction in main channel discharge
compared with a straight a l igned channel system . (The main channel section
is defined by vertical divisions at the banks . ) This reduction is based on
the premise that the f lood p lain flows themse lves may be assumed unchanged ,
compared with their " straight channe l " value s , so that any deficit is
ascr ibed solely to the main channel component . When f low first submerges

26
the f lood p lain , there i s already a 3 0 - 45% reduction in main channe l

conveyance , depending on the s inuosity . This was exp l ained earl ier in this
Chapter as being due to the extra stream l ength du� to s inuos ity p lus the
component o f energy d i s s ipation arising from p l anform l o s s e s . When overbank
depth reaches 0 . 6 x channel depth ( H* = 0 . 3 7 5 ) this reduct ion in

apparent main channel conveyance has increased to between 45 and 7 7% when

the f lood p l ains have the same roughnes s as the main channe l , yet with very
rough f lood p l a ins ( n /n = 3 approx) the increase with depth is more
F C
mode s t .

8.2.6 The as sumption that the l o s s o f conveyance should a l l be a l l ocated

to the main channe l whi l s t the theoretical f l ood p lain f l ows are unchanged
is hardly a rea l istic mode l in terms of f l ow detail s , a lthough with straight
channe l s it was found in the FCF work that the main channel di s charge
de ficits were much greater than any compensating addition to f l ood p lain
f l ow . However , it is quite l ikely bearing in mind what is now known about
the detail o f the f l ow exchanges that the main channel c omponent o f
discharge must suffer considerab ly from the periodic inf luxes o f f lood p l ain
f l ows and compensating e f fluxes o f main channel f l ows , so the concept o f
l o s s o f main channel c onveyance was a far-sighted cont r ibution .

8.2.7 Tab le 8 . 3 shows the US WES results in two d i f ferent ways : F l is the
factor by whi ch the measured straight channe l dis charge at the g iven depth
woul d have to be mul tip l ied to yie ld the measured discharge under meandering
conditions ; F2 is the factor by which the discharge e stimated by adding
together the main channe l f l ow extrapo lated f rom the observed meander ing
bank ful l condition and the estimated f lood p l a in f l ow , neglecting
int e r fe rence e ffects , wou l d have to be mul t ip l ied . This provides a matrix ,
albeit sparse , o f resu l t s for a range o f s inuo s ities , range o f re lative
depths and range o f roughnes s ratios .

27
TABLE 8 . 3 : Discharge and conveyance reduction factors for meandering
channel s flowing above bank-full .

F 1 : Factor by which the exper imentally observed al igned channel discharge

would have to be modified to account for meandering o f main channel .
F2 : Ratio o f observed total dis charge to that obtained by summing the main
channe l discharge . as extrapolated from the observed bank- ful l meandered
flow . and the experiment al flood p l ain discharge . proportioned down to the
actual flood p lain width .

Ratio of Relative
Manning ' s n flow depth .
on F P to Sy 1 . 20 Sy = 1 . 40 Sy = 1 . 57
main channel
value H F1 F2 F1 F2 F1 F1
m

1.0 0 . 167 0 . 804 0 . 960 o . 729 0 . 937 0 . 617 0 . 854

0 . 286 0 . 8 19 0 . 974 0. 7 1 3 0 . 89 1 0 . 65 1 0 . 85 4
0 . 375 0 . 7 83 0 . 9 44 0 . 686 0 . 860 0 . 63 0 0 . 82 1
2.0 0 . 167 0 . 698 0 . 741 0 . 6 16 0 . 718 0 . 541 0 . 696
0 . 286 0 . 867 0. 816 0 . 78 4 0 . 790 0 . 696 0 . 75 1
0. 375 0 . 8 30 0 . 85 4 0 . 775 0 . 843 0. 714 0 . 820
3.0 0 . 167 0 . 62 4 0 . 69 1 0 . 5 76 0 . 69 3 0.511 0 . 695
0 . 286 0 . 7 13 0 . 735 0 . 687 o . 7 47 0 . 605 0 . 73 1
0 .375 0 . 80 2 0 . 802 0 . 767 0 . 819 0. 717 0 . 817

8.2.8 The above re fer to a particular flood plain width . corresponding to

B/b = 4 , which in these tests was a varying amount in excess o f the meander
belt width . It would be reasonable to as sume , perhaps , that if the flood
plains were wider those reduction factors might still apply to the zone
within 2B = 4 x 2b whilst the sect ions outs ide might be relative ly
unaffected . This quest ion also il lustrates how difficu lt it is to
genera l i s e from a limited range o f experiments when so many geometric
parameters can be involved - and why a sound co-ordinat ing theory is
required to generalise any design procedures for irregular channe ls . It can
not be as sumed that the above empirical adj ustment factors apply to
geometries dif fer ing from those tested by US WE S .

28
8.2.9 Turning next to recent ly acquired data , only the FCF work with a
6 0 ° cross-over ang l e and s inuo s i ty 1 . 3 7 has been analysed for the treatment
that fol lows . Other cases studied in the FCF and e l sewhere inc lude a
"naturalised" main channel cross-section with outer bank deep zones and
inner bank shoa l s , and a l s o var ious distributions o f f lood p lain roughening .
Work with a 1 10 ° cross-over has a l s o been carried out at Wal l ingford , and in
thi s geometry the direction o f the deep channel partial ly reverses to
s imulate high s inuosity . The s imp l e case of equal roughnesses on f lood
p lain and in main channel with a trapezoida l main channel cross-se ction
forms the first progr e s s ion from the a l i gned and skewed channel s considered
in earl ier Chapters , so provides a first tentative picture o f broad
e f fects .

8 . 2 . 10 The geometry s e l e cted for this meandering system was not a direct
deve lopment of one of the cross-sections used in the main series of straight
channel tests : the constraints of the flume width and typical meander
geometries dictated otherwis e . Whereas al l the straight channel s had a bed
width of 1 . 5m , the meander ing channel had a bed width of 0 . 9m , so that its
aspect ratio ( width to depth ratio) was 0 . 9/ 0 . 1 5 = 6 , rather than 1 0 . It
had 1 : 1 s i de s lopes and was insta l led in a total f lood p l ain width , 2B , of
1 0m , so that B/b = 5 . 55 6 . Thus there were no actual measured discharges
either within bank or over bank for an exact straight channel equivalent .
Instead , any comparisons with the corresponding a l i gned system had to rely
on computations for the latter , using the we l l estab l ished basic resistance
function for this method of construction . Thes e calcu lations coul d either
be the basic flows obtained as the sum of the calculated zonal flows , or
cou l d include the a l lowance for interference e f fects deduced from the
comprehensive data analysis o f a l i gned systems reported in Chapters 3 and
5.

8.2. 11 Figure 8 . 5 shows four p lots :

I. The predicted f l ows for an equival ent a l igned system , with the main
channel calculated basi c discharge a l l owing for both the s inuosity and the
a l location of 1 / 3rd of the ava i l ab l e gradient along the thalweg to p l anform
losses , and also for main channe l / f lood p lain interaction effects using the
straight channel procedures deve loped in Chapters 3 and 5 ; ARF = 0 . 6 . These
results are shown as the ratio of the predicted f l ows to the basic zonal
calculation , (i.e. D ISADF ) .

29
II. The predicted flows for an equivalent straight system , no a l l owances
being made for sinuosity , but inc luding the interaction effects worked out
us ing the methods of Chapter 3 and 5 ( ARF = 0 . 6 = width/depth ratio/ 1 0 ) ,
shown as DISADF .

I I I . The observed stage discharge data for the meandered channel , in

comparison with the basic zonal calculation , with the main channel component
a l l owing for both the s inuos ity and p l anform losses assessed on the basis of
in-bank performance , but not for interaction e f fects .

IV . The ratio of observed discharge to that predicted under I above .

8 . 2 . 12 Some features o f figure 8 . 5 . require exp l anation . In computing

graph I , the main channel is effective ly much rougher than the flood p l ain ,
because of sinuosity and p l anform losse s . In consequence the interference
e f fects calculated from the methods deduced from the analysis of straight
aligned channel s are somewhat reduced , and at the upper l imit the velocity
difference calculated from the basic resistance of main channel and flood
pl ain has been reversed : the main channel flow is theoretica l ly moving
s l ower than the fl ood p lain flow . This is not a condition ever covered with
aligned channe l s : there the presumpt ion is a lways that the flood p lain
o ffers greater resistance than the main channe l . This diminution o f
ve locity di fference as depth increases , unti l they become equal and even
reverse , causes the " standard" a ligned channel inter ference equations to
show quite sma l l effects , reaching zero ( D I SADF = 1) when H* = 0 . 43 approx .
Thus any attempt to a l low for meandering as an extension o f the computation
procedures for straight channels must fai l , as the extra resistance of the
main channel diminishes rather then enhances the calculated interference
effect . The exchanges of flow between flood p lain and main channel , with
radical ly dif ferent flow structures and discontuity of fluid fluxes in both
the deep and sha l l ow sect ions , rule out any extension of the methods for
straight aligned or mi l dly skewed systems to ful ly meandered or irregular
channe ls .

8 . 2 . 13 With the type of flow structure i l lustrated in figure 8 . 3 , it is to

be expected that the observations o f stage discharge p lotted as D ISADF
against H* on figure 8 . 5 ( points shown by circles , I I I ) wi l l show much more

30
interference than either of the above computations . The gros s exchanges o f
f low and momentum between the f lood p lain flow and main channel flow over
the meander belt width induce much more powerful mechanisms for energy
dissipation than the dispers ion acros s the shear zone at the bank l ine in
straight al igned systems : there are l arge s econdary circulation ce l l s with
secondary ve locities perhaps an order of magnitude greater than in straight
channe l s , and something akin to expansion and contracti on los ses as f l ows
move from flood pl ain to main channel and vice vers a . The crosses o f plot
I V on figure 8 . 4 show this excess in e ffect . The values o f D ISADF for plot
IV are the ratios of observed discharge to predicted discharge using scheme
I above .

8 . 2 . 14 At shal low overbank depths , the conveyance o f this syst em with

s inuo sity 1 . 3 7 is about 7 0% of the sum of the zonal flows , calculated as i f
there was no inter ference or added energy l o s s due to flow and momentum
exchange . This drop in conveyance is additional to that which comes from
the extra resistance o f the main channe l itsel f . We saw in paragraph 8 . 1 . 8
that the bank ful l capac ity reduces from 0 . 120 m 3 / s to 0 . 082m 3 / s due to
s inuosity ; a factor of 0 . 68 . As depth inc reases , the ratio to the basic sum
of meandered main channe l flow p lus f lood plain f low increases and steadies
at about 0 . 80 . Thus the exchange of momentum for this particular geometry
reduces the conveyance compared with a basic calculation , such as might be
extrapolat ed from a knowledge of the meandered channe l ' s resistance
coefficient plus a calculat ion for the flood p lain , by 2 0% , for H* from 0 . 2 5
to 0 . 5 0 , but can be as much as 3 0% at l ower depths . This is a somewhat
greater influence than shown by the comparable F2 factors in Tabl e 8 . 3 , for
Sy 1 . 4 and Manning ' s n ratio 1 , obtained from the US WES test s .

8 . 2 . 15 The se few examples o f the influence o f meandering on the stage

di scharge funct ion for overbank f l ow clearly demonstrate that the effect is
important in an eng ineering context and should be a l lowed for in hydraul i c
computations . The b e s t method for doing this must awai t the completion of
the current research programmes , especi a l ly the l arge scale work in the FCF
at Wal l ing ford , inc luding the comp lex analyses that wi l l no doubt be
required in order to quanti fy the results in a general form to provide a
rel iab l e des ign method . Although systems with a sma l l ang le of skew were
shown in Chapter 4 to be amenab le to a s imple extens ion of straight aligned
channe l methodol ogy , the evidence for meandered channe l s is that the

31
interference effects with overbank flow are o f a radical ly dif ferent
character , rendering any extension of s traight channe l procedures
inappropr iate : a quite dif ferent co-ordinating theory is required .

8 . 2 . 16 Tab l e 8 . 3 above provides a matrix o f p l aus ib l e adjustment factors

for discharge and conveyance in comparison with equiva lent straight aligned
systems : the F 1 value s . However , for engineering purposes the concept o f an
equival ent a ligned system is less useful than an extrapolation from observed
or estimated conditions at bank ful l flow in the actua l meandered river : F2
is then the appropriate adj ustment for interference e f fects . Thes e were
based on l imited data from 1956 . Figure 8 . 5 provides a first example using
more recent and more detai l ed research results o f the inf luence of flow and
momentum exchange under overbank conditions on the extrapolation o f stage
discharge funct ions beyond bank ful l . However , neither Tab le 8 . 3 nor figure
8 . 5 is put forward as an estab l ished design procedure for irregular
channel s . Some indications of the way forward to such a method are
suggested in the fo l lowing section .

8.3 Flow model s for s inuous , meandering and irregular channe l s

8.3. 1 Flow in meandering channe l s with over-bank flow is a comp lex

three-dimens ional sys tem , with reversal s of secondary currents and maj or
exchanges of discharge and momentum between the main channel and flood
p lain . These interactions between the different f low regions differ
radica l ly in their mechanisms o f energy loss from those found in s traight
channel s a ligned with their flood p lains . There have , however , been several
attempt s at developing theoretical mode l s of the comp lex flow structure .

8.3.2 Ervine and E l l is ( 1987 ) offered a hydromechanics approach to the

prob l em , considering the flow over the meander bel t width ( see figure 8 . 1
for i l lustrat ion) as i f it repeatedly expanded and contracted as i t passed
from the flood p lain to ang le across the main channel and then on to the
opposite flood p lain . They considered four main s ources of energy loss in
the main channe l and three in the f lood p lain .

32
Main channel

1. Frictional resistance of the wetted perimeter of the channe l itse l f ,

which could be assessed from a knowledge o f bed material size or other
information on its sur face condition , using the Colebrook- White
equation ( or Manning with n re lated to the surface texture ) .

2. Meander bends with their secondary currents akin to large scale

turbulent eddies occupying most o f the cross-section . Energy loss
would arise from interna l shear and also transverse shear at the
boundaries .

3. Turbulent shear stress on the horizontal sur face at bank top leve l due
to the overflowing flood plain flow . The ve locities o f the two streams
wi l l di ffer in magnitude and direction , with the e f fect that an
apparent shear stress wi l l be generated on the inter facial plane , with
an influence on the main channe l that might be positive or negative
depending on the direction of momentum exchange .

4. There may also be poo l-riffle sequences in the meandered channe l , and
indeed the characteristic deepening on the outs ide o f the bend and
shoa l ing on the ins ide with cross-overs between bends induces the flow
along the thalweg to fol low a sequence o f deeps and sha l lows as with a
poo l-riffle sequence . This source o f energy loss is l ike ly to be more
s igni ficant at shal low flows than with over-bank f l ows .

Fl ood plain

1. Friction losses over the wetted per imeter , as determined from a

conventiona l resis tance equation .

2. The expansion loss where the f lood plain flow encounters the deep main
channe l .

3. Contraction losses where the flow leaves the main channe l to re-enter
the flood plain from the oppos ite bank .

8.3.3. These assumptions s imp l ified the flow situation to permit

conventional hydraulic assessments to be made of the various sources o f

33
energy dis sipation . Ervine and Ellis pro ceeded on the se l ines , treating the
f lood plain flow wi thin the meander belt width separate ly from that outside
the belt width , as the latter would not suffer the expansion and contract ion
losses ment ioned above . The va lues for expansion and contraction losses
came from work by Yen and Yen ( 1983 ) : the losses due to secondary currents
were assessed us ing a method pub lished by Chang ( 1983 ) : the Colebrook-White
or Manning equation would provide the boundary friction loss : but losses due
to the interface shear were omitted . Assembl ing the various head loss terms
and taking ac count of continuity , an equation was devel oped that could be
used to obtain the s tage-discharge function for a given geometry .

8.3.4 This method was tested against the same US WES data used ear lier ( US
WES , 1 9 5 6 ) with promising resul t s . These are i l lustrated in figure 8 . 6
( taken from Ervine and E l lis , 1987 ) . This shows for two o f the sinuosities
and two o f the f lood p lain roughnesses tested at Vicksburg how the
theoretical prediction compared with observation . Ervine and E l l i s
commented as fol lows : "The most obvious conc lusion i s that the predicted
discharge is general ly underestimated at higher flood p lain depths and
overestimated at lower flood p lain depths . The reason for
overestimating discharge at low flood plain depth may be related to the
omission of the eo- flowing turbulent shear stress term . . . . In thi s region ,
the predicted di scharges are of the order o f 0 - 20% greater than the
experimental data . For larger flood p lain depths the predicted discharges
are of the order o f 10% too low . It should be noted that the assumption for
energy loss due to s econdary c e l l s in the main channe l was derived for
in-bank flow and may be great ly repre ssed at higher f lood p lain depths .
This is combined with the fact that the assumption o f three sub-sections of
cros s-sectional area acting independently o f each other , with no interaction
between each section , represents a crude attempt to rationalise a comp lex
three-dimensiona l s ituation . 11

8. 3 . 5 . Ervine and E l l i s a l so compared their theory with smaller scale

research by Toebes and Sooky ( 1 967) , which had the advantage of separate
measurements of discharge distribution across the f loodway . They conc luded

" . . • predicted discharge is low compared with the experimental data in the
area of the main channe l and inner f lood p lain , with the opposite occurring
in the outer flood p lain . This woul d imp ly an overestimation of head loss

34
in the inner regions either in the secondary cel ls or expansion and
contraction losses " . The graphical comparison be tween the Ervine and E l lis
theory and Toebes and Sooky data is shown in figure 8 . 7 .

8.3.6 Pre liminary analysis o f the very detai led data on flow pattern .
velocity vectors and secondary currents obtained from the FCF at Wal ling ford
indicates that the above hydromechanics model would require appreciab l e
modification t o conform c losely t o the real ity of flow i n meandered
channe l s , but the general approach via component head losses remains a val id
avenue of development .

8.3.7 A somewhat more fundamental hydromechanic s model might be deve loped

through the momentum equations . though their ful l so lution depends on a
knowledge of a l l boundary shears and pressures . I t woul d be easy to assume
hydrostatic pressures . but this begs the question : the water sur face is not
a simp l e s loping p lane . and the flow separations where the f lood p lain
discharges enter the deeper main channel wi l l create non-hydrostatic
conditions at the channe l margins . However . it i s pos s ib l e that a model
could be developed with the s imp l i fying a ssumptions that could then be
calibrated against the available data .

8.3.8 A stage beyond the above woul d be to use re fined grid numerical
mode l l ing in two dimensions on p lan . so lving the 2-D St Venant momentum and
cont inuity equations for the geometries for which data is ava i l able . No
doubt there woul d be some need for empirical adj ustments to obtain good
agreement . but once achieved the model could be used to generate stage
discharge functions . or conveyance functions . for any other geometry o f
sinuous . meandering or irregular compound channe l with over-bank flow .
However . i t may be necess ary to add a third dimension , effect ive ly us ing a
l ayered mode l , to obtain satis factory simulation . The se are questions for
future study but are essential if hydraulic design information on complex
channe l s i s to achieve the standard o f accuracy and rel iabil ity that is now
becoming ava i lable for stra i ght compound channels . In the meantime , the
methods availab l e are somewhat crude and inadequately confirmed by wide
ranging data at l arge scale .

8.3.9 Design methods for irregular compound channe l s are outs ide the scope
of this report . The treatment of the subj ect here is intended to i l lustrate

35
the need for and importance of further analys is before a comparable design
method for irregular channe l s can be prepared , whil st providing the reader
with some indication of the order of magnitude of the inter ference e ffects
in such systems .

36
 (a)

- v,
Total
width
28

Flood lain

1 Width of meander belt Wm 1

(b) ,. .,

�' � lH
�
Rood plain Area 1
\
"
�
Rood plain
Area 2
'
1 2b
--

f't---l'f
ll' "'" '''"''''" '" ''
Main channel area Rood plain
Area 2

Fig 8.1 Plan and cross-section of meandered channel, with definition of

main :;ymb�ls used. <? ross-section shows subdivisions of area

General direction of
sediment movement

Sense of secondary
circulation

JBW/51/10-91/30

Fig 8.2 Established secondary currents in bend and main direction of

bed movement in meandered channels: (a) within bank flow;
(b) over-bank flow
 .: net outflow
before bend
-< net inflow
before bend
-

Flood plain flow

crosses over
main channel

-< net inflow

before bend

JBW/52/10-91130

Fig 8.3 Illustration of flow exchanges and secondary cel l development

in s uccessive meander bends, Ervine and Jasem (1 991 ) .
 0.60 21 .6
!1:
ne

Ill \I!
tD tl)
0.40 /::;. 1 6.6 Cii (ij
..1<: -
Sinuosity 1 .20 � t: tD
II! C:
w .o t:
2if 0.58 "- Ill
tD ..c:

ao
=

0.20 9.8
:tl..c:

0
0 10 20 30 40 50 60 70 80 90 1 00
% reduction in channel discharge

0 .60 21 .6
!1: 2
nc - �

= :E Ill \I!
tl) a.. <l> t»
"0 � 0.40 Sinuosity 1 .57 1 6.6 111
....
11!
....

w
..1<: -
c: tD ..1<: -
II! C: 2if = 0.91 t:
\Il
<P
l::
.o t:
"- Ill .o t:
tD ..t: '- Ill

oo
Sinuosity 1 .20 <P ..t:

w Sinuosity 1 .40 oo
2if = 0.58
� 0.77
:tl ..c:
0.20 =
9.8

0
0 10 20 30 40 50 60 70 80 90 1 00
% reduction in channel discharge

0.60 21 .6
!1:
nc - 3�

0.40 Sinuosity 1 .57 1 6.6

w
2if 0.91
=

Sinuosity 1 .20 __­ -- Sinuosity 1 .40

W
28" �
0.20 = 0.58 = 0.77 9.8

0
0 10 20 30 40 50 60 70 80 90 1 00
% reduction in channel discharge
JBW/53110-91/30

Fig 8.4 Red uction i n channel discharge (assuming flood plain flow is u naffected)
for th ree channel sinuosities a nd th ree ratios of flood plain roug hness to
main channel roughness: US WES (1 956)
 0.5
t>
I
I. G I
- - - -
11. /
G /
0.4 0 Ill. /
01- '
+ IV. '
8� '
�
0 + /
0.3 0 + /
/
0 +
:I: /
.

0 + /
/
0 + /
0.2 /
/
0 + /
/
0 + /
/
0 + /
/
/
0 + -
0.1 -
0 + -
-
.......
.......
.......
.......

0
0.70 0.80 0.90 1 .00
DISADF

Key
I. Predicted function for e9 uivalent straight system , but with main channel flow computation based on
thalweg slope and allow1ng for planform losses.
11. Predicted function for straight system of same basic cross-section.
Ill. Observed data in comparison with basic zonal computations, with main channel calculation using thalweg
slope and allowing for planform losses.
IV. The ratio of observed discharges to the computations under I, i.e. the interference effect with meandering
additional to the allowance for thalweg slope and planform losses in main channel.

JBW/5411 0-91/"JD

Fig 8.5 Discharge adj ustment factors for straight and meandered channels, FCF
experimental results for meandered case fo r 8/b = 5.56, 2b/h = 6, 60°
cross-over, LM/W = 1 0, Sy = 1 .37, with prediction for equivalent straight
c
system.
 1· ·1
30.5m

b... � � � �j
Plan of model with sinuosity r "" 1 .2
-- 4.88m

�idth of meander belt 2 .82 m �

1 I
!Y

�'''' "'�'\;:;'""'
l?�l
� 0 . 1 52m
, ,', � '''�
,,
Cross-section through model at apex of a bend

0.4
- 0.1

&J
"'
.s �;>"-"' " o
>: <:)" I !;)· " "
.17
I "
-� 0.08 11
'
(\\ "
"'
..r:. c:::',
1S.. ,
"0 a 1 "'
0.3 � +1 '
0 , 0
,g 0.06
c:
� ,' , "'
0 ]! I /
I "
§
;;::
0.04 0.2 0:
CD
I '
+ I
I cl
' 0 Experimental data, n1 0.01 2
"8. +
=
I
I
0.02 Experimental data, n1 0.035
I I
=

Cl)
0.1
11 Authors' prediction
0
- - - ·

11
0
0 50 1 00 1 50 200 250 300
Total discharge Q(l/s)

Plan of model with sinuosity r "'" 1 .57

Width of meander belt 4.42m

�''''''�'''''
0.6m
0 . 1 52m

Cross-section through model at apex of a bend

,,,,, , , , �
0.4
0.1
e
>:
-o
.?::
$/
<:)" I 'f-
�<"�- •
!;)t;J "' , 0
* ,
c: � If I
<:'-'"' ,
·a 0.08 t:::':: l
1S.. ..r:. I "'
"0 a I ,
0 0.3 � ,
+. I '
,g 0.06 Cl)
I , 0
c: I ,
0 �� I ,
!U
:;: Qj I ,
0 I
;;:: 0.04 0.2 0: I I
+ l
0 I o
..r:. I 0 Experimental data, n, = 0.01 2
a 0.02 I I
Cl)
+ Experimental data, n1 0.035
0.1 I I
=

0 11 - - - · Authors' prediction
"

0
0 50 1 00 1 50 200 250 300
JBWISS/1 0-91130
Total discharge 0{1/s)

Fig 8.6 Compa rison by Ervine and Ell i s (1 967) of their theory with selected cases
from the US WES (1 956) report: rati os of fl ood plain rough ness to main
channel rough ness of 1 and approx. 3 (Manning coefficients) ; for two
sinuosities. Geometries shown above.
 7.32m
1· ·1
] 1 . 1 85m

Plan of model with sinuosity r = 1 .1 35 (Toebes and Sooky)

���1"n\1 �0.0762m
Cross-section through model at apex of a bend

0.05 - t- 0.4
:[ '() 0 Experimental data, (Toebes and Sooky)
>;
c: 0.04 -
� - - - · Authors' prediction
·a ..c:
a. a.
-c (I)
1- 0.3 -c
0
0 0.03 - ...
;;::: � 0
A]
c: "
0 1ii "
" 0
3: Gi
0 0.02 -1- 0.2 a:
"
.,[J
;;:::
0 .d
"
..c: "
a. 0.01 -I- 0.1 o
(I) "
,
c ,
�' I I I
0
0 5 10 15
Total discharge Q(Vs)
JBW/56/1 0-91130

Fig 8.7 Comparison by Ervine and Ellis (1 967) of thei r theory with
Toebes and Sooky (1 967) report: equal flood plain and mai n
channel roug hness; Sy = 1 .1 35. Geometry shown above.
9. SEDIMENT TRANS PORT

9.1 General aspects o f sediment transport

9. 1. 1 The deve lopment and uti lisation o f water resources for irrigation ,
hydro-power and pub l i c supp ly can be severely affected by sediment in many
parts of the wor ld . Where there is a mature and we l l vegetated l andscape ,
sediment prob l ems may be re lative ly minor ; but where s l opes are steep and
vegetation sparse , the yie l d of sediment from the catchment gives high
concentrations in the rivers . In uti l is ing these water resources , and also
in managing rivers in terms o f flood protection , an understanding o f the
hydraul ics of a l luvial channe ls is vita l . So far only the water conveyance
aspects have been considered , but it is a l s o important to review the impact
of the research on flow and resistance in channel s with flood berms in the
context of its imp l ications for sediment movement . Only through better
understanding of fluvial morphology can the r ivers be controlled and managed
sympathetica l ly in respect of the environmental requirements , and the long
term success of engineering projects in r ivers carrying sediment be
secured .

9. 1.2 Sediment may be transported either in suspension ( fine mater ia l in

turbulent flow) or as bed l oad ( by the creep and saltation of particles
c lose to the bed) . These modes o f transport are governed by somewhat
different laws . In practice a range o f sediment s izes may exist in a r iver
- or an intermediate s ize even if of narrow grading - so that des ign methods
for handl ing sediment probl ems have a l s o to deal with a l l conditions between
the extremes of ful ly suspended wash loads and coarse grave l and boulders
moving in contact with the bed . Over the years , many formulae have been
derived re lating the transport of bed material , whatever its size , to the
hydraul i c properties of flow , but in the last decade or so it has been shown
that few are of acceptab le accuracy and even the best are far from precise
as predictive equations . Neverthe l e s s , recent theor ies are sufficient ly
comprehensive to represent not only fine and coarse material but the
intermediate sand s izes which dominate many al luvial systems .

9. 1. 3 Because o f the range o f sediment s izes o f interest and differing

transport mechanisms invo lved , a few de finit ions are in order :

37
Bed loa d : The material that moves in close contact with the bed .

Suspended bed ma teri a l load : That part o f the suspended load

cons isting o f particle sizes found in samples taken from the bed.
To tal bed ma teri a l l oad : The sum o f the above , i . e . the total
transport of those particle sizes present in bed sample s .
Wash load : That part o f the total sediment discharge cons isting of
partic l e sizes smal ler than are present in the bed ; frequently taken to be
sizes below 0 . 06 mm .

9.2. Transport process and theory

9.2. 1 The transport of sediment by even a steady uni form fl ow is a comp lex
process as yet incompletely understood . Many theories have been put forward
to provide frameworks for the analysis of data on sediment t ransport , and
very many experiments have been carried out over a period of some 50 years
under controlled conditions in laboratory flumes . S ome theor ies begin from
the analys is of the mechanics o f motion of individual particles , others use
s imi larity princip les or dimensional analys is as the starting point . Al l ,
however , include a measure o f empiricism in providing coe fficient values
based on laboratory experiments or field measurement s . Dimens ional analysis
provide s a set of governing variables as fo l l ows :

9.2.2 The minimum set o f basic quant ities which influence the process o f
sediment t ransport in two-dimensional , free-sur face flow are the unit mas s
o f f luid , p , the unit mas s o f solids , p , the vis cosity o f the f luid , v ,
s
particle diameter , D , water depth , d , shear velocity at the bed f ( gdS ) ,
denoted v* ' and acce lerat ion due to gravity , g . Dimens ional analysis yields
four group s :

. . . 9. 1

y • • • 9.2

z d/D • • • 9. 3

s . . • 9.4

9.2.3 One of the most signi f icant contributions to the science of

sediment motion was made by Shiel ds ( 19 3 6 ) , who ana lysed exper imenta l data

38
on the initiation o f movement o f granular material us ing the first two o f
the above four non-dimensiona l groups . For establ ished motion , an
additiona l parameter is needed to represent the transport rate , for exampl e
Einstein ' s ( 19 5 0 ) non-dimensional expression :

qs/p [ ( s - l ) gD) 3 / 2 • • • 9.5

where qs is the sediment transport rate as submerged weight per unit time
per unit width . I t fo l l ows that :

function [ Re* ; Y ; Z ; s ] • • • 9.6

Most transport theories use the above parameters or their e quivalent . For
examp l e , Ackers and White ( 19 7 3 ) r ep laced the above par t i c l e Reynol ds
number , Re* , by :

D • . • 9.7
gr

9.2.4 One o f the more s ignif i cant studies o f the total l oad o f
non-cohesive sediments was b y Engelund and Hansen ( 1967 ) . They used a
sub-set o f the functions indicated by e quation 9 . 6 , to provide a s impl e
re lationship between transport and channel hydraulics :

• • • 9. 8

9.2.5 The Ackers and White ( 19 7 3 ) theory considered coarse sediment and
fine sediment s eparate l y , and then sought a trans it ional function between
them . These transitiona l sizes include the sands and s i lts that are o f
great practical interest in a l luvial systems . Their analysis i s typical o f
several in the last fi fteen years which have used the power o f modern
computation to make the ful lest use o f the mass o f data availab l e from
laboratory and fie l d . Their results are a l s o typical in that the
optimisation procedures used to " ca l ibrate" the theory provided a set o f
equations from which the total transport o f bed mater i a l could b e
c a l culated within a factor o f two on about two occasions out o f three . The
transport rate was based on the stream power concept introduced by Bagno ld
( 19 6 6 ) , and the di fferent mechanisms app li cable to coarse and fine
sediments led to two sets of parameters derived from Y , Z and D which were
gr

39
linked by a trans ition parameter n which was expected to be - and
gr
confirmed as a function of Dgr .

mgr
G - C [ (F - A )/ A ] • • • 9.9
gr gr gr gr gr

where :

Transport rate -

ngr
Xd
G . • • 9 . 10
gr SD

Sediment mobil ity -

( 1-ngr)
V
F { ( gD ( s - 1 ) ) {32 log ( lOd/D) • • • 9. 11
gr

and

A , C , m , n = funct ions ( D ) • • • 9 . 12
gr gr gr gr gr

X is the transport rate expressed as the ratio of sediment flux to fluid

flux , by mass or weight , akin to a concentration which wil l be re ferred to
as the " sediment charge" . Data corre lations provided s imp le a lgebraic
formulae for all the functions 9 . 1 2 , so forming a direct method of
prediction . The original data analys i s of Ackers and White ( 19 7 3 ) has
recently been updated , providing improved formulae for the funct i ons 9 . 12 ,
HR , Wal l ing ford ( 1990c) .

9.2.6 White , Mi l l i and Crabbe ( 19 7 5 ) reviewed the then availab l e methods

and found that few could approach the level of prediction of the Ackers and
White method , the nearest comparabl e formulation being by Enge lund and
Hansen ( 19 6 7 ) . S ince then , there have been other contenders , for example
van Rij n ( 1984) , ( rather complex to detai l here ) as we l l as the
contemporary mult i-dimens ional empirical corre lation of Yang ( 19 7 2 ) .

40
log X ' 5 . 4 35 - 0 . 286 log (wD/ u ) - 0 . 45 7 log (v* /w)
+ [ 1 . 799 - 0 . 409 log (wD/u ) - 0 . 3 14 log (v*/w) ]
X log [ (V - v ) ( S/w) ] . . • 9 . 13
cr

In the above , X ' is in parts per mil l ion , V is the mean channe l velocity
cr
at initial motion , w is the fal l velocity o f the particles . The above
formula requires evaluat ion o f V and Yang gave a group o f express ions
er
depending on the value v*D/v .

9.2.7 These formulae for the total load o f bed material are o f similar
reliability . but are not precise predictors . In fact it is by now clear
that sediment transport is so s ensitive to the hydraulics of the stream and
the grading and condition of the stream bed that it is unl ikely that it
wi l l ever be possible t o predict t ransport rates from the overa l l hydraulic
parameters to much greater accuracy than at present . I t is this
sensitivity to hydraulic conditions , especial ly the mean flow velocity and
the consequent stress on the bed , that makes a cons iderat ion of the
inter ference e ffects o f flood plain flows on main channel flows
particularly s ignificant . In what fol lows in this Chapter , the 1990
up-date of the Ackers-White transport functions is used to examine the
effect of compound flow on bed material transport , though very s imi lar
conclus ions would emerge whichever of the relatively reliab l e calculation
procedures mentioned above was used .

9.2.8 The suspension of finer material by the s tream turbulence i s

broadly described b y the theory deve l oped by Rouse ( 19 3 7 ) . In this the
gravitat iona l effect through the fal l ve locity of particles is countered by
the upward turbul ent movement that arises from the vertical dis tribution of
sediment . The concentration , C , at e l evation , z , above the bed i s related
to that at some re ference e l evation , C at z , through the equat ion :
0 0

c (y - z) z
=
0
9 . 14
c z (y - z )
• • •

0 0

where

y f low depth
0 a turbulence parameter given by w/v *K

41
v* = shear ve locity , { ( gRS )
K = van Karman turbulence constant

9 . 2.9 The period of deve lop ing understanding has brought a number of text
books on the subj ect of sediment transport , for examp le Yal in ( 19 7 7 ) , Graf
( 19 7 1 ) , Garde and Ranga Raju ( 19 7 7 ) , A ll en ( 19 8 5 ) and Thorne , Bathurst and
Hey ( eds ) ( 19 8 7 ) . Papers by the originators o f the more reliable transport
functions should be read for ful l details of their methods .

9.3. The influence of compound flow on bed material transport

9.3. 1 It is clear that there is cons iderab le inter ference o f the f lood
p l ain f l ow with the main channe l f low through lateral shear e f fects and
exchange of momentum via secondary c ir cul ations , and that this interference
increases the apparent hydraulic resistance of the main channel , reducing
the mean ve locity therein s ignificant ly . It is this change in hydraul ic
conditions in the main river that wil l give rise to change s in the rate o f
transport of bed sediment . This sediment charge i s a function o f velocity
( and other factors to a l esser degree) , and ve loc ity dependence is to a
power above one : hence the sensitivity to the hydraulic conditions . Using
sediment transport theory , it is therefore possible to assess the e ffect o f
compound flow , taking account of the int er ference effect by using the
cal culation procedures devel oped ear l ier . This is the basis of what
fol l ows : in e f fect they are computed examples using the best of the
available knowledge , though it is hoped in due cours e that research on
transport in compound channe l s wil l be carried out to confirm , or up-date as
necessary , these forecasts .

9.3.2 The Ackers-White functions are straight forward to app ly : no

iterations are required and the equations for the various parameters are
readi ly programmable for computation . They have the advantage of covering
a very wide range of sediment s izes , and also by working in terms o f the
total transport of bed material they automat ically t ake account of the
balance between suspended load and bed l oad . The trans ition parameter
n a l l ows for the fact that the transport of fine material is a suspension
gr

42
proces s . depending on the overal l turbulence leve l in the stream , whi l s t the
transport o f coarse material is a bed process depending more on the bottom
shear stress deve l oped by the average ve locity . In app lying the e quations ,
some as sumptions were necessary :

the turbul ence leve l relevant to s ediment processes in the main channel
o f a compound section is determined by the stream gradient , through the
shear ve locity , v* = { ( gRS ) .

the bed stress relevant to sediment transport i s dependant on the mean

channel ve locity in the same way in a compound section as in a s imp l e
section .

the e ffect o f the f lood p lains on the main channel veloci ty is g iven by
the methods desc r ibed in Chapters 3 and 5 . This is direc t ly c a l culab l e
in f low Region 1 , but in the higher Regions o f f low , the assumption is
made that the main channel discharge adjustment factor remains the s ame
as at the l imit o f Region 1 . See Chapter 3 , para 3 . 5 . 10 .

ther e i s no transport o f channel bed sediment over the f lood p lain .

This is consistent with non-ava i l ab i lity o f such material on the f lood
plains , which would typica l ly be vegetated , and so no material
comparabl e to the main channe l bed sediment would be exposed .

9.3.3 For purposes o f i l lustration , a cross-section was chosen typical o f

sma l l rivers : bed width 1 5m , channel depth l . Sm , flood plain width , 2 x
20m , channel and flood p lain s ide s l opes 1 in 1 , thus 2b/h = 1 0 , and B/b =

3 . 87 . Two gradients were examined : 0 . 3 / 1 000 which might represent a r iver

with an active sand bed , and 3 / 1 000 to represent a grave l bed stream . The
Manning e quation was used for the bas ic hydrau l i c c a l culations , with n =
e
0 . 03 and two va lues of flood p lain coefficient , n = 0 . 0 3 and 0 . 06 . The
F
smoother of the flood p lain conditions gave equal roughnesses over the
perimeter , so was akin to most of the tests in the FCF , with f low
progressing with depth through Reg ions 1 to 4 . With the rougher f lood
plain condit ion , the flow stayed in region 1 , as expected from the work
with roughened f lood plains . The range of depths considered covered
within-bank f lows and r elative f low depths , H * , up to 0 . 5 , i . e . a depth o f

43
fl ow on the flood p l ain equal to the actual main channe l depth , l . Sm . The
channel of lower gradient was assumed to have a bed of 0 . 25mm sand , while
the s teeper channel bed was taken as gravel of 3 0mm dia . These were chosen
with some trial and error to provide interesting il lustrations : obviously
the sediment size chosen should provide transport at bank ful l f l ow if it
was to provide any simulation of a real a lluvial channel .

9.3.4 Figure 9 . 2 shows the calculated s tage-dis charge functions for all
four cases . These show the by now characteristic change o f s lope and
curvature when f low first goes over-bank . Figure 9 . 2 shows the calculated
sediment charge in the main channel for the four cases cons idered . Taking
the sand bed case first ( shown by ful l l ine s ) sand movement occurs from
quit e shal low depths in the main channel increasing to over lOO x l 0 - 6 ( 100
mgm/ 1 ) at bank ful l . With the rougher f lood p lain , this i s e f fectively the
maximum sediment charge at any dis charge . Above bank ful l the charge
diminishes because of interference e ffects from the flood plain , before
r i s ing again at depths above about 2m . With the smoother flood p lain ,
interference effects are less , and so after some hes itation in the rate o f
increase in charge with depth above bank ful l , a t higher depths the charge
rises above 200 x 1 0 - & .

9.3.5 The grave l bed case in figure 9 . 2 shows initial sediment movement in
the main channel at a flow depth o f l . Om , 2/ 3 rds channe l depth . The charge
increas e s to about 60 x l 0 - 6 at bank ful l , with a steep rate of increase up
to that depth . Beyond bank ful l , with the rougher flood p lain the
diminution o f transport above bank ful l due to interference e f fects is also
sharp , with the charge , X, dropping to perhaps a third o f its bank ful l
value a t depth 2 . 0m , i . e . O . Sm depth over the f l ood plain . There is a
sharp drop in X just above bank ful l f low with the smoother f lood p lains
too , though the drop is short l ived giving a return to rapid increase with
depth again at depths over 2m .

9.3.6 These estimates o f how compounding may a ffect sediment transport in

natural rivers are of considerab le interest , and a lthough dif ferent
examp les would show somewhat different results , the broad picture would be
expected to remain : a significant change in the s ediment transport function
when flow goes above bank , with the main channe l becoming less e f fective
than it woul d be in the absence of flood plains or channel berms .

44
9.3.7 Figure 9 . 3 shows the information in several di fferent ways . Here
sediment charge , X, is plotted against water discharg e , Q . The upper
diagram is for the sand bed river , with the rougher of the two flood plain
conditions . Graph I is the same data as shown in figure 9 . 2 , the charge
obtained by cal culating conditions in the main channel . Graph I I also
re fers to the main channel , but here the inter ference effect from the flood
plains has been ignored to demonstrate what the trend in the transport
function would have been l ike if there was no information on interference
and the sediment t ransport calculation had been based on the main channel in
isolation . The latter becomes s erious ly in error as depth increases , by a
factor exceeding 2 . Graph I I I converts the estimated charge shown in graph
I to the average over the who le stream , on the basis that there wi l l be no
additional transport of this material generated by the f low over the f l ood
plain , but the flow over the flood p lain is an additional di luent . Graph IV
wi l l be re ferred to later .

9.3.8 The lower diagram of figure 9 . 3 i s simi lar information for the
grave l bed river . Graph I and I I I show a very pronounced peak at bank
ful l , indicating that in terms o f transport capacity the system becomes
much less efficient above bank ful l , with the interference e ffect g enerated
by rough flood p l ains s o s evere in the examp le given that transport a lmo s t
stops again . In fact para l l e l computations were a l s o made for grave l s izes
o f 40rnrn and 5 0rnrn : 50mrn material is just mobile at bank ful l but not at
lesser or greater depth s , 40 mm material is mobi le over a range of depth
but virtua l ly ceases moving again at about O . Sm depth on the flood p l ain .

9.3.9 The results for f l ood plains with equal roughne ss to the main
channe l show s imi lar but rather less dramatic e ffect s . They form graphs V
( charge based on main channe l discharge ) and VI ( charge based on total
discharge ) .

9 . 3 . 10 There is often discussion as to whether a compound channe l is more

effic ient at t ransporting sediment than a s ing le section without berms . To
examine this question , some as sumption has to be made about the hydraulic
equivalence , and a s imp le trapezoidal section has been as sumed , with s ide
s lopes 1 / 1 , giving the same conveyance at depth o f 3m as the compound
section with flood p l ains having double the Manning ' s n value o f the main
channe l . The equiva lent s imple section has bed width 1 7 . 1 5m compared with

45
1 5m in the compound section : their stage discharge curves intersect at 3m.
The sediment cal culations for this equivalent section form graphs IV on
figure 9 . 3 . Below bank ful l this wider section is less effic ient in terms
of sediment transport that the compound one : graph IV l ies be low graph I
( graphs I , I I and I I I are ident ical below bank ful l , of course ) . The drop
in efficiency within bank is sma l l with sand , but rather more significant
in the gravel bed case where initial motion is del ayed to a higher
discharge . However , for above bank f l ow the inter ference effect diminishes
the compound channel transport e fficiency s o much that the singl e channel
has a much better performance : thi s i s shown by comparing graph IV with
graph I I I .

9.3. 11 This comparison o f graph IV with graph I I I may a l so be regarded as

representing the situation where a smal l river confined in a narrow val l ey
without flood p l ains disgorges into a wider va l l ey or on to an alluvial
plain . Under the confined va l l ey condit ion , maj or floods wil l carry
sediment according to graph IV , but where the val l ey opens out to provide
flood plains then graph I I I wil l app l y . The morphological inference is
that in high floods the reach with flood p lains can not carry forward an
excess o f s ediment del ivered from the confined val ley . The river channel
it s e l f wi l l accrete , partially b l ock , force more flow on to the flood berms
and in due course depo s it the load of sand or gravel on those berms . These
theoretical sediment forecasts for typical systems are cons istent with
geomorphological information on the devel opment of al luvia l plains and with
experience of river behaviour in maj or f loods .

9 .3 . 12 The purpose here is to draw attention to the influence of a

compound cro s s section with flood berms on the transport o f bed material .
The improved knowledge of the conveyance of such sections , a l l owing for the
interference between f lood p lain and main channel , can feed forward into
improved computations for river morpho logy. Even these pre liminary sample
calculations are of signi ficance in re spect o f fluvial morphol ogy , but they
are theoretical and involve assumptions , and so should not be regarded as
factual unt il there has been some experiment al confirmation . There is
ampl e s cope for research on transport in compound channe l systems , with
much of interest in straight systems as wel l as in meandered systems .
Until the results o f such res earch are avai lable , the method used here

46
provides evidence of the importance o f compounding on s ediment transport
capacity and a provisional calculat ion procedure .

9 . 4 . Suspended sol ids in compound channel s

9.4. 1 The turbulent s tructure in compound channel s i s undoubtedly

different from that in s imp l e channel s without flood berms , and so the
basi c turbulent suspension theory summa rised in equation 9 . 1 4 wil l not be
app l i c ab l e without some modi fication . In terms of broad e f fects , however ,
the mos t s ignificant feature that arises with c ompounding is the s trong
l ateral shear generated around the bank l ine which wil l diffuse the
s ediment in the upper l ayers of the main channe l across on to the flood
p l a ins , wher e the f low ' s capacity for keeping the sediment in suspension
wi l l be less . Hence there wi l l be depos i t ion on the f lood p lains o f
suspended load originating in the main channe l .

9.4.2 A l l en ( 1948 ) has described this process in some detai l , as wel l as

r eminding us of the amp le evidence from the field through levee buil ding
etc . that this is indeed a we l l authenticated process . Using the general
concept that the capacity to maintain mater ial o f a g iven s ize in
suspension is a function of the ve loc ity , and that the vel ocity wil l be
less on the berms than in the deep channel , A l l en shows that the
concentration sustainab le on the flood p l ains woul d be less than that in
the main channel by the factor (h/H) . The lateral dispersion from main
channel to flood p l ain thus increases the concentration above the
sustainabl e value , so a balance can only be achieved through depos it ion .

9.4.3 Much detai led information on turbul ence under the conditions o f
compound flow has been obtained from the programme o f research o n the FCF ,
inc luding dispers ion tests using dye s . This new information should provide
a basi s for better understanding of s ediment dispersion , but it is hoped
that a subsequent research programme wi l l examine this topic directly , by
us ing suspended s o l ids .

47
 3 . 0 .--------,--�---------:-r- ------------.,

�/·
""" . -
,.
_, . -·

nF = 0.03
/ - -.-·
--
·

.- -
·

. .
-
.

nF = 0.06 ';> ././ .- .-

- ·

.
S = 3 x 1 04 ./ ....- · ..... .
/ ./
•

2.0
/"">:·""
E
1
�:/ · · "' S = 3 X 1 0"3
i Bankfull
�

0
0 1 00 200 300
Discharge m3/s

Fig 9.1 Stage d ischarge curves for example rivers : main channel 1 5m
bed width, 1 .5m deep, 1 /1 side slopes; flood p lai ns 20m wide, 1 /1
. . . - '

3.0
I
I
:__._--- nF = 0.06
... ..
I ... ..
I

... ... ...

2.0
I
'
\ , , .. ..
I
'\
I
E '
. '
... \
..c
Ci..
..... ........ . Bankfull
a> ...
.. ... ..
c

...
.. ...
Sed Slope
... Dia mm /1 000
1 .0
0.25 0.3
----- 30 3.0

0
0 1 00 200 300
Sediment charge X 1 o..S JBW/571!0-91130

Fig 9.2 Sedi ment ch arge - X, versus stage, for sand river at S = 0.3/1 000
and for g ravel bed river at S = 3/1 000
 400 .------,
0.25mm sand

0 25 50 75
Discharge m3/s

400 .-------��--�
30mm gravel

300

,
0
X
G)
Cl
(ij
..c 200
0

c
G)
E
'6
G)
(f)
VI
1 00 IJ.

0
0 1 00 200 300
Discharge m 3/s

Key:

Graph I, X calcu lated for main channel allowing for interference from flood plain with n F double that in main channel
Graph 11, values of main channel X if interference is ignored
Graph Ill, sediment transport as calculated in I but expressed as ratio to total fluid flux
Graph IV, X calculated for an equivalent trapezoidal section (equal conveyance at depth 3m)
Graph V, as I but with flood plain roughness same as main channel
Graph VI, sediment transport as calculated in V but expressed as ratio to total fluid flux

JBW/S811 0·!11130

Fig 9.3 Sediment charge - X, versus total d ischarge: upper diag ra m for
0.25mm sand, lower diagram for 30mm gravel
10 . CONCLUD ING REMARKS

10 . 1 Summa ry of hydrau l i c design formulae for the conveyance o f straight

c ompound channe ls .

10 . 1 . 1 As a result o f the analys is of the new information from the l arge

s cale flood channel fac i lity at Wal l ingford , as described in Chapter 3 , a
set o f equations was derived for assessing the s tage/dis charge relationship
for compound channel s , or in other words the conveyance of their
cross-section . The bas i c method is a devel opment of the common approach
dividing the cross-section into zones by verti ca l inter faces at the channel
bank l ine . The basic discharges in the main channel and on the flood
p lains are first cal culated separate ly from an appropriate conventional
friction formula and roughnes s coefficients consistent with the character o f
the boundar ies , exc luding the vertical division p l anes from the wetted
perimeters . The sum of these basi c dis charges have then to be adj usted to
a l low for the e ffects of the interaction between the zones , which has a
s ignificant e f fect on the channel conveyance . Sever a l alternative methods
of adjustment were cons idered within the broad framework provided by
dimensional analys is , and they were progressively devel oped to be abl e to
cope with the ful l r ange of conditions teste d . Because of the comp l exity
o f the flow structure involving different regions o f behaviour , no s ingl e
formu l a cou l d cover a l l conditions . Moreover , the form o f equation and the
parameters i t depends on were found to differ from one region to the next ,
and so a logical method was estab lished for determining which flow region
applies at any particular f l ow depth . The e quations derived are bas i ca l ly
simple in form , with l inear variation with the governing parameters , but i t
is expected that app lication i n practice wil l uti l ise a computer program
that incorporates the logic for determining which region o f flow app l ie s .

10. 1 . 2 Being empirical functions based on data from channel s with a main
channel bed width/depth ratio of ten , it was necessary to confirm the
general appl ication o f these e quations by r eference to data from other
sources . Although the predic t ive e quations turned out to be robust in the
sense o f being trans ferabl e in the main to most cross-sections and types
and comb inations o f roughne s s , as exp lained in Chapter 5 , the equations for
Region 1 flow that covering the lower range o f depths required revision
for general app l ication . This possibi l i ty had been envisaged in the

48
dimens ional analys is o f Appendix 1 : the quest ion hinged on whether or not
the main channe l was wide in relation to the zone of interaction from the
fl ood p lains . The wide channe l assumpt ion proved not to be valid for
width/depth ratios below , say , twenty but a relative ly s imp l e modificat ion
was found adequate to cover the range of width/depth ratios for which
research data was available , and those o f practical interest . This
involved introducing an aspect ratio factor , ARF , proportional to the main
channel width/depth ratio . The resulting set o f design equations for
straight compound channel s is as fol lows :

10 . 1 . 3 REGION 1 : This is the region o f relative ly shallow depths where

inter ference e f fects increase progres s ively with depth . The formula for
this reg ion is based on Q* 2 , the discharge deficit normal ised by the product
(V -V ) Hh ( see Appendix 1 for nomenclature ) , with adjustment for aspect
C F
ratio . The discharge deficit is the deduction required from the bas i c
calculat ion , i . e . the sum of the basic f lows i n the flood plains and main
channe l , to obtain the ' true ' discharge . This is calculated as the sum of
the separate de ficits for f lood plains and main channe l . The flood p lain
deficit proved to be the minor part and is negat ive , i . e . an addition to
flood plain f low . I t depends l inear ly on the depth ratio , H * , but to cover
the case with roughened flood plains is progres s ive ly reduced by the factor
f / f as the f lood pl ain friction increased . The maj or part , the main
C F
channe l deficit , depends l inear ly on both the width ratio (width over f lood
plains divided by top width of main channe l ) and relat ive depth , the
relative depth multip l ier also depending ( though not very strongly) on
fri ction factor ratio and channel s ide s l ope . Thus for region 1 :

• . • 10 . 1

Q*
2C
= - 1 . 240 + 0 . 395 B/w + G H* . . . 10. 2
C

(Q * is never permitted to be negat ive , and perhaps should not be l e s s than

2C
0 . 5 , to provide some minimum interaction effect : if this l imit app lies then
a l s o Q* = 0 : see para 5 . 5 . 1 0 )
2F

For s e 2!: 1 .0:

G . . . 10 . 3

49
For s < 1.0:
e

G • • . 10 . 4

10 . 1 . 4 REGION 2. This is the zone of greater depth where the

interference e f fect reduces again . The most appropriate form o f design
function for this region relates the requis ite discharge adjustment factor
to the channel coherence , which is an express ion for the degree of
s imilar ity of hydraul ic conditions within the main channel and on the f lood
p lains . Channe l coherence itsel f is dependent on the section geometry and
roughnes s e s invo lved and is de fined and explained in paras 3 . 3 . 4 to 3 . 3 . 6 ,
and by equat ions 3 . 1 , 3 . 2 and 3 . 3 . It was found that the correction factor
to a llow for inter ference e f fects is rather more than the calculated value
of coherence at that depth : it is nearer to the coherence value at a
s omewhat greater depth , in other words requir ing a shi ft in relative depth .
The basic dis charge calculation has thus to be factored as foll ows in region
2:

D I SADF ( H * , channel geometry and roughnes s )

= COH { ( H* +shi ft ) , channel geometry and roughnes s ) • • • 10 . 5

where for s � 1.0,

e

shi ft = 0 . 05 + 0 . 05 N . . . 10 . 6
F

for s < 1.0,

e

shift = -0 . 0 1 + 0 . 05 N + 0 . 06 s . . . 10 . 7
F e

In the above N is the number of flood p lains . The test conditions in the
F
FCF did not provide any Region 2 results with dif ferent roughnesses on flood
plain and in main channel - the FCF rough flood p lain results remained in
Region 1 at a l l depths - but data from other sources has provided reasonable
confirmation of this approach for more modest di fferences o f roughnesses .

10. 1 . 5 REGION 3 . This is a relatively narrow transit ional region o f

flow , for which alternative approaches were cons idered , a s impl e constant
discharge adustment factor for the zone or an equation g iving DI SADF as a

so
function o f COH • Further data analysis showed that the latter was a
3
somewhat more accurate representataion o f the FCF data and so for Region 3
the fo l l owing function is recommended :

D I SADF = 1 . 5 6 7 - 0 . 66 7 COH • • • 10 . 8

However , the alternative form :

D ISADF 0 . 95 . . . 10 . 9

is a lmos t as accurate overa l l .

10. 1 . 6 REGION 4 . This is the region where the coherence o f the

cross-section is such that it may be treated as a s ingl e section when
c a l culating overa l l f l ow , with perimeter weighting of friction factor s .
This doe s not , howeve r , mean that the separate zonal flows c a l culated
provide accurate asses sments of the flows in those zone s : significant
interaction e f fects remain , and the method for adj usting the main channe l
f low separately is given later . For total f l ow computation however , in
region 4 :

D I SADF COH . . . 10 . 10

10 . 1 . 7 ASPECT RATIO FACTOR . The aspect ratio factor , ARF , i s

gene ra l ly given b y the main channel width/depth ratio / 10 , i . e . 2b/ 1 0 h .
Howeve r , i f the aspect ratio exceeds 20 , the channe l should b e assumed t o be
"wide " , when ARF = 2.

10 . 1 . 8 COMPUTATION PROCEDURE . The actual computation o f discharge

depends on the choice of basi c friction formu l a and assoc iated coefficient
for the conditions under review , as wel l as on the cross-section geometry
and hydraulic gradient . Nothing in the derivation o f the set o f predictive
equations l imits that choice o f friction formu l a : the engineer is free to
choose Manning , C o lebrook-White or whatever i s most appropriate for the
particular s ituation . The cross-section geometry provides the va lues o f
area , wetted perimeter ( exc luding the vertical division p l ane ) and hence
hydraulic mean depths for the main channe l and fl ood p l ain zone s . The
friction formula then provides "bas i c " values o f Q
FB '
Q
CB
and hence �B '

51
The cal culation of the various parameters to permit the use of equations
based on a c l ass ical compound channel form was explained in section 7 . 1 .
The best estimate for flow i f in region 1 is then obtained from:

• • . 10. 11

I f f l ow is in regions 2 , 3 or 4 , then the best estimate is obtained from:

Q = O I SADF �B • • • 10. 12
R2 , 3 or 4 R2 , 3 or 4

10 . 1 . 9 CHOICE OF REGION . The l ogic behind the sel ection o f the

appropr iate predictive e quation is dependent upon the calcu lation of
discharge for all reg ions in turn , re ferred to as Q • Q • Q and Q
R1 R2 R3 R4
respective ly . The choice o f the appropriate region and hence appropriate
total discharge proceeds as fol lows :

Region 1 or 2 ?

• • • 10 . 1 3

Region 2 or 3 ?

. • • 10 . 14

Region 3 or 4 ?

10 . 15

10. 1 . 10 The calculation o f Q etc uti lises the equations summa rised
R1
above , e qus 1 . 1 to 1 . 1 2 , together with the respective de finitions of the
dimensionless groups used , namely Q* . Q* and D ISADF . The l ogic route
2F 2C
given in the previous paragraph then s e l ects the appropr iate value . There
is no trans ition between them , in accordance with the individual test
results : there was l ittle if any evidence of a curved trans ition between
the regional equations .

52
10 . 1 . 1 1 TOLERANCES . The performance of this set o f predictive
equations was checked by reference back to the original Wal ling ford
experimental data , and the percentage discrepancies between the individual
results and the predicted discharge s for the observed depths , geometries etc
were assessed . These dis crepancies were subj ected to statistical analysis ,
to obtain mean errors and the standard error of es timate . The former
statistic indicates the overa l l goodnes s of fit , and the latter the
var iabil ity . This variab i l ity can have two components : any imperfection in
the trend of the predictive equations and a l s o the inevitabl e experimental
scatter due to random errors of measurement . With main channel and flood
plains of equal roughness , the mean errors for the various groups of tests
were all found to be under a third of a percent ; and variab i l ity under ha l f
a percent ( standard error o f estimate : some two-thirds wil l l i e within this
with normal distribution o f errors ) . The former shows the excel lence o f the
set of predictive equations in fitting the experimental trends ; the latter
could hardly be bettered in terms of consi stency of laboratory measurement .
The tests with roughened f lood plains were not represented quite so wel l :
a lthough the mean error o f 0 . 07% indicates good agreement on average , the
standard error o f 1 . 5% indicated greater variab i l i ty in the predictions .
The complete data set was fitted almos t exactly on average by these
predictive methods : mean error -0 . 00 1% . The variability o f 0 . 8% was very
satis factory , bearing in mind that perhaps 0 . 5% arose from the experimental
observations themse lves , and that the one set o f equations was app l ied to
smooth and rough conditions , to asymme tric as we l l as symme tric cases , to a
range of flood plain widths and channel bank s l ope s , over a range o f flow
depths covering four dif ferent regions of flow .

1 0 . 1 . 12 The purpose of the analys es o f data from other sources covered in

Chapter 5 was to validate and to adj ust and cal ibrate further as
necessary - the method based on the FCF results by comparing its
predict ions with a wider range of exper imental data , covering many more
geometries and roughne ss combinat ions . The only adj ustment found necessary
was the inc lusion of the parameter ARF in the formula for Region 1 :
otherwise the formulae trans ferred we l l and were able to explain severa l
unsuspected differences i n trends of behaviour . However , many o f these
other result s were obtained at sma l l s cale , when measurement problems ,
especial ly the setting of uni form flow and measurement o f gradient , give
higher tolerances than were obtainab le in the l arge FCF at Wa l l ingford .

53
Consequently the degree o f agreement between prediction and observation
was variable and in nearly a l l cases not as good as for the main data base .
S ome o f this increased discrepancy was undoubtedly due to wider
experimental tolerances , but some may have ar isen because of inaccuracy in
trans ferring the set o f empirical equations to geometries and roughnesses
wel l outs ide the r ange covered by the original derivation . I t i s therefore
di fficu l t to spe c i fy tolerances on the formulae themselves . The probable
e rror in predict ing discharge at 9 5 % confidence l evel due to deficiency in
the prediction method could be as l ow as 2% , but for mos t circumstances is
a lmos t certainly below 5% . To this must be added the tolerances in the
basic friction formula and the knowledge o f the roughness o f the channel
boundaries .

10 . 1 . 1 3 CALCULATION OF MAIN CHANNEL CONDITIONS . For some purposes it

i s not sufficient to c a l cu l ate the stage/discharge curve : separate
assessments of discharge in the main channel and f l ood p lain are required ,
duly corrected for interaction e f fects . One such exampl e i s in the
calculation o f bed material load in the r iver itse l f . The method o f
obtaining the adjusted value o f the main channel and f l ood p lain flows in
Region 1 wil l be evident from the def initions o f Q* and Q* ' Equations
C F
1 0 . 1 to 1 0 . 4 yie l d those value s , and then :

• • • 10 . 16

• • • 1 0 . 17

Other parameters such as the mean ve locities in thos e zones can then be
c a l culated .

10 . 1 . 1 4 Extending this separate zone adj ustment to the higher regions o f

flow has not been so wel l estab l i shed , because o f l ack o f data . Howeve r ,
this may be achieved to engineering accuracy by the method indicated in
paragraph 3 . 5 . 10 . As the c a l culations proceed from sha l l ow depths through
Region 1 , the value of D ISADF may be calculated from Q /Q depth by
C CR1 CB '
depth . The logic for choosing the regions ( based on total f l ow ) wil l in due
course indicate a change to Region 2 , but the value of the main channel
discharge adjustment factor for region 1 i s then taken to apply at al l
higher flows . This i s s imply achieved by retaining the value o f D ISADF
C

54
calculated at the l imit o f region 1 at all higher stages , so that in
Regions 2 , 3 and 4 :

* DISADFC at R1 l imit 1 0 . 18
QCR2 , 3 , 4 = QCB • • •

10 . 2 The advantages of compound channels .

1 0 . 2 . 1 The environmental and ecological advantages of two-stage channels

stem part ly from their more natural appearance , but also because berms or
f lood plains provide useful amenities . Their use has , of course , to be
compatable with inundation from time to time . The most general use of
flood p lains is for agriculture , especially where they are a natural
feature of the landscape , but they may also form parks or playing fields ,
and even relatively narrow berms alongside urban drainage channels can be
developed as linear parks . There are , however , precautions to be fol lowed ,
such as making good forecasts of frequency and duration of inundation , and
the elevation of the normal water table which wil l have an important
bearing on the vegetation growth , and hence the cost of maintenance and the
consequent hydraulic resistance .

10 . 2 . 2 A case is described by Sellin , Giles and van Beeston ( 1990) o f a

small river improvement project which was designed with the ecology very
much in mind . This proved less efficient than expected hydraulically and
also in terms of the need for and access arrangements for maintenance .
This is the River Roding in Essex , draining a catchment o f 250 km to the
project l ocation . The scheme in reality forms a three stage system. There
is a curvilinear main channel with some straighter but skewed reaches , with
berms out to a fairly regular retired bank l ine , all set below the general
level of an extensive flood p lain . Because the berms are not much higher
than normal water level , they provide a wet habitat , which may be the
delight of ecologists but because of the luxuriant growth of water-loving
reeds and other plants , offers a very high resistance except just after
cutting . The obj ect of the scheme had been to provide a 70 year standard
of protection to a neighbouring town through the provision of flood berms ,
but in practice this is not normal ly achievable , even with considerable
maintenance effort . Cutting the growth on the berm increases the flood
capacity at 1 . 3 5m depth on the berm by S O% . The actual capacity can be as
low as 15 m 3 / s with uncut vegetation , trees , tussock development and debris ,

55
but is typically 25 m 3 /s after a ful l season' s growth , rising to to 40
m3/s after cutting , compared with a design standard of 50 m3/s . This
example shows that the use o f two-stage channels is not without its
problems . What is ecologically highly bene ficial in a river corridor has
to be reconciled with the social requirement of l imited flood frequency.
This depends on hydraulic performance , which in turn depends on the normal
use of the berms or flood plains and on the vegetation thereon.

10. 2 . 3 The environmental imp l ications of river engineering are covered in

a report by Hey ( 1 990) . He also draws particular attention to the problems
of a high water table relative to the berm elevation and to the impact of
vegetation on flow capacity , whil s t pointing the way through river corridor
surveys and post-construction audit surveys to achieving the desired
balance between what is environmental ly desirable and what may be essential
in meeting hydraulic obj ectives . This involves not only vegetation , o f
course , but also within channel features such as shoals and embayments
which may attract fish and provide attractive habitats for much wi ld l ife
and plant species . B rookes ( 1 988) treats the environmental management of
channel ized rivers in detail , with many examples of good and bad practice .
This contains a wealth of experience and expertise on all aspects of
environmental assessment : habitat evaluation procedures , biotic indices ,
aesthetic evaluation , stream morphology , fish and fisheries , aquatic
plants . Figure 1 0 . 1 i llustrates improvements to river cross-sections to
increase their conveyance whilst retaining desirable ecological features
within a two-stage channel , though of course the river engineer has sti l l
to assess the l ike ly roughness coefficients and meet the hydraulic
obj ectives of the proj ect .

The hydraulic advantage of a compound channel when drainage

10 . 2 . 4
improvements are required is the increased flood capacity for a given
increase in stage , arising from the flow over the berms or flood plains .
This advantage may not always be as great as might have appeared from the
traditional methods of calculation , because the interaction between the
zones of different flow depth increases energy dissipation , as c learly
demonstrated by research and now calculable with the methods recommended
herein . Nevertheless , the advantage is very real in practice where the
available flow depth is l imited . Irrespective of any hydraulic advantage , a
knowledge of two-stage channel behaviour is necessary because they occur

56
natural ly : the c ross-section o f typical r iver channel s is determined by the
dis charge which o c curs for a combined period o f the order o f 2 days per
year , so it is obvious tha� maj or floods wi l l inundate their associated
flood plains , so �hat the ces ign condition i s when they are indeed two-stage
systems .

10 . 2 . 5 The prov i sion o f berms a longs ide artificial channel s has a l s o

advantages i n terms o f a c c e s s f o r maintenance . There i s n o need f o r such
access to be above water l eve l even during f loods , as maintenance work i s
a lmost invariably carried out when the system i s not near capacity . It
therefore makes good sense to provide two-stage channe l s , even in
c i rcumstances where they may have little or no amenity value : they combine
good access with increased hydraulic conveyance .

10. 3 State of knowledge acd need for further reseach

10. 3 . 1 The detai leci and extensive resear ch programme carried out on the
SERC-FCF at Wal lingfo�d has reaped the bene fit of being the first maj or
programme to combine large s c a l e , a high s tandard o f accuracy of
measurement , attentio� to cetail and c o l laboration between different groups
with comp lementary in�ere s � s . The large s c a l e has permitted the use of a
width/depth ratio more in : ine with. the practical range o f main channel
aspect ratios , anc in seve� a l respects has provided results quite different
from thos e reported f�om s=al l scale narrow fac il iies that typ i fied much o f
the ear l ie r resea�ch . Thi s research investment has been rewarded with a
detailed know ledge o f the : : ow in two-stage channe l s that was not previous ly
availab le from any so�rce , and this new data base confirms that a radical ly
different approach is requ�=ed to the hydraulic des ign and assessment o f
such systems . Previc�s me��ods were serious ly in error .

10 . 3 . 2 The ana l ys i s of t�e resear ch results and the appl ication o f those
findings to related topics such as the extrapol at ion of stage/discharge
functions and the tra�spor� of bed material shows that the conse quences o f
the new knowl edge a r e not �onfined t o improved estimates of channel
conveyance . They cut acro s s e stabl ished practice by providing new ins ight s
into channe l morphology , t�e c omputational model l ing o f river systems , the
hydrau l ic consequences of : o l lowing environmental ly des irab le river
management pract ices , etc . Thi s repor t dea l s only with the results

57
relevant to straight channels aligned with , or only mildly skew to , their
berms or flood plains , but subsequent aspects of the research programme
have concerned meandering channe ls and will undoubtedly give rise to
equa l ly significant changes in the approach to the hydraulics of irregular
channels when the analysis and interpretaion is completed .

10. 3 . 3Even with straight aligned channels , the research programme in the
FCF has left some quest ion marks ; some gaps in the coverage and
confirmation of ideas and concepts . The fol lowing items deserve further
study at large scale when opportunity and funding levels permit :

Stage discharge data is required of comparable detail and accuracy for

channels with differing aspoect ratios . The limitation of the FCF to a
main channel width/depth ratio of 1 0 has made confirmation of the
influence o f this feature somewhat elusive .

Stage discharge data is also required with boundary roughness of

various degrees on the f lood plains . Any artificial roughness would
require accurate and detailed cal ibration , of course , but without such
research there remains a possibility that the influence o f flood p lain
roughening in the form of sur face piercing rods may not be the same as
that of boundary-type roughening : the des ign equations using the ratio
of the flood plain to main channel friction factors may leave scope for
refinement for application to the more usual roughness condition .

The different regions of flow apparent ly inc lude Region 3 as a

transition between Regions 2 and 4 , and this may be associated with an
unstable re-organisation of secondary circulations . There is scope for
using the existing detai led information on flow structure to seek a
cause for this transition in the stage discharge function .

Turbulence methods wi ll undoubtedly oust the empirical procedure

recommended here in course of time , but their adoption requires better
understanding of the role of different interactions between the main
channe l and flood plain , and any dependence of their relative
importance on flow depth and cross-section geometry . A careful study
of the formulation of the turbulence coefficent in the lateral
distribution method is required , making ful l use of the data set now

58
 availab l e , with particular attention paid to accuracy o f s imulation o f
total flow and a l so the division o f flow , and any variation with flow
depth .

The imp l ications for s ediment transpor t o f the r eduction in main

channel discharge and velocity conse quent upon interaction from the
flood p lains requires experimenta l s tudy. Without experimenta l
ver i fication , the methods used t o indicate the order o f magnitude o f
the l ikely e f fect used in Chapter 9 are open to question , and i n any
event are unl ikely to have taken adequately into account a l l the
comp lexities of the f l ow structure .

10.3 . 4 Regarding the l ater phases o f r esearch on meandering channel s , i t

is important that the results a r e reviewed within the context o f s imulating
the fluid and momentum exchanges in computational model s with two-dimens ions
on p lan . Only thus can the work on meandering channe l s be extrapolated to
i rregular p lan geometries in general : they c learly can not be handled as a
s imp l e extension o f the methods developed here for straight channe ls . The
new insi ghts a l ready gained on the flow structure s in such systems with
overbank flow , as i l lustrated in Chapter 8 , provide a vital s tarting point
for incorporating appropriate mechanisms into any mode l .

59
 Overhanging
� tree trimmed

Retained submerged
vegetation
Sedge

n.w.l. - Normal water level

Tree planted
to shade berm
"'-...

Riffle

Before After

JBWI59/10-91fJO

Fig 1 0.1 Ecologically attractive two-stage chan nel design proposed for
River Ray in Oxfordshire (Brookes, 1 988, after Hinge and Hol lis)
11. ACKNOWLEDGEMENTS

This analys is of compound channe l s would not have been possible without the
support of funding from several Water Authoritie s , whose river duties have
now been taken over by the National Rivers Authority . I t has depended on
support from staff o f HR , Wal l ingford , as wel l as l iaison with the SERC
Proj ect Co-ordinator , Dr D W Knight , of the University of Birmingham . The
carefu l and detai led work of the UK research groups involved deserves
special mention . Without their research and the avai labil ity o f their high
c a l ibre results , the work described here could not have proceeded to a
successful conclus ion . The co-operation o f many other research groups and
individual s is a l s o acknowledged : they have readily supplied additional
details of their research , as wi l l be c l ear from Chapter 5 , and have a l s o
provided valued comments o n the Technical Reports which in essence provided
drafts of sections of the manual . Dr J ames Wark , now at HR , Wal l ingford ,
carr ied out the deta iled work on turbulence methods reported in Chapter 6 ,
as part o f the research programme for the Ministry o f Agricu luture ,
Fisheries and Food ; Mark Morris o f HR per formed the ana lys is that formed the
basis o f Chapter 3 ; their help was invaluab l e . I am a l s o indebted to the
many authors and pub lishers o f figure s and other information re-used here :
the references to them provide their s ources .

60
12. REFERENCES

Ackers P , ( 19 5 8 ) , Resistance o f fluids flowing in channels and pipes ,

Hydraulics Research Paper no 1 , HMSO , London , 19 5 8 .

Ackers P , ( 1980) Use o f sediment transport concepts in stable channe l

design , Int . Workshop on Al luvial River Problems , Roorkee , India , 1980 .

Ackers P and White W R , ( 19 73 ) , Sediment transport : new approach and

analysis , Proc American SCE , Nov 1 9 7 3 , vol 99 , HY1 1 , pp 2041 - 2060 .

A l l en J R L , ( 1985 ) , Princip les of physical sedimentatology , George Al l en

and Unwin , London , ( 1985 ) , I S B 0-04-5 5 1096-2 .

Asano , T , Hashimoto , H and Fuj ita , K , ( 1985 ) , Characteristics of variation

of Manning ' s roughness coefficient in a compound cross section .
International As sociation for Hydraulic Research , Proc 2 1st Congress ,
Melbourne , Vol 6 , August 1985 , pp 3 0-3 4 .

Baird J I and Ervine D A , ( 1984) , Resistance to flow in channel s with

overbank flood p l ain flow , Proc 1st Int Conf on Channe ls and Channel Control
Structures , Southampton , UK , p 4- 1 3 7- 1 5 0 .

Bagno ld R A , ( 19 6 6 ) , An approach t o the sediment transport problem from

general physics , US Geo logical Survey , Pro f . paper 422- 1 , US Gov . Printing
Office , Washington , 1966 .

Blasius H , ( 19 1 3 ) , Das ahnlichkeitsgesetz bei r eibung svorgangen in

flus s igkeiten , V D I Forschung shaft , no 1 3 1 .

Bray D I , ( 1982) , Flow resistance in grave l bed r ivers , I n : R D Hey et al .

( 19 8 2 )

Bridge J S and Jarvis J , ( 1982 ) , The dynamics o f a river bend : a study in

flow and sedimentary processes , Sedimento logy , vol 29 , 1982 , 499- 5 4 1 .

Brookes A , ( 1988) , Channe lized rivers : perspectives for environmenta l

management , John Wi ley and Sons , Chichester , UK .

61
Chadwick A and Mor fett J , ( 1986) , Hydraul ic s in c ivil engineering , Al len and
Unwin , L ondon .

Chang H H , ( 19 83 ) , Energy expenditure in curved open channel s , Proc ASCE ,

vol 109 , J Hydrau l i c Engineering , no 7 , July 1983 , 1 0 12- 1022 .

Chang H H , ( 19 8 8 ) , F luvial proces s e s in river engineering , John Wiley and

Sons , 1988 .

C l auser F H , ( 19 5 4 ) , Turbulent boundary layers in adverse pressure

gradients , J Aeronuatical Sc iences , Feb 195 4 .

Colebrook C F , ( 19 3 9 ) , Turbul ent flow in pipes , with particular re ference to

the transition region between the smooth and rough p ipe laws , Journal Instn .
Civil Engineers , London , vol 1 1 , p 1 3 3 .

Cowan W L , ( 19 5 6 ) , E stimating hydraul ic roughnes s coefficients , Agricultural

Engineering , vo l 3 7 , no 7 , July 19 5 6 , 473-47 5 .

Cunge J A , ( 1969 ) , On the subj ect of a flood propagation computation method

(Muskingum method) , J . Hydraulics Res . , vo l 7 , no 2 , 1 9 69 , 205-230

Einstein H A, ( 19 5 0 ) , The bed load for s ediment transportation in open

channel flow , Ser . Tech . Bul l . US Dept Agricul ture and Soi l Conservation ,
1026 , 1 9 5 0

E l sawy E M , McKee P M and McKeogh E J , ( 198 3 ) , Appl ication of LDA

techniques to vel o c ity and turbulence measurements in open channel s o f
compound section , Proc 20th Int Congress IAHR , vo l 3 , Moscow , p 2 5 5 - 263 .

E l l iott S C A and Sel lin R H J , ( 19 9 0 ) , SERC flood channel faci l ity : skewed
flow experiments , Journal o f Hydraulics Research , Vol 28 , 1990 , no 2 ,
p l97-2 1 4

Engelund F and Hansen E , ( 1967 ) , A monograph o n sediment transport in

a l luvial streams , Teknisk Vorlag , Copenhagen , Sweden , 1967 .

Ervine D A and E l l is J , ( 1987 ) , Experiment a l and computational aspects o f

overbank f lood plain f low , Trans . Royal Soc iety of Edinburgh : Earth
Sciences , vo l 7 8 , 1987 , 3 15 - 3 2 5 .
62
Ervine A and Jasem H K , ( 199 1 ) , unpubl i shed PhD the s i s and private
communication .

Fread D L ( 1984) , DAMBRK : the NWS dam-break flood forecasting mode l , Office
of Hydrology , Nat ional Wether Service , National Oceanic and Atmospheric
Admin . , U S Dept . Commerce , S i lver Spring , Md , USA .

Garbrecht J and Brunner G , ( 199 1 ) , Hydro logic channel-flow routing for

compound s ections , J Hydrau l i c Eng . , American Soc . Civi l Engrs , vol 1 1 7 , 5 ,
May 199 1 , 629-642 .

Garde R J and Ranga Raj u K G , ( 19 77 ) , Mechanics o f sediment transportation

and a l luvial stream prob l ems , Wi l ey Eastern Ltd , 19 7 7 ,

Graf W H , ( 19 7 1 ) , Hydraulics of sediment transport , McGraw-Hi l l , 1 97 1 .

Herschy R W , ed . ( 19 7 8 ) , Hydrometry , John Wi l ey and Sons , Chichester , UK ,

197 8 .

Henderson F M , ( 19 6 6 ) , Open channel f low , Macmil l an Company , New York ,

1966 .

Hey R D , ( 19 7 9 ) , F low resistance in gravel-bed rivers , ASCE , J Hyd . Div , vol

105 , HY4 , Apr i l 1979 , 3 65 - 3 7 9 .

Hey R D , ( 19 9 0 ) , Environmental r iver engineering , J . Inst . Water an

Environmenta l Management , vol 4 , no 4 , Aug 1990 , 3 3 5 - 3 40 .

Hey R D , Bathurst J C and Thorne C R , ( editor s ) ( 19 8 2 ) , Gravel bed river s ,

fluvial processes , engineering and management , Wi l ey , New York , 1982 .

Higginson N N J , Johnston H T and Myer s W R C , ( 19 9 0 ) , E f fect o f scale

di s t ortion in a compound r iver channel model study , Proc . Int . Con f . on
River F lood Hydraulics , Wal lingford , Sept 1990 , ed . W R White , John Wi ley
and Sons , 3 9 1-403 .

Hogg I G G , Guganesharaj ah K , Gunn P D S and Ackers p ( 1988) , The influence

of river regime on the flood management of the Sukkur Barrage , Pakistan , Int
Conf . on River Regime , ed White W R , Hydraulics Resear ch , Wal l ingford , UK ,
May 1988 .
63
Ho l l inrake P G , ( 1987 , 1988 , 1 9 89 ) , The structure o f flow in open channe ls ­
a l iterature search , vo l s 1 , 2 and 3 , Reports SR 9 6 , S R 1 5 3 and SR209 ,
Hydraulics Research , Wal l ingford , UK.

Hydraulics Research , ( 19 8 8 ) , Hydraul i c per formance o f environmenta l ly

acceptable r iver engineering s cheme s , Report EX 1 7 9 9 , Hydraulics Research ,
Wal l ingford , UK .

Hydraulics Research , ( 1 990a) , Charts for the hydraul ic design o f pipes and
channe l s , 6th edition , Hydraul ics Research , Wa l l ingford , UK , pub . Thomas
Te l ford , London .

Hydraulics Research , ( 1990b ) , Tab les for the hydraul i c des ign o f p ipes , 5 th
edition , Hydraulics Research , Wa l l ing ford , UK , pub . Thomas Te l ford ,
London .

Hydraulics Research , Wal l ingford ( 1990c) , S ediment transport : The Ackers and
White theory revised , Report SR 2 3 7 , HR Wal l ingford , Oxon , UK

Jaeger C ( 1 9 5 6 ) , Engineering fluid mechanics , Bl ackie and sons , London ,

19 5 6 .

Jansen P Ph , van Bendegom L , van den Berg J , de Vries M and Zanen A , ( 19 7 9 ) ,

Princip les o f river engineering , Pitman , London , 1 9 7 9 .

Kiely G , ( 1 99 1 ) , unpubl ished thesis p lus personal communication .

Ke l l er R J and Rodi W , ( 1984) , Prediction o f two-dimens ional flow

characteristics in comp lex channel cross-sect ion , Proc Hydrosoft 8 4 conf . ,
Portorez , Yugo s l avia , E l s evier , p 3 - 3 - 1 4 .

Ke l l er R J and Rodi W , ( 19 8 8 ) , Predic tion o f main channe l / f lood plain flows ,

J Hydraulic Research , vo l 2 6 , no 4 , p 425-44 1 .

Keulegan G H , ( 1 9 3 8 ) , Laws o f turbulent fl ow in open channe ls , Journal o f

Re search , Nat . Bureau o f Standards , Washington D C , Res . Paper 1 1 5 1 , vo l
2 1 , no 6 , pp 7 0 7 - 7 4 1 .

64
Klaas sen G J and Van der Zwaard J J , ( 19 7 4 ) , Roughness coefficients o f
vegetated f l ood p lains , J Hydraul ics Research , vol 1 2 , 1 97 4 .

Knight D W and Demetriou J D , ( 19 83 ) , Flood p lain and main channel

interaction , Proc ASCE , J Hydraul ic Eng , vol 109 , no 8 , Aug 1983 ,
1 0 7 3 - 1092 .

Knight D W , Demetriou J D and Hamed M E , ( 19 83 ) , Hydraulic analysi s o f

channels with f lood p lains , Proc Int Conf on Hydrol Aspects o f F loods , BHRA ,
Bedford , UK , p 1 29- 1 44 .

Knight D W , Demetriou J D and Hamed M E , ( 1984) , Stage discharge relations

for compound channe ls , Proc 1 s t Int Conf , Channel s and Channel Contro l
S tructures , Apr i l 1984 , e d Smith K V H , Springer Verlag , 1984 , 4 . 2 1 -4 . 36 .

Knight D W and Hamed M E ( 19 8 4 ) , Boundary shear in symme trical compound

channel s , Proc ASCE , J Hydreauli c Eng . , vol 1 10 , HYl O , Oct . 1984 ,
1 4 1 2- 1 43 0 .

Knight D W and Se l l in R H J , ( 1987 ) , The SERC F l ood Channel Faci lity , J Inst
Water and Env . Management , London , vol 1 , no 2 , 1 9 8 7 , 198-204 .

Knight D W and Shiono K , ( 19 9 0 ) , Turbul ence measurements in a shear l ayer

region o f a compound channel , J Hydraul i c Research , vol 28 , no 2 , 1990 ,
IAHR , Del ft , 1 7 5 - 196

Knight D W , Shiono K and Pirt J , ( 1989 ) , Prediction o f depth mean velocity

and discharge in natural r ivers with overbank flow , Conf . on Hydraulic and
Environmental Mode l l ing o f Coastal , Estuarine and River Waters , ed .
Falconer , Goodwin and Matthew , Gower Tecnical Press , 1989 , 4 19-428 .

Knight D W , Samuel s P G and Shiono K , ( 19 9 0 ) , River f l ow s imulation :

research and deve lopments , J . Inst . Water and Environmenta l Management , vol
4 , no 2 , Apr i l , 1 99 0 , 163- 1 7 5

Knight D W and Yuen K W H ( 19 9 0 ) , Critical flow in a two stage channel , Int .

Conf . on River Flood Hydraulics , Sept 1990 . Hydraulics Research ,
Wa l l ingford , Wi l ey and Sons , 1990 .

65
Kouwen N , Li R N and S imons D B , ( 19 8 1 ) , F l ow resitance in vegetated
waterways , Trans Amer . Soc Agric . Engrs , vol 24 , ( 3 ) , 684-69 0 , 698 .
Larsen T , Frier J -0 and Vestergaard K ( 19 9 0 ) , Discharge/ s tage re lations in
vegetated Danish streams , Int . Conf . on River Flood Hydraulics , Wal l ingford ,
Sept 1990 , John Wiley and Sons , 1 8 7 - 19 6 .

Leopold L B and Wo lman M G , ( 19 5 7 ) , River channel patterns : braided

meandering and straight , USGS Prof . P aper 282-B , 195 7 .

Limerinos J T , ( 19 7 0 ) , Determination o f the Manning coefficient from

measured bed roughness in natura l channe l s , U S Geological Survey Water
Supply Paper 1898-B , 1 9 7 0 .

Manning R , ( 189 1 ) , On the flow o f water in open channel s and pipes , Proc
Inst Civil Engrs , Ireland , 189 1 vol 20 , p 16 1 , 1895 vol 2 4 , p 1 7 9 .

Morris H M , ( 19 5 9 ) , Design methods for f low in rough conduits , Proc American

Soc . Civil Engineers , vol 85 , HY7 , July 1 9 5 9 , p 43 .

Myers W R C , ( 19 7 8 ) , Physical model l ing o f a compound r iver channel , Proc .

Int . Conf . on River F l ood Hydraulics , Wal l ingford , S ept 199 0 , e d . W R
White , John Wiley and Sons , 3 8 1 - 3 9 0 .

Myers W R C , ( 19 7 8 ) , Momentum trans fer in a c ompound channel , J Hydraulic

Research , vo l 16 , IAHR Delft

Myers W R C , ( 19 84 ) , Frictional Resistance in channel s with flood p lains ;

Channel s and channel control structures , 1 s t Int Conf , Southampton , Engl and ,
1984 , ed K V H Smith , pub Springer-Ve rlag , 198 4 , p 4 . 73 - 4 . 87 .

Myers W R C , ( 19 85 ) , Flow resistance in smooth compound channe l s ,

experimental data , Univers ity o f Ul ster , March 1 9 8 5 .

Myers W R C and Brennan E K , ( 19 9 0 ) , F l ow resistance in compound channel s , J

Hydraulic Research , vol 28 , no 2 , 199 0 , IAHR , D e l ft , 1 4 1- 1 5 6 .

Nal luri C and Judy N D , ( 19 85 ) , Interaction between main channe l and f l ood
plain f l ows , Proc 2 1 st Congress IAHR , Melbourne , Australia , p 3 7 7-382 .

66
Nikuradse J , ( 193 3 ) , Stronungsgesetze in rauhen rohren , VDI Fors chungsheft ,
1933 , p 361 .

Nikuradse J , ( 19 3 2 ) , Gesetzmabigkeit der turbulenten stromung in glatten

rohhren , VD! Fors chungsheft , 193 2 , p 3 5 6 .

Patel V C , ( 1965 ) , Calibration o f the Preston tube and l imitations on its

use in pressure gradient s , J Fluid Mechanics , vol 2 3 , Part 1 , 1965 , 185- 195

Petryk S and Grant E U , ( 19 7 8 ) , Critical flow in compound channe l s , ASCE , J

Hyd Div , vol 104 , HY5 , 1978 , 583-594

Powe l l K E C , ( 19 7 8 ) , Weed growth - a factor in channel roughness ,

Hydrometry, ed Hers chy R W , John Wi ley and Sons , 197 8 , 3 2 7 - 3 5 2 .

Prandt l 1 , ( 1 93 3 ) , Neuere Ergebni s s e der Turbulenz forschung , ZVDI vol 77 , p

105 .

Prinos P and Towns end R D , ( 19 83 ) , Estimating discharge in compound open

chane l s , Canadian Soc for Civi l Engineering , 6th Canadian Hydrotechnical
Conference , Ottawa , Ontario , June 1983 , p 1 29- 1 4 6 .

Prinos P and Townsend R D , ( 19 8 4 ) , Comparison o f methods for predicting

dis charge in compound open channels , Advances in Water Resources , 1984 , vol
7 , Dec , CML Pub lications , 180- 1 87 .

Radojkovic M , ( 19 7 6 ) , Mathematical mode l l ing o f rivers with flood pla ins ,

3 rd annual symp . Waterways , Harbours and Coastal div , ASCE , vol 1 , rivers .

Radojkovic M and Dj ordj evic S , ( 1985 ) , Proc 2 1 s t Congress IAHR , Melbourne ,

Austra lia , vo l 3 , p 3 68- 3 7 1 .

Raj aratnam N and Ahmadi R M , ( 1 9 8 1 ) , Hydraulics o f channels with flood

p lains , J Hyd Research , vol 19 , 1 , 19 8 1 .

Ramsbottom D M , ( 1989 ) , F lood discharge assessment - interim report ,

Hydraul ics Res earch , Wal l ingford , S R 195 , March 1 9 89 .

67
van Rijn L C , ( 19 8 4 ) Sediment transport : part I : bed load transport ; part
I I : suspended load transport ; part I I I : bed forms and a l luvial
roughness , Proc ASCE , vol 1 10 , HY 1 0 , Oct 8 4 , p 1 43 1 ; HY 1 1 , Nov 8 4 ,
1 6 1 3- 164 1 ; HY1 2 , Dec 8 4 , 1 7 3 3 - 1 7 5 4 .

Rouse H , ( 19 3 7 ) , Modern conceptions o f mechanics o f f luid turbulence , Trans

Amer . Soc . Civ . Engrs , vo l 102 , 1 9 3 7 .

Rouse H , ed ( 19 5 0 ) , Engineering Hydraulics , John Wiley and S ons , New York ,

1 9 5 0 , reprinted 19 6 1 etc .

Samuel s P , ( 19 89 ) , The hydraulics o f two-stage channel s - review o f current

knowledge , Inst Water and Environmental Management , River Eng ' s Conf . ,
Univers ity o f Loughborough , Jul y , 1989 .

Samuel s P , ( 19 8 8 ) , Lateral shear layers in compound channels , Fluvial

Hydraulics 88 conf , V ITUKI , Budapes t , Hungary .

Sargent R J , Variation o f Manning ' s n roughnes s coeffic ient with f low in

open river channel s , J . Inst Water Engineers , vol 3 3 , no 3 , 1 9 7 9 , 290-294

S e l l in R H J , Giles A and van Beeston D P , ( 19 9 0 ) , Post-implementation

appraisal of a two-stage channel in the River Roding , Es sex , J . Inst Water
and Environmental Mnagement , vol 4 , no 2 , Apr i l 1990 , 1 19 - 1 3 0

Shields A , ( 19 3 6 ) , Anwendung der Ahnlichkeitsmechanik und Turbulenz­

forschung auf die Geschiebebewegung , Mittie l . , PVWE S , Ber l in , no 26 , 193 6 .

Shiono K and Knight D W , ( 19 8 8 ) , Two dimens ional analytical solution for a

compound channe l , Proc 3 rd Int . Symp . on Refined Flow Mode l l ing and
Turbulence Measurement , Tokyo , Japan , July 1988 , 5 0 3 -5 1 0 .

Shiono K and Knight D W , ( 1990a) , Mathematical model s o f flow in two or

multi-stage straight channel s , Int . Conf . on River F l ood Hydraul ics , S ept .
1990 , Wal l ingford , ed White W R , John Wiley and Sons , Chichester , 1990 ,
229- 2 3 8 .

Shiono K and Knight D W , ( 199 1 ) , Turbulent open channe l f l ows with var iab l e
depth across the channel - a genera l mode l and new data , J Fluid Mechanics ,
vol 2 2 2 , September 1 99 1 , 6 17-646 .
68
Strickler A , ( 19 2 4 ) , Beitrage zur frage der ges chwindig-keits formal und der
rauhigkeitsahlen fur strome , kanale und geschlossene l eitungen , Eidg .
Wasserwirtshaftsamt Bern , Mil le i lung , 1923 ; Schweitz . Bauztg , 1924 , vol 83 ,
23 .

Thorne C R , Bathurst J C and Hey R D , ( eds ) , ( 19 87 ) , Sediment transport in

grave l bed rivers , John Wiley and Sons , 1987 .

Toebes G H and Sooky A A , ( 19 6 7 ) , Hydraulics o f meander ing rivers with flood

plains , Proc ASCE , vol 93 , J Waterways and Harbours , no 2 , May 1967 ,
2 1 3-236 .

US WES , ( 19 5 6 ) , Hydraul ic capacity o f meandering channe ls in straight

floodways , Tech . Memo . 2 . 249 , Waterways Experiment Station , Vicksburg ,
Mississippi , March 195 6 .

Ven te Chow , ( 19 5 9 ) , Open channel hydraulics , McGraw-Hi l l , New York .

Wark J B , Samuel s P G and Ervine D A , ( 19 9 0 ) , A practical method of

estimating veloc ity and discharge in compound channe ls , Int Conf on River
Flood Hydraulics , September 1990 , HR , Wal l ingford , Pub J Wiley and Sons ,
163- 1 7 2 .

Wi l letts B B , ( 199 1 ) , The hydraul ics of meandering channe ls , Seminar on

River and Flood Plain Management , 22 March , 199 1 , Scottish Hydraulics Study
Group , Glasgow .

White W R , Paris E and Bettess R ( 19 8 0 ) , The frictional characteristics o f

al luvia l streams : a new approach , Proc . Inst . C ivil Engrs , v o l 69 , ( 2 ) ,
73 7 - 7 5 0 .

White W R , Bettess R and Wang Shiqiang , ( 19 8 7 ) , Frictional characteristics

o f al luvial streams in the lower and upper regimes , Proc .
Inst . Civil Engrs , vo l 83 , ( 2 ) , 685-700 .

White W R , Mi l l i H and Crabbe A D , ( 19 75 ) , Sediment transport theories : a

review , Proc Inst Civ Engrs , Part 2 , June 1975 , vol 59 , 265 -29 2 .

69
Wormleaton P R , ( 19 8 8 ) , Determination o f dis charge in in compound channel s
us ing the dynamic equation for for l ateral veloci ty distribution , Proc IAHR
Conf . on Fluvial Hydrau l i c s , Vituki , Budapes t .

Wormleaton P R , Al l en J and Hadj ipanos P , ( 19 8 2 ) , Discharge assessment in

compound channe l flow , Proc ASCE , vol 108 , J Hy D iv , p 9 7 5 -9 9 4

Wormleaton P R and Merrett , ( 19 9 0 ) , An improved method of calculation for

steady uniform f l ow in prismat ic main channel / f l ood p la in sections , J
Hydrau l i c Research , vol 2 8 , no 2 , 1990 , IAHR , De l ft , 1 5 7- 17 4 .

Ya l in M S , ( 19 7 7 ) , Mechanics o f sediment transport , 2nd ed . , Pergamon Press ,

1977 .

Yang C T , ( 19 7 2 ) Unit stream power and sediment transport , Proc ASCE , vol
98 , HYl O , Oct 1 9 7 2 , 1805 - 1826 .

Yen B C and Yen C L , ( 1983 ) , Flood flow over meandering channe ls , Rivers
Meandering Conf , 1983 , 5 5 4-5 6 1 .

Zhe l eznyakov G V , ( 1985 ) , Probl em o f the interaction o f the main channel and
flood plain f l ows , Proc 2 1 s t Congress IAHR , Me lbourne , Australia , vol 3 ,
p373 - 376 .

The fo l l owing form a l imited circu lation unpubl ished series " SERC f l ood
channel research , Water Authority funded proj ect for the deve l opment of a
design manual , Technic a l Report Series , Consultant P Ackers " .

1, Resistance funct ions for the Wal l ingford faci l ity , August 1989 .

2, Resistance functions for the Wal l ingford faci lity , rod roughne s s , Nov
1989 .

3, The hydraul i c resistance of compound channe l s , review and proposed

method of approach , Nov 1989 .

70
4, Review and ana lysis of research at Wal l ing ford , in the context of
empirical design methods , October 1990 .

5, Review and analys is of other sources of data in the context o f

empirical des ign equations based on WFCF data , March 199 1 .

6, Skew channel s , February 199 1 .

7, Turbulence methods , Apr i l 199 1 .

8, Irregular p l an form , April 199 1 .

9, Sediment transport in compound channe l s , Apr i l 199 1 .

71
13 . NOMENCLATURE

NOTE : The nomenc l ature is not unique . To fol l ow establ i shed convention and
to avoid an excess o f subscripts and Greek symb o l s . some characters are
used in more than one sens e . The context in whi ch they appear wil l make
c l ear which is intended .

A Cross s ectional area

Parameters in the l ogarithmic smooth-turbul ent velocity
distribution
A. B Empirical coefficients
A , C , F , G , m , n Parameters in sediment transport function
gr gr gr gr gr gr
ARF An adjustment factor in the Region 1 functions to a l low for the
e f fect o f main channel width/deth ratio ( aspect ratio)
B Hal f total width o f channel p lus berm or berms ( flood p l ains ) , at
the e levation o f the berms ( flood p l ains ) . I f the berms s l ope and
are partial ly inudated , B i s taken as hal f the water surface
width
b' Mean width , defined as area/f l ow depth (normal ly with sub script
for main channel o r flood p l ain)
bw Bed width (norma l ly with subscript for channel , flood p lain)
b Hal f bed width o f main channel
e
b Bed width o f one o f a pair o f berms or f l ood p lains
F
B, b Parameters in a genera l i s ed form o f the exponential
smooth-turbulent Blasius e quation
C, D Parameters in a generalised form of the l ogarithmi c
smooth-turbul ent l aw

Concentration o f suspended s o l ids , reference value at prescribed

e l evation , z0
C Drag coefficient o f rods
D
COH Channel coherence ; subscript indicates method of calculation
D Pipe diameter ; s ediment diameter
D A dimens ionle s s indicator of grain size
gr
d Flow depth : diameter o f rods forming roughnes s
D I SADF Factor b y whi ch zonal calcu lation has t o b e multipied t o a l l ow for
interference
D ISDEF D i fference between zonal calculation of discharge and actual flow
D ISDEFBF Ratio of D I SDEF to bank ful l discharge
e Base o f Naperian logarithms , denoted by ln
F ,C,F Adj ustment factors
T
72
F form drag o f rods per unit channe l l ength
ROD
f Friction factor , 8gRS/V�
f Friction factor aris ing from smooth channel boundaries
s
f Total of rod roughness and smooth perimeter resistance , expressed
TOT
as friction factor
G A parameter in predictive equation for region 1
g Gravitational acceleration
H Total f l ow depth ; depth o f f low in main channel
H Flow depth on f lood p lain , H - h
F
h Depth o f main channel below berm l evel
H* Ratio o f f l ow depths o n f lood p lain and i n main channel
K Van Karman turbul ence constant

Ky Conveyance as c onventional l y defined , e . g . in Ven Te Chow

K Conveyance , Q/{ ( 8g S ) = A/{ (A/ fP)
D
k Linear measure o f roughne s s , after Nikuradse and Colebrook White
5
ln Logarithm to base e
log Logarithm to base 1 0
N Number o f roughening rods per unit channel l ength
n Manning ' s roughne s s coefficient , from 1 /n = V/ ( R 2 � 3 {S )
n Number o f rods in a transvers e row
N Number o f flood p lains
F
Q Discharge
Q* D i s charge deficit ( D ISDEF) norma l ised by (V -V ) (H-h) h
l C F
Q* D is charge def i c it ( D ISDEF) norma l ised by (V -V ) Hh
2 C F
q D ischarge intens ity , i . e . discharge per unit width
q Sediment transport rate as submerged weight per unit time per unit
s
width
R Hydraulic radius ( or h . m . d . ) , cross section area/wetted perimeter
Re Reynolds number , 4VR/u
Re* Grain Reyno lds number in s ediment transport , v*D/u
Re* Roughness Reyno lds number , u* kS/u
s Channe l / flood p lain side s l ope , horizontal/vertical
s Re l ative specific weight/mass o f sediment to f luid .
s S l ope of flood p lain towards main channel
BF
S Hydrau l i c gradient o f channel
SF Shear force at interface
1
tw Top width o f channel
c

73
u Mean ve locity over the flow depth
u Local mean stream ve locity
u• • v. Friction ve locity , { ( gRS )
V Average flow velocity through cross-section , or with suscript
through one zone o f cross-section
w Water surface width
w Fal l velocity of partic les of sediment
Hal f width o f main channel at elevation o f bank top
WC
y A non-dimens ional form for bed shear stress in sediment transport
y Local flow depth at point in cross-section
z Ratio of flow depth to sediment diameter
z* Ratio of local flow depth to rod diameter , z/d
z Distance from s o l id boundary ; local flow depth
a Velocity distribution coe fficient
� Blockage coe fficient arising from rod roughness
0 A correct ion term
� Radojkovic interaction index ; a function o f . . . .

p Dens ity o f fluid

T Shear stress
� Fluid viscosity
u Kinematic viscos ity of fluid , � /p
Q turbulence parameter for s o l ids suspension

74
Subscripts :

AV Ave rage
BF Bank fu l l
B , basic Basic values be fore a l l owing for interac tion
c Main channe l
CALC Calcul ated va lue
F Flood p lain
i Interval : one of a series of value s
I Inter face
MEAS Measured value
Rl , R2 , R3 , R4 Regions of flow behaviour
T Total i . e . main channel p lus flood p lains
*
Ratio between flood plain and main channel values ( except where
otherwise defined)

75
Appendices
APPENDIX 1

DIMENS IONAL ANALY S I S APPLIED TO COMPOUND CHANNELS

1. The independen t vari a b l es that determine steady uniform

friction-contro l led flow in a pri smati c compound trapezoidal channel are the
f luid properties , the roughnes s of the sur faces , gravity , channe l s lope and
cross-section geometry :

p - mas s density o f f luid

� - f luid viscos ity
k e - roughness of main channel perimeter
s
k - roughness of f lood p lain
SF
g - acce leration due to gravity
S - s lope of channe l
b e - hal f bed width of deep channel
b bed width o f each flood p l ain
F
s e - s ide s l opes of deep channel , 1 vertical to se horizontal
s F - s ide s lopes o f flood p l a in edges
h - depth of main channe l
H - tota l flow depth , measured over deep channel

The dependent variables inc lude :

� - total discharge in the compound section

Qe - component discharge in deep channel
Q - component discharge in the flood p lains
F
V - average ve locity over whol e cross section
Ve ; V - component average velocities etc .
F

Any dependen t vari a b l e is a function o f the independent variab les

l is ted :

. . . Al . 1

2. These 1 2 independent variabl es wi l l yiel d 9 dimensionless groups ,

which could be derived in many different ways . To proceed , it is desirab l e
t o introduce s imilarity concepts , such a s those fami l iar when deal ing with
friction contro l led f l ow in simp l e cross-sections . Taking total discharge
as the dependent o f immediate practical concern , and also l inking S with g
because i t is the weight component down the s lope that has phys ical
s igni ficance , and considering the sys tem as one channe l :

3. Singl e channel method

QZ P QI P H Q k
se
k b
e
SF
= � ; ; ; H ; Al . 2
8gS 3 { J.l/ p H-h { (gA/W) H-h H
• . •

where

A = total cross-section area = � [ b ; b ; s ; s ; h ; H ]

A e F e F
p total wetted perimeter = � [ ditto ]
p
w = total water sur face width = � [ ditto
W
\) kinemat ic viscosity, J.l/p

These dimensionless groups are recognis ab l e as :

= 1 / f where f is the frict ion factor t reating the

who le section as one

= Reynolds number o f whole section

Q Froude number for whol e sect ion

{ ( gA/W)

H/ (H-h) ratio of channel f low depth to that on flood p lain

re lative roughness of channel

= relative roughnes s of f lood p l ain

= width/depth ( aspect ) ratio of main channel

width/depth (aspect ) ratio o f f lood p lain

side s lope of main channel

s ide s l ope of f lood p l ain

4. I f we are to proceed on these l ines , treating the who le section , then
some simp l i fication is possible i f we confine attention to :

(i) Rough-turbulent conditions on flood plain a s wel l a s in the deep

channe l (when viscous e f fects are no longer s ignif icant )

( ii ) Velocities low enough to avoid energy lossses due to surface waves

(when the Froude number becomes irrelevant ) .

Also the side s lopes o f the main channel may become less s ignif i cant i f the
aspect ratios are based on mean channe l widths , b ' and b ' , so that we may
e F
then reduce the prob l em to one with 7 independent variab les , with two
( bracketted) of lower significance perhap s :

Q
Al . 3
( 8g SA 3 / P )
• • •

I f there is equa l roughnes s on f l ood p l ain and in channe l , this further

simp l ifies to :

f • • . Al . 4

5. In genera l , s ing le channe l approaches use a restricted sub-set o f the

above :

f = �
1
[ k 5 ! R 1 x � [ H / ( H-h) ; b ' /b c ' J
F
• . • Al . 5
. • . Al . 5

where R is the hydrau l i c mean depth o f the who le section ( replacing H in the
preceding form) and the aspect ratios o f the separate zones are rep l aced by
the ratio between their widths .

6. Splitting dimensionless statements in this way i s stric t ly not

permissib l e , however . Consider the general relationship :

c = � [ a ; b] • • • A1 . 6
Rep lacing this by :

c = � l [ a ] x �2 [ b ] • • • Al . 7

l imits the relat ionship to one that would plot as a set of para l le l curves
on a log-log p l ot , c learly much less general than the function it seeks to
replace .

Rep lacing it by :

• • • Al . 8

means restricting it to a set o f para l l e l curves when p lotted to l inear

axes . So a lthough split dimensionless statements are convenient ,
especial ly in terms of data analysis , whether or not they are acceptab le is
a matter requiring j ustification . It is common in des ign practice to spl it
complex rel ations by a ssuming that the various factors influencing a
proces s can be a l l owed for separate ly , i . e . as suming the various factors
do not interact , but the leg itimacy of this procedure requires testing in
the particular case under review .

7 . Divided channel method

Cont inuing the as sumption that vis cous and surface wave inf luences are
neg l igible , the divided channel approach in its basic form s eparates the
cross-s ect ion into deep channel and f lood plain zones . Various as sumpt ions
about the planes of division between the zones have been considered in the
past , but here they are vertical p lains at the bank line , and the inter face
so created is not inc luded in the wetted perimeter of either zone . With
the further as sumption that the e ffect of the aspect ratio is sol e ly on
frict ion loss so that it may be accounted for by us ing the hydraul ic mean
depth , the method is basical ly a s fo l lows :

• • • Al . 9

where
 . • . Al . l O

and

• • • Al . l l

• • • Al . l 2

• • • Al . l 3

8. The above is defective as it ignores any interaction effect between the

separate zone s . Hence we must introduce correct ion factors , F and F to
e F
adjust the calcu lated channel and flood plain flows for the e ffec t o f the ir
interaction , or an overal l correction factor for the total flow , F ;
T

• . . Al . l 4

9. and Q are the basic values ca lcu l ated from the

In the above , Q
C F
friction formulae appropriate to their part icular features , e . g . smooth ,
rough or t ransit ional . The correction factors wi l l be functions of other
flow or geometric parameters . The minimum requirement from purely
dimens ional considerations is :

• . • Al . l5

• • • A l . l6

and

• . . Al . l 7

10 . Different authors have used different restricted forms o f the above ,

perhaps re lated to the scope o f exper iments they have made , for example :

F . • • Al . l8
C,F,T
For equa l roughnes s on flood p lain and in main channe l , this would reduce
to sets of curves giving corre ction factors p lotted against relative depth ,
H / ( H-h ) , with the width ratio as the third variab l e for each set . Different
sets wou ld be required for different values of the frict ion factor ratio -
but a further four possibly relevant parameters have also to be accounted
for , or shown to be insignifi cant . ( Four because the seven parameters o f
equs A1 . 15 t o 17 have been reduced t o three)

11. Let us cons ider the features to be expected where both the flood plains
and main channel are very wide relative to the ir flow depths : in the limit
this is the j unction at a bank l ine o f two semi-infinite sheet flows . The
expectation in that case is for the flow to be a ffected for a limited zone
e ither side o f the bank l ine , and is is worth consi dering therefore using an
addi t i ve/subtrac t i ve correction to the flows either side of the bank
rather than using mul tiplying correct ion factors . In the wide channel
case , then :

• • . A1 . 19

where

• • • A 1 . 20

• • . A1 . 2 1

Q and Q are the channe l and f lood p lain flows calculated by the
Cb Fb
appropr iate resi stance equation for the main channel and flood p lains
respective ly , ignoring any interact ion e ffect s : the additive/subtractive
corrections take account of the actual interaction . The question then
arises as to how best to "norma l ise" , i . e . convert to non-dimensional
form , these discharge correct ions . An argument can be developed that ·they
should be norma l ised by the local flow parameters characterising the
j unct ion between the as sumed semi-infinite sheet flows , e . g .
2
( H-h) (V -V ) .
C F

12 . The imp lication behind this is that the cross-section of the zone o f
inf luence has dimensions related solely t o the depth o n the flood plain ,
and any veloc ity defect/increment is bas ically proportional to the
difference between main channel and flood p l ain ve locities . However ,
Raj aratman and Ahmadi ( 19 8 1 ) considered this very point on the bas is o f
experiments i n a vertical s ided compound channel and demonstrated that the
width of the zone of influence was proportional to bank ful l depth , h , so
2
that norma l i s at iom by h (V -V ) or hH (V -V ) might prove pre ferab l e .
C F C F

13 . Zheleznyakov ( 1985 ) had ear l ier sugges ted the concept o f additive or
subtractive corrections to the flood p la in and main channel f l ows and
demonstrated that it was the loss o f main channel f l ow rather than any
increase in flood p lain flow that dominated the s ituati on . He went on to
suggest that changes could be expressed in terms of bank full f low . This
would not be appropriate for very wide systems approaching the semi- infinite
s ituation hypothecated above , however .

14. Thus there are several a lternative concepts for norma l i s ing the
suggested discharge correction . From the pragmatic point o f view , bank ful l
discharge i s a straight forward quantity to use , but s o a lso are the
a lternatives . Only by comparing the possible normal i sing procedures when
analysing experiments can we decide on the mos t relevent non-dimensi ona l
groupings . For very wide systems , the semi-infinite concept would l ead to :

. • . A 1 . 22

though from the dimensional analysis viewpoint the term ( H-h) in the l e ft
parameter could equa l ly we l l be any combination of H and h with the
dimensions o f an area , e . g . ( H-h) h or Hh .

15 . To cover compound channe l s in general , the function wou l d have to a l l ow

for the restriction o f the f l ow interaction by the channel width and edge o f
the flood p l ain . The relevant aspect ratios have there fore to be
re-introduced. Also , because of viscous influences , especial ly in smooth
laboratory channel s , it is conceivable that the Reynolds number may be
s igni ficant , a lthough it is hoped that the friction factors woul d be a
sufficient flow description for both smooth and rough s ituations . The
general ised s tatement thus becomes :
where (Area) is intended to cover any product of two independent variab les
of length dimensions to give a plausib le area of influence . Combinations
o f interest might be Hh , ( H-h) h , 2bH etc . As the FCF at Wal l ingford has a
fixed value o f main channe l aspect ratio . 2b/h = 1 0 , in terms o f data
analysis the use of Hh or 2bH would not be distinguishable : the latter is
always 1 0 t imes the former , and the issue could only be resolved by
resorting to other data sources with different main channel aspect ratio .

16 . This concept , equivalent to additive/subtractive correction to the

overall flow (or more basically in the separate zones of f low) can be tested
alongside other methods , for exampl e the use of discharge adj ustment factors
either for the separate zonal calculations or for the sum of thos e basi c
discharges . There is n o reason from the dimensional analysi s viewpoint to
prefer any particular method o f express ing the required correction to the
basic calculated flows . The criteria for choice might include :

- goodnes s o f fit
- insensitivity to some variables
- s implicity of function e . g . linearity
- convenience of appl ication in hydraulic design
APPENDIX 2

RES I STANCE FUNCTIONS FOR THE SERC-FCC AT WALLINGFORD

! . Background

1 . 1 The F l ood Channel Facil ity at Hydraulics Research , Wal l ingford ,

cons ists o f a compound channel moulded in cement mortar . I t is o f fixed
gradient , although a number of alternat ive cross-section geometries has
been tested . Tests were conducted by several independent research groups
concentrating on different aspect s , but in a l l cases imp ortant flow
parameters were the discharges at which the experiments have been conducted
and the corresponding mean depths of flow relative to the bed of the main
channel . From discharge and depth , with the known fixed gradient of about
1 in 1 000 , a l l other conventional measures of channe l performance can be
calculated , e . g . mean ve locity , Froude number , friction factor , Reynolds
number etc .

1 . 2 In the context of analys ing the experimental results for the

preparation o f a design manual , the res istance of the channel is of prime
importance , and this required the establishment o f the mos t appropriate
resistance formula for the Wal l ingford channel , based on the analysis o f
experiments conducted on channe ls without f lood plains . I f the funct ion
were not a good fit for simple channels , there would be much less prospect
of identifying and formulat ing the additional resis tance ari s ing under
compound flow conditions ; and indeed mi s leading conclusions could emerge .
Hence a priority task was a review of the basic resistance function . It
must be stressed that the conclus ions in this Appendix relate specifical ly
to the Wa llingford channe l ; they do not apply to the rougher channe ls in
engineering practice .

1 . 3 It might be thought that the choice o f the Colebrook-White transit ion

funct ion would have been automatic and uncontentious bearing in mind that
it has been in gene ral use , at least for the smoother range of manufactured
and constructed sur faces , for both pipes and channe l s . I f that had been
the case , all that would have been necessary was to assess a suitable value
of k . Analyses of the data by the various research teams , however , left
s
the matter in some doubt , because the different groups had used different
basic functions , including forms o f smooth turbulent equation , the
turbulent transition function o f Colebrook-White and Manning ' s equation , as
wel l as friction factor/Reyno lds Number p lots .

1 . 4 A perhaps surpris ing feature of the previous analyse s was that several
a lternat ive resistance laws were , at face value , equal ly val id , even
equations such as Manning ' s , which is normal ly regarded as restricted to
rough channel s , providing a good fit with constant n for the s imp l e
channel s . To avoid continuing confusion , as wel l as to provide a sound
basi s for further analysi s of stage/discharge data , reconsideration o f the
basic resistance function was therefore considered essential .

2 . Brief review of resis tance fun c tions

2 . 1 Most text books on hydraul i c s contain a review of hydraulic resistance ,

and inc lude a friction factor/Reynolds number diagram ( o ften ascribed to
Moody , see for exampl e Chadwick and Morfett , 1 9 8 6 ) based on the
Colebrook-White transition funct ion . Almost without exception , this
diagram wil l relate to pipe f l ow but a method of conversion to open
channe ls or non-circular cross-sections may be given , using the equival ence
for circular sections , namely :

R D/4 • • • 2. 1

2 . 2 This is an oversimp l i fication , because the Colebrook-White formu l a

derives from the smooth turbulent and rough turbulent functions obtained by
integrating the logarithmic vel oc ity distribution law over the f l ow
cross-section . There are constants o f integration that depend on the shape
of sect ion , and so an additional adjustment i s required over and above the
R = D/ 4 c onvers ion . In normal engineering the distinction is not very
important (because of uncertainties e lsewhere in the design process ) , but
in the accurate analys is of research results it could be s ignificant . The
two ver s ions of the rough turbul ent equation in the l iterature , for
circu l ar and wide open-channel s respectively , are :

Ci rcul a r sections :

1 /ff = 2 log ( k 5 / 3 . 7D ) 2 l og ( k / 14 . 8R)

5
. . . 2. 2
 Wide open-channels :

1/ {f = 2 log (k / 1 2 . 3R) • • • 2.3

8

This last form was obtained by Keulegan ( 19 3 8 ) who also showed that a
s imilar correction for channe l shape i s required for smooth turbulent flow .
The corol lary t o this i s that i f the Co lebrook-White transition function i s
to be adjusted for shape , botkb e l ements require the same adjus tment
factor ( Ackers , 1 9 5 8 ) so that the two ver sions of the function are :

Circular sec tions :

1/ {f = - 2 log [ (k / 1 4 . 8R) + ( 1 . 25 5�/R{ ( 3 2gRS ) ) ] • • • 2.4

8

Wide open-channels :

1/ {f = -2 log [ (k / 1 2 . 3R) + ( 1 . 5 1 0�/R{ ( 32gRS ) ) ] • • • 2.5

8

Unfortunately the need for adjustment to the smooth term i s not so

genera l ly recognised so that sometimes the wide channel adj ustment i s made
only to the roughnes s term in the transition function . In what follows ,
the wide-channel transit ion function will be taken as e qu . 2.5.

2 . 3 A maj or distinction between smooth and rough turbulent flows is the

influence o f f luid viscosity on the resistance function : in the former
case , the friction factor , f , depends on Reynolds number , decreasing with
increasing s ize and velocity ; in the latter case , the friction factor is
independent o f Reynolds number . This distinction also results in only
rough turbulent flow following a square law of resistance , i . e . velocity
be ing proportional to the s quare root of hydraulic gradient ( other
parameter s o f flow being unchanged) . The very popular " s quare law" Manning
equation may be thought of as an approximation to equation 2 . 2 or 2 . 3 , with
116
Manning ' s n given by k !26 (k in mm) It can be shown that the
8 8
approximation i s c lose for re lative roughnesses R/k between about 1 0 and
8
100 , which confirms application to many o f the rougher engineering
constructions and to natural channe ls . It fol lows from the form of the
Manning e quation that it should only be app lied to rough turbulent flow ,
when the f low is not being influenced by viscosity and so is independent o f
Reynolds number .

2 . 4 The transition between rough and smooth turbulent f l ows embodied in the
Colebrook-White function a l l ows for the progressive change from viscosity
dominated flow to roughnes s dominated f low as the Reynolds number
increases . The form this transition takes in a friction factor/ Reynolds
number p lot is not fixed . but depends on the character o f the roughnes s :
for examp l e , the Nikuradse diagram for a uni form coverage o f g lued-on sand
grains has different transitions from the Moody/Colebrook White chart which
i s for isolated protuberances . The latter has a much more gradual and
extended curve into the smooth law and this is thought to arise because the
i s o l ated roughnes s e l ements continue to exercise a local disruption on the
laminar sub-layer : they are not so readi ly submerged as is the uni form
coverage of grains with equal k value .
8

2 . 5 An important distinction between experiments conducted on pipes and

those on open channe ls is that the former wi l l be at a constant relative
roughness , because in any test series both sur face texture and flow cross
section remain constant : it is the hydraulic gradient ( s lope) which varies
as the discharge is varied . Open channel s , on the other hand , have the
f lexibi l ity o f cross-section change : whether they are at constant s lope or
variab l e s l ope wil l depend on the experimental arrangement , but the
Wal l ingford flood channel faci l ity imposes constant s lope , so that relative
roughness is not constant . Thus any sequence o f test results for a fixed
gradient open channel runs across the constant r e lative roughness l ines on
a Moody- type diagram rather than fo l lowing one o f them.

2 . 6 Myers and Brennan ( 1989 ) when analysing the Wal l ingford data
general ised the smooth turbulent function to the fo l lowing :

1 /{f C log ( Re{f) - D • • • 2.6

The smooth component o f the Colebrook-White formula for p ipes sets C = 2 . 00

and D = 0 . 80 , values deduced from Nikuradse ' s tests on pipes many years
ago . They were empirical adjustments to values deduced theoretica l ly from
the ve locity distribution : 2 . 03 was considered a more fundamental value o f
C , corresponding t o the generally accepted value for the turbulence
constant , K , o f 0 . 40 . For wide open channel s , Keul egan had suggested D =
1 . 08 . Us ing a s light variant o f equ . 2 . 4 ( Henderson , 1966) , a k value o f
8
0 . 06mm was deduced for the Wal l ingford channel . Myers and Brennan reasoned
that as Re* = u*k /� had a maximum value about the s ame as the value o f 4
S
suggested by Henderson as an upper l imit for smooth turbulent flow , the
f l ow should indeed be considered as smooth turbulent throughout . However ,
the l imit o f 4 is more appropriate to a uni form sand textured surface ,
being based on Nikuradse ' s rough p ipe exper iments : the Colebrook-White
transition extends from Re* = 0 . 3 to 5 0 , so the question about whether the
flow shoul d be considered as smooth rather than t ransitional was
unresolved .

2 . 7 The form o f equation 2 . 6 indicates that the main influence o f any

variation in the term D is to provide a vertical shift to a plot of 1{/f
against Re . The e f fect o f channel shape obtained by integrating the
velocity distribution a ccounts for a change in D o f 0 . 17 :
( 0 . 5 - ln 2 ) /K{/8= 2 ( l og 1 4 . 8 - l og 12 . 3 ) . However , s ince the time when
Keulegan studied the question , a great deal o f additional experimenta l
evidence o n turbul ent flows has become ava i l ab l e which has providing updated
equations for the turbulent velocity distribution in smooth channel s , which
in turn l ead to up-dated open channel friction functions .

2 . 8 The logarithmic vel ocity distribution l aw for smooth boundaries takes

the form :

• • • 2.7

where v is the local vel ocity at distance z from the wal l , and v* is the
shear ve locity at the wal l ( ={ ( • /p ) ) . There is s t i l l some contention over
the best values o f A and B , which are essent ial ly empirica l as they depend
on experimental measurement of ve locity distribution . The gener a l l y
accepted values i n a historic context have been those due to work by
Nikuradse ( 19 3 2 ) , C l auser ( 19 5 4 ) , Patel ( 19 6 5 ) and the consensus from the
Stan ford conference o f 1968 .

2 . 9 Integrating the velocity distr ibution over the depth for a wide open
channel g ives equation 2 . 6 , where :
c 2 . 3026 A/{8 • • • 2.8

D - (1 + ln 4{8 ) A + B . . .2.9

and on this basis it i s poss ib l e the smooth channel res i stance functions
that may derived from the ' hi s toric ' velocity distributions may be
compared :

Source Date A B c D

Nikuradse 19 3 2 2 . 50 5 . 50 2 . 03 5 1 . 083
C lauser 1954 2 . 49 4 . 90 2 . 027 1 . 283
Patel 1965 2 . 39 5 . 45 1 . 946 0 . 96 8
Stanford 1968 2 . 44 5 . 00 1 . 986 1 . 188

2 . 10 From the appearance and feel o f the moul ded surface a t Wal l ingford , i t
is apparent that the surface i s not far from smooth : it certainly can not be
characterised as rough . In those c ircumstances the Manning equation would
not be expected to provide a good fit to resi stance data without using the
coefficient , n , in an art i fi c ial way to i l lustrate the departure from rough
turbulent " s quare l aw" resistance : yet as previous ly mentioned it provides a
surprisingly good fit without varying n . The reason for this wi l l be
discussed in detail l ater , but one factor is that open channel tests at
constant gradient are not we l l conditioned to distinguish between smooth
turbulent , rough turbulent and trans itional conditions , given that a l l
contain at least one parameter whi ch is based on the parti cular set o f
experiments . A shift from the c l as s ical smooth turbulent l ine o n an f , Re
p lot , whi l s t remaining approximately para l l e l to it , can be obtained by
altering D in the l ogarithmi c smooth-turbulent formul a , by a non-zero value
of k in the transition formula for a wide channe l ( or indeed in the
S
corresponding p ipe equation) or by varying depth and hence relat ive
roughness in a rough turbulent formula .

2 . 1 1 These three equations are shown in fig A2 . 1 : the Myers and Brennan
( 19 9 0 ) modi fied smooth turbulent equat ion , the wide channel transition
function with kS = 0 . 07 mm and the Manning equation with n = 0 . 0 10 . The
upper diagram shows the conditions for the Wal lingford main channel
cros s-section at a gradient of 1 : 1 00 0 : only at the sha l l owes t flows is
there much difference between the three funct ions .

2 . 1 2 On the basis o f this review , from pure ly theoretical considerations

the Manning equation should be ruled out on the grounds that app lying a
rough turbulent equation to a rather smooth channel at slack gradient would
be misleading . Any agreement with the Manning equation is the fortuitous
result of having a fixed gradient . Had the gradient varied , a fixed
Manning ' s n could not have provided an acceptab l e fit to the data . In
fact , the flow is not rough turbul ent and some viscous influence is
expected . The appropriate resistance function from theoretical
considerations alone is expected to be the transition law or a modification
o f the smooth turbulent law .

3 . Analysis of resistance cal ibra tion da ta

3 . 1 Depth discharge data were obtained in the various phases of the

invest igation for ' simple ' , i . e . non-compound flow conditions , with bank
s ide-s lopes of 2 : 1 (hor : vert ) , 1 : 1 and vertical . In a l l cases the bed
width was 1 . 5 m and the channe l hydraulic gradient very c l ose to the mean
channel gradient o f 1 . 027/ 1000 . The average stage was obtained by taking
the mean of the measured depths over the experimental length of wel l
e stab l i shed uni form flow , the dis charge was measured by orifice meters in
the supply lines , the hydraulic gradient was asses sed from the s lope of the
total energy line . Ve locity was derived from continuity , knowing the mean
cross sectional area . Viscosity was obtained from the water temperature .
On the who le , the water temperature remained close to 1 5 ° C . Unfortunate ly
at the time thi s analysi s was carried out not all the measured temperatures
had been added to the data set , and where they were mi ssing a figure of
1 5 ° C was assumed , but in view of the small variation from this standard
value any error introduced would be ins ignificant . (This assumption did not
apply to the ana lys is of compound channe ls : measured tempe ratures were then
available . )
3 . 2 There were four sets of information referred to as channe ls 1 to 4 :

C hanne l Side Date o f Number Depth range

number s lope , s experiments of tests mm

1 2 Jan-Feb 89 14 45 - 150
2 1 Nov 86-S ept 87 28 25 - 1 49
3 1 Sept 87 13 150 - 296
4 0 Oct-Nov 88 11 40 - 148

(These dates refer to the bulk of tests at the stated s ide-s lope : a few
tests at other dates are included)

Channe l 3 relates to tests carried out as an extension to those on channel

2 , having extended the banks upwards with temporary side s lopes to cover
the range of depths of interest with over- bank flow . One of the tests on
channe l 2 seemed to be away from the general trend , so some analyses were
made omitting it , though with very l ittle effect on the overall picture .
As channe l s 2 and 3 were in e s s ence one channe l tested over different depth
ranges , some analyses were made combining these data : and also combining
all four data s ets .

3 . 3 There has been an observab le change in the surface texture o f some

sections o f the main channe l with time . There are some hard , possibly
cal careous deposits over sections where the bed level was marginally below
the perfect l ine . These have perhaps occurred because of s lower draining
and subsequent drying out in these zones in the periods between tests , so
giving a deposit from the hard laboratory water . The texture of these
zones seemed even smoother to the touch than the original surface ( they
were somewhat s lippy) but they may have introduced minor irregularities
that could have increased roughnes s rather than reducing it . A secondary
obj ective of the data ana lys is was to show whether any change was
s ignifi cant .

3 . 4 The equations cons idered were :

- Colebrook-White transit ion in original form , equ 2 . 4

The same but modi fied for wide channe ls , equ 2 . 5
 Manning equation
Generalised smooth turbulent , e qu 2 . 6

The aim o f the analysis was to find the best fit coefficient values for the
data set , and then to assess the variation about that funct ion by
c a l culating the root mean s quare error , both as a percentage and as an
abso lute error in predicted ve locity . The first two equations have k as
8
the empirical coefficient ; Manning ' s n app l ies to the thir d . The
general ised smooth turbulent function has two parameters , C and D , whi ch
g ive greater flexibilty in the fitting procedure - e f fectivel y a shift from
the smooth turbul ent l aw and a t i l t i f need b e . The results are given in
Table A2 . 1 .

3 . 5 The broad conc lus ion from the results in Tab l e A2 . 1 is that there is
not much to choose between the four equations tested . The Colebrook-White
transit ional e quation is in mos t cases margina l ly the worst whether in
terms of percentage discrepancy or absolute discrepancy , but it is also
apparent that most o f the discrepancy from any o f the equations i s due to
inevitab le experimental tolerances rather than basic inadequacy of the
theoretical functions . Apart from channe l 1 ( 2 : 1 s ide s lopes ) , the
wide-channel modification of the C-W equation is a s l ightly better fit .
Perhaps surpris ing ly , the Manning equation is better than either form o f
transition , whether taking data s e t s separately o r in combination . Despite
its extra degree o f freedom , only in one case ( channel 4 , vertical s ides )
does the general ised smooth turbulent function turn out to be s ignificant ly
better than the o thers . In this parti cular cas e , values o f A and D have
emerged from the best-fit routine that differ cons iderab ly from those for
the other data sets : they have provided a t i l t that better accommodates
these particular results .

3 . 6 The k values for the C-W and wide-channel equations provide a

s
sensitive measure o f any roughnes s changes with time . For the wide-channel
formula , the mean value o f k for the firs t set o f experiments carried out
S
from November 1986 , to January 19 87 on the channe l with 1 : 1 s ide s lopes
( channel 2) was 0 . 07 1 mm : a low value indicating how near ly smooth the
steel f loated finish to the channe l was . This increased in tests made in
September , 1987 on channe l 2 to about 0 . 09 0 mm , showing marginal roughening
but within the range o f experimental error . Further tests in with the same
side-s lope o f 1 : 1 but at depths above 0 . 1 5 m , a l s o in S eptember 1987 , gave
k = 0 . 046 mm , seemingly smoother . Later tests in October and November
s
1988 with vertical s ides ( channel 4 ) gave a value of 0 . 1 1 1 mm , but then
tests made in January and February 1989 with side s lopes of 1 : 2 ( channel 1 )
yie l ded a mean roughness o f 0 . 0 10 mm , implying virtual smoothnes s , and
suggesting that the deposition has smoothed rather than roughened the
channel . However , the deposits were a lready present in the 1988 tests .

3 . 8 The Manning roughnes s values provide a somewhat less sens itive measure
of any change in resistance of the basic channel . The first tests results
may be regarded as setting a standard and then the average values from
subsequent groups of tests can be used to indicate a percentage increase or
reduction in c a l culated vel ocity or discharge , as in the fol lowing tab l e :

Channel S ide Tes t date Mean n Percentage change in V

number s lope Increase Reduction
H:V Smoother Rougher

2 1:1 Nov 86 - Jan 87 0 . 0 1000

2 1:1 Sept 87 0 . 0 10 1 4 1.4
3 1:1 Sept 87 0 . 0 1 002 0.2
4 Vert Oct - Nov 88 0 . 0 10 0 1 0. 1
1 2: 1 Jan - Feb 89 0 . 00965 3.5

3 . 9 Factors apart from change o f surface texture with age may influence
calculated values of k and Manning ' s n , for exampl e any effect of change
s
in cross sectional shape not ful ly accounted for by the hydraulic radius R .
The conclusion , however , i s that any changes o f roughness were minor and
with no apparent direct association with age . Thus a l l test data may be
regarded as a sing le set for a channel o f constant roughness . Bearing in
mind experimental tolerances , real change can not be identi fied with any
confidence . Thus al l test data were regarded a s a singl e set for a channel
o f constant roughness .

3 . 10 Returning to the choice o f a pre ferred resistance function , tests over

a wide range of depths are best suited to this purpose . The 40 tests on
channel s 2 and 3 covered depths from 2 5 - 3 0 0mm , and a Reynolds number range
from 20 000 to 9 0 0 000 . Tabl e A2 . 1 shows the order of per formance to be :
 1- Manning ;
2- modified smooth-turbulent equation ( S -T) ;
3 - wide channe l ;
4- Co 1 ebrook-White equation ( C-W) .

The distinction is smal l , and does not by itsel f provide a rational basis
of selection . Al l have a percentage discrepancy of between 1 . 7 and 2 . 0
percent . However , careful inspection o f the detailed p lots showed that the
modi fied S -T equation had best fo l l owed the trend o f data at shallow
depths . Figs A2 . 2 and A2 . 3 show the discrepancy between computed vel ocity
and measured vel ocity for channel s 2 and 3 ( 1 : 2 side s lope s ) and for a l l
results respective ly . (Actual velocities were o f the order o f 0 . 5 m/s at
minimum depth , 0 . 8 m/s at a depth o f 0 . 1 5m , bank ful l when operating as a
compound channel , and 1 . 2 m/s at maximum depth o f 0 . 3 m . ) Although shal low
depths in the main channe l are not important , they are signi ficant in the
analysi s o f compound channe l s : some o f the most il luminating results were
expe cted to be those with shal low flows over the flood plain . Thus the
preferred equation was the general ised smooth- turbulent function with C =

2 . 02 and D = 1 . 38 , i . e .

1/{ f 2 . 02 log ( Reff) - 1 . 3 8 • • • 3. 1

3 . 1 1 This is is remarkably c lose to the resistance law deduced from the

velocity distribution by integration for the wide channel case with
C lauser ' s ( 19 5 4 ) parameter values ( C = 2 . 02 7 , D = 1 . 28 3 ) . It is a l ittle
further removed from the classic smooth turbulent law o f Nikuradse , in
e f fect shift ing several percent towards increased resistance . As already
mentioned the c l assic Keulegan value for the second parameter for wide
channe ls is 1 . 08 , so equation 3 . 1 is a further shift o f 0 . 3 in the 1/ f f
value . Reference to the more recent work by C lauser suggests that this
might largely be explained without recourse to any s ignificant increase in
resistance over a smooth sur face , although the S tanford consensus would
s t i l l leave a change of 0 . 20 in the value of D to be explained as increased
resistance .

3 . 1 2 Morris ( 19 5 9 ) put forward some novel concepts on resistance functions ,

distinguishing between three types o f flow :
 - semi-smooth turbulent generated by isolated roughness elements
- hyper-turbulent , where there is inter ference between the
wakes from roughness elements
- quasi-smooth , where there were additional localised sources
o f energy loss , such as flow skimming over grooves

The last of these provided a shi ft in the friction factor I Reynolds number
p lot whil s t remaining paral l e l to the smooth- turbulent l ine . Possibly
the Wal l ingford faci l ity is fol l owing this quasi-smooth function because of
such localised additional energy losses , even if they cannot be identified .
Whether the resistance l aw o f equ 3 . 1 represents quasi-smooth or ful ly
smooth f l ow with ful l a l lowance for channel shape is immaterial in terms o f
data analysis . Because many readers wil l be unfamil iar with the term
quas i-smooth , the l atter explanation has been adopted in the main report .

3 . 1 3 There remains the question o f why the Manning e quation provides a good

fit to data from a smooth channe l . The Manning equation is exponential

rather than logarithmic , and a wel l-known exponential smooth-turbulent
formula is that proposed by Blasius ( 1 9 1 3 ) :

114
f = 0 . 08 Re . . • 3.2

This was derived as a good fit to experimenta l data over a particular range
of Reynol ds number , but let us examine a more general form of the Blasius
type o f formula :

-b
f = B Re . . • 3.3

which can be expressed in detail as :

-b
8 gRS/V= B ( 4VR/� ) • . • '3 . 4

This in turn yie lds :

. • • 3.5
3 . 1 4 For R to appear to the 2/3 power as in the Manning equat ion , b mus t
equal 0 . 2 , not very different from the Blasius value o f 0 . 25 . Inserting
this value then gives a Manning " look a l ike " :

• • • 3.6

Thus one could consider the appl ication of Manning in the present context
to be for quasi-smooth conditions with the coefficient n depending on
viscos ity , gravity and hydraul i c gradient as opposed to its usual role as a
descriptor of surface texture :

n = [ ( 8 g/B) 5 /9 ( 4/ u ) 1/9 S l/ 1 8 3.7

] -1 • • •

Because s lope remained constant and water temperature approximately s o in

the Wa l l ingford tests , the Manning equation provided a good fit : n as
defined by equation 3 . 7 remained cons tant . If the Blasius value o f b had
been retained , name ly 0 . 25 , then the hydraul ic mean depth would be rais ed
to power 0 . 7 1 4 rather than 0 . 667 , but conceivably would have also given a
good fit .

3 . 1 5 For the best fit value of n = 0 . 0 1 0 to apply , with S = 1 . 027x10-3 and

u = 1 . 1 4x1 0-6m/s , then it can be shown that B = 0 . 20 , and hence the
Blasius/ Manning smooth equation become s :

-0 20
f = 0 . 20 Re · • • • 3.8

This would be a more appropriate express ion o f the Manning- like resistance
of the Wal lingford channe l . I t would p lot on fig A2 . 1 virtua l ly identical
with the Manning l ine shown , pass ing through f = 0 . 02 , Re = O . lx 1 0 6 ,
4 . Rod roughness

4 . 1 Some o f the experiments on the F lood Channel Facility at Wal l ingford

have been carried out with the flood p lains roughened by vertical rods
extending through the ful l depth of water . In order to estab l ish the basi c
resistance function for this type o f roughening , data i s available from a
set o f seven tests carried out with the main channel roughened with the
same pattern o f rods used under c ompound channel conditions . Thes e basic
sing l e channel tests covered depths of f l ow from 44mm to 1 19mm , a l arge
part of the range of f lood p lain depths observed in the roughened flood
p l a in tests .

4 . 2 The pattern o f rods used consisted o f a triangular distr ibution , of

ang le 6 0 ° . This was designed to have a density o f 1 2 rods per m3 , and so
the s ides o f the e quilateral triangles forming the grid was 0 . 3 10m . This
was the spacing between the rods transverse to the f l ow , and the
longitudinal spacing o f rows was there fore 268 . 5mm . See fig A2 . 4 .

4 . 3 Under these conditions the res istance to f l ow i s made up o f the drag o f

the rods and the shear force a t the channel boundaries . I t i s assumed that
these may be treated separate l y as if the presence of the rods does not
influence the boundary drag of the channel sur face , except through the
increase of ve locity imposed by the b l o ckage e f fec t . Also it i s assumed
that any influence of the vertical vel oc ity distribution on the drag on
each rod can be accommodated by incorporating a suitab l e distribution
coefficient into an equation that uti l i s e s the mean channel ve locity
c a l cu lated a l lowing for the b lockage e f fect o f the rods , by using the net
cross sectional area in the p lane of the row of rods .

4 . 4 The equation for the drag at the s o l id sur face was derived ear l ier and

is given by equ 3 . 1 , app l ied in this case with subscript s denoting that
part of the total friction factor arising from shear at the s o l id surface :

1 /{f = 2 . 02 log ( Reff ) 1 . 38 4. 1

s 5
• • •

where :

the friction factor arising from the drag on the channel perimeter
fs
Re the Reynol ds number o f the f l ow as a whole
4 . 5 The drag o f the rods arises from three sources : internal vortex sheet

drag due to flow separation behind the rods ; free sur face drag ar ising from
induced waves ; and skin friction on the rods themselves . The se e f fects
might be af fected by wake inter ference , i . e . each rod may provide some
she lter ing o f the rod next downstream . The first o f the se components is the
dominent one : the so-ca l led form drag .

• • • 4. 2

where

F the form drag o f the rods per unit length o f channe l

ROD

c0 = the drag coe ffic ient

N number o f rods per unit channe l length
d dia o f rods
z = flow depth

p speci fic mass o f fluid

a = ve locity distr ibution coefficient
U the mean ve loc ity over the flow depth

The ve locity distr ibut ion coefficient al lows for the variation o f ve loc ity
over the length of the rod , i . e . the depth of flow . The depth mean
ve loc ity U is calculated a l lowing for the b lockage of the transverse rows
of rods :

u Ql (A - n z d) 0 0 0 4 3
0

where

Q discharge
A channel cross-section
n number o f rods in each row

On this bas is , a b lockage coeffic ient , � . may be defined :

2 2
� = ( U/V) = n z d/A)
-

(1 - . . .4.4
4 . 6 The total drag per unit length o f channe l is then given by :

. . .4.5

where

P wetted perimeter o f channe l

4 . 7 The drag equat ion can be converted into a conventional form o f

res istance equat ion us ing the force ba lance equat ion :

F = pg A S pgRSP 4. 6
TOT
• • •

where

g acceleration due to gravity

A cross sect ional area of flow
S channe l gradient
R hydraulic mean depth , A/P

f = 8gRS /V 4. 7
TOT
• • •

where

f =
overa l l friction factor
TOT

4 . 8 Equation 4 . 7 with the value of f obtained from equ . 4 . 1 formed the

S
basis o f analysing the test data . Note that the Reyno lds Number , Re , for
calculating the sur face resistance incorporates U rather than V . For the
main channe l calibrat ion tests , N was given by 1 0 / 0 . 2685 18 . 62 per unit
length. Taking CD as the unknown , all other parameters in equ . 4 . 7 were
known , so that CD could be ca l culated . ( When us ing equat ion 4 . 7 in the
reverse direction with f or V as unknown , iteration is required because
TOT
f depends on the overa l l Reyno lds Number , which in turn depends on the
S
unknown mean channe l ve locity . )
4 . 9 The fact that res is tance is generated over the ful l depth o f flow by the
rod roughnes s , with a minor part generated by the sur face drag at the solid
boundaries g ives a more-than-usual ly uni form velocity distribution in the
vertica l . Thus the role of a is probably sma l l compared with the blockage
coefficient , in this case approximately [ 3 1 0 / ( 3 1 0 - 25 ) ] = 1 . 18 , though
varying with fl ow depth because the cross section i s trapezoida l .

4 . 1 0 The bas i c as sumption that the surface drag can be asses sed by ignoring
the presence o f the rods ( except , o f cour s e , in respect o f the blockage and
the reduction in channel Reynolds Number because the extra drag reduce s
mean velocity ) was open to question . I t might be argued that the surface
drag would be reduced b ecause in the wake of the rods the velocity close to
the boundary would be les s than averag e , and could even be reversed over
some area behind each rod . On the other hand , the e ffect o f the rods wil l
produce irregularity o f the transverse distribution o f velocity and this
would increase the average sur face drag . Furthermore , the di s ruption to
the boundary layer might modify the basic smooth l aw . There was no way o f
knowing which direction any change would be , l e t alone quantifying i t from
previous knowl edge . The maj or e f fect is a lmos t certainly due to blockag e ,
which was taken into account through � . Some preliminary analyse s with this
calculated sur face drag modi fied by factors e ither above or below unity did
not provide any improvement in the correl ations .

4 . 1 1 Figure A2 . 5 shows the variation o f C calculated as above with the

D
channel Reynol ds Number . Figure A2 . 6 shows e ssentially the same
information , but p l otted against the ratio o f flow depth to rod diameter :
Figure A2 . 6 al so shows the calculated values o f the blockage coefficient .
The experimental resul ts display smooth trends with very l ittle s catter . C
D
varies from a minimum o f about 0 . 9 7 at the sha llowest flow tested (minimum
Re ) through a maximum approaching 1 . 22 at intermediate depths , dropping
again at the deepest flow (maximum Re ) to 1 . 1 6 . These values may be
compared with values for the drag coefficient o f isol ated cyl inders in the
l iterature : see for examp le Rouse 1 9 5 0 . The Reyno lds number of the rods
themse lves varies over a very narrow range , 4 1 0 0 to 4600 , because the mean
channe l ve loc ity varies only from 0 . 19 to 0 . 22 m/s . This is a range where
the drag coe fficient is not expected to show any rapid change with the rod
Reyno lds Number , so it is most unlikely that the variation of drag
coe ffic ient is a Reyno lds number e f fect .

4 . 1 2 The drag coe f f i c ient to be expected wi l l depend on whether the rods

are effectively smooth or rough . The former would g ive about 0 . 9 5
according to previous ly avai l able informat ion : the latter would be expected
to be higher though data at this range of Reyno l ds Number are lacking . A
further possible influence is the e ffect of the wake from the rods on those
downstream from them . However , the actual long itudinal spac ing between
rods in this case was over 20 diameters , so al though undoubtedly there
would be some res idual ef fect , it was probably very smal l . It would ,
moreover , remain the same in a ll tests s o would not cause variation in the
drag coefficient .

4 . 1 3 The above analys i s takes no account of wave drag that could arise
because the rods are " sur face pier cing " elements . It might be anticipated
that if wave drag were s igni fi cant , following the work of Froude on ship
resistance , it would be expected to depend on channel Froude Number ,
V/{ (gz) . Thi s would also a f fect any sur face interference between the rods
as the pattern of waves would be ang led to the channel axis as a function o f
Froude number . Although the channe l Froude Number var ies over quite a
narrow range , from 0 . 3 at the sha l l owest flow to 0 . 2 at the deepest , it is
conceivable that the sur face wave pattern could pa ss through some sens itive
zone within the range of these tests . Indeed Froude found that the wave
drag showed a peak value at modest Froude number s , fal l ing before
increasing again at high Froude numbers . The variation of overa l l drag
coe fficient may therefore arise from the free surface e ffects . The
conclusion was that the drag coefficient is best expressed as a function o f
relative depth , i . e . flow depth/rod diameter .

4 . 1 4 A curve fitting exercise gave the following formula for the drag
coe ffic ient :
For 1 . 7 5 < Z* < 6.6 . . . . • . .

acD = 1 . 1 8 4 - 0 . 277 z * + / ( 0 . 529 z* - 0 . 843 ) . • • 4.8

where

z* = flow depth/rod diameter

This is shown also on Figure A2 . 6 and provides a good fit over the range o f
data for the Wa llingford tests .

4 . 1 5 There remains the problem o f extrapol ation outside the range o f test
results . From the evidence of past research on the drag o f cyl inders , it
seemed inappropriate to al low the value of C to drop below 0 . 95 at sha l l ow
D
depths , which was the value obtained at the sha l l owest flow considered . It
might be argued that allowing for int eraction between the rods ( the
influence of the wake from one on the rod in l ine downs tream) might reduce
the expected drag coefficient , there is no direct evidence o f this . At
higher depths in the Wal l ingford f lume , up to the maximum Z* value o f 6 , the
empirical equation predicts a drop in C towards 1 . 05 . It is not necessary
D
to extrapolate beyond this for this series of tests , but work at Glasgow by
Ervine and Jasim (private communication) suggests that higher values o f Z*
woul d give a continuing downward trend towards a value o f perhaps 0 . 8 or
0 . 85 as a limiting value for Z * > 10.

4 . 1 6 When app lying the functions to the tests o n compound channel s , there
was a minor comp lication , because the number o f rods in alternate rows
differed . Thus the basic equations for the combination of drag on the rods
and the surface drag were re formulated , as with dif ferent rod numbers in
alternate rows , the blockage coef ficients differed in alternate rows .

5 Conc l usions

5 . 1 The SERC Flood Channe l Facili ty at Wal l ing ford is effectively smooth .

5 . 2 From the empirical and pragmatic point of view, there was l ittle to
choose between four resistance functions that were compared with the data
from tests on simp l e channe l s . Thes e were the Co lebrook-White equation for
t ransit ional turbulent f l ow in pipes , the convers ion of this to a
wide-channel form , the Manning equation and a generalised logarithmic
smooth turbulent function.

5 . 3 Most variation between the test data and any of these established
equations for flow resis tance was due to experimenta l s catter : tests in an
open channel at fixed s lope are not wel l conditioned to differentiate
between resi s tance funct ions .

5 . 4 On theoretical grounds , the Manning equation should not be used for

.
analysing flow in smooth channels . I t i s most appropriate for
rough-turbulent flow , which is not the situation in any of the tests in the
Wal l ing ford f lood channe l faci l ity .

5 . 5 However , the Manning equation provided a good fit to the experimental

results . This was so because an exponential formula with the hydraulic
mean depth raised to the power 2/3 can also be derived from a power-law
smooth-turbulent function , with the coef ficient n dependent on vis co sity
and gradient , rather than surface texture .

5 . 6 In the context o f the Wal l ing ford f lood channel facil ity , a preferable
exponential formula is a modification of the Blas ius smooth turbulent
equation to suit wide open-channel conditions :

- 0 20
f = 0 . 20 Re ·
5 . 7 The wide range o f depths covered in tests on the channel with 1 : 1 s ide
s l opes made these data the mos t suitable for selecting a preferred
equation . Paying due weight to sha l low depths of flow , which are o f
significance in the analysi s o f the tests on compound channel s with flood
plains , a special form of the smooth-turbulent e quation was derived and
recommended for a l l subsequent ana lyse s o f Wal l ingford test data :

1 / {f 2 . 02 log ( Re{ f ) - 1 . 3 8

5 . 8 Al though there was an ob se rvab le change in the sur face character o f

some sections of the t e s t fac i l ity due to a hard depo s i t , this appears t o
have had very l it t l e influence o n the f low resistance , a lthough
experimental tolerances make any influence difficult to detect or
quanti fy .

5 . 9 The use o f other equations in ear l ier analyses of the Wal l ingford
results , including some pub l i shed paper s , did not introduce s ignif icant
errors , but was potential ly confusing .

5 . 10 I t is preferab l e to dist inguish between roughnes s coefficients and

resis tance coe fficients . The former wi l l not change because o f the flow
cro ss section becoming compound as they de fine the phys ical roughnes s o f
the channel which wil l be unchanged . The extra resistance arising from the
interference e f fects with flood p lain flow is best expressed as an
adj ustment to the shear streses , friction factor , velocity or discharge ,
leaving the bas i c roughness coe fficients unchanged .

5 . 1 1 The bas ic resis tance o f the Wal l ing ford channel when roughened with
ful l depth rods can be des cr ibed by combining the speci fic form of smooth
law for the channel surface with the additional drag due to the rods .

5 . 1 2 A form o f drag coefficient is used which also incorporates a ve locity

distribution factor . The values obtained are within the range of
expectation , bear ing in mind the va lues of drag coefficients for cyl inders
g iven in the l i terature and the actual blockage ratio . The drag
coe fficient has been expressed as a funct ion of the relative depth o f
flow .
5 . 1 3 Basic resistance calculations for rod roughness as in the Wal l ing ford
tests may be based upon the fo l lowing set of formulae , which al low for
different numbers of rods in al ternate rows :

�1

For 1 . 7 5 < Z* < 6 . 6 :

acD = 1 . 184 - 0 . 2 7 7 z * + { ( 0 . 5 2 9 z* - 0 . 843 )

e l se aCD = 0 . 9 5

1/{f = 2 . 02 log ( Reff ) - 1 . 3 8

s 8

where

Re = Reynolds number o f blocked channe l = 2 V R ({ � 1 +{ � 2 ) /�

� 1 • 2 = blockage effect , i . e . s quare o f area ratios for a lterate rows

n = number of rods of dia d across channel / fl ood p lain , rows 1 and 2
1 •2
N = number o f r ods per unit length o f main channel / f lood p lain , in ditto
1 ,2
z = depth of f l ow in main channel / on f lood p lain
A gross cross sectional area o f zone under consideration
f = friction factor due to smooth boundary
s
f ove ra l l friction factor
TOT
V nominal ve locity given by component di scharge/A
aCD = effective drag coe fficient of r ods
z = z/d
*
R = hydraulic mean depth A/P , for zone under consideration
S = hydraul ic gradient (water surface s l ope )
 TABLE A2 . 1 . ANALYS I S OF RES I S TANCE DATA

Channe l Side Range o f No . o f Colebrook-White Wide channe l trans Manning equation Mod . smooth turb ' t
s lope depths tests ks RMS error ks RMS error n RMS error A , D RMS error
hor/vert mm mm % of V mm % of V % of V % of V
(V , cm/s ) (V , cm/ s ) ( V , cm/ s ) (V , cm/ s )

1 2 45-150 14 0 . 028 1 . 66 0 . 0 14 1 . 84 0 . 0096 1 . 61 2 . 00 t o . 9 6 1 . 64

( 1 . 20) ( 1 . 33) ( 1. 19) ( 1 . 18)
2 1 25- 1 49 28 0 . 085 4 . 77 0 . 05 0 4 . 64 0 . 0096 4 . 36 2 . 20 , 2 . 1 2 4 . 27
( 2 . 86 ) ( 2 . 85 ) ( 2 . 58) ( 2 . 84 )
2 1 25- 149 27 0 . 096 3 . 96 0 . 059 3 . 82 0 . 0 100 3 . 66 2 . 08 , 1 . 67 3 . 49
( 2 . 15 ) ( 2 . 13 ) ( 1 . 96) ( 2 . 12)
3 1 150-296 13 0 . 065 1 . 17 0 . 04 1 1. 11 0 . 0 100 1 . 10 1 . 9 1 t 0 . 88 1 . 06
( 1 . 12) ( 1 . 08) ( 1 . 05) ( 1 . 03 )
>
4 0 40- 1 50 11 0 . 075 3 . 86 0 . 04 1 3 . 62 0 . 099 1 3 . 46 2 . 53 , 3 . 56 2 . 22
( 1 . 81) ( 1 . 72) ( 1 . 65) ( 1. 26)
2 & 3 1 25-296 40 0 . 08 1 3.51 0 . 05 0 3 .31 0 . 0 100 3 . 06 2 . 02 , 1 . 3 8 2 . 97
( 1 . 94) ( 1 . 89 ) ( 1 . 72) ( 1 . 84)
1,2,3,4 0, 1,2 25-296 65 0 . 072 3 . 57 0 . 04 1 3 . 44 0 . 0994 3 . 30 1 . 9 1 , 0 . 84 3 . 38
( 1 . 99) ( 1 . 97 ) ( 1 . 90) ( 1 . 99)

The bracketted error figures are the root mean square errors i n ve locity for the series o f experiments ,
comparing measurement with best-fit formula predi ction . The unbracketted figures are the RMS per centage
di fferenc�s .
 •'

�
Ul
0 t ? 10
.- I
X X

I I -
E' -
10 (/)'?
.- a� O
• a_.-
2-..Q
i If 5 -
.J: Ui
X
c I -
Ci. m 3: • I
.I I
a> C O
-c ·H -
-c as (/)
� > ::1 ,
u::: .......
62 -
I
I
-
l
·�
1
-

jl 0.5

-
1/
If
a>
a: I· 0.2 a>
a:
cpf
li
I·
.- " .-
0 l it- 0.1 ci
-
?i .
-
I
I I
0.05 -
? If -
I
I I -
I .
I I 0.02
-
I l
p .
I .I
I
.I
0.01 -
I
I .I
0

C\1 I C\1
C! C!
C\1 .- 0 M C\1 .- 0
C! C! C! C! C!
0 0 0 0 0
..... .....

0 Modified smooth
+ C-W for wide channel
• Manning
JBW/60/10-91/30

Fig A2.1 Comparison of friction formulae: (a) for fixed slope;

(b) for fixed depth
 ,
ea·
l>
1\)
.
1\)
SERC-FCF data - generalised smooth turbulent results
0.1 00 ,....
en
3
0
0 [Link] 1--
-
:::T
(')
:::T
Q) 0.0600
:::s D
:::s
('f)
-

(') 0.0400
Q)
-
-·
0"' D D
...,
Q) 0.0200 1- D D
-
D
5' -
D D

I
.i!! D ctJ D D
:::s E D D D
:- 0.000 ! ! I
o '
D ,o I I n f!J - g I I I I
0 0 D
_,.
..
.. dJ D
-
1\) w 0.0200 0.0400 0.0600 [Link] 0.1 00 0.1 20 0.1 40 0.1 60 0.1 BO 0.200 0.220 0.240 0.260 0.280 0.300
D 0 D
en D D D
-·
-0.0200 1-- D
c.
('f)
en
-
0 -0.0400 I- D
"'C
('f)

-0.0600 1--
D

-[Link] I-

'-

!
0
-0.100 L..
Stage (m)
..
D Chan 2 and 3
§
.,
cc
-·

)>
1\)
.
(.,.)

0.1 00 r-
SERC·FCF data · generalised smooth turbulent results - all data
en
3
0
0 0.0800 1-
-
::r
(")
::r X
m 0.0600 1-
:::J
:::J
(1)
-

(") 0.0400 1-
� X X
C'
-·
X � X X
""'

a 0.0200 r X
X
x >'x X X
X X
X X
0 X X xxX X X
-·

:::J X X
�X , x ! X
� I I � I I I I
. . I I J I X I
.... 0.000
....
0
.... 00 0 .0400 0.0600 o.o�o �o � .1 20 : 0.14� 0.1 Go 0 . 1 80
X X
0.200 0.220 0.240 0 .260 0.280 0.300
�
- w x x X X
0x xx X X X
; -0.0200 1-
(fJ X X
c X
;:::::; X
(fJ X
-0.0400 1-

X X
-0.0600 1-

-0.0800 1-

....

I ·0.1 00 L
Stage (m)
�
g x All data
 -
55mm (standard)
---
1. 31 0mm .1
2.25m
flood plain width
0 Standard rod pattern 0 Removed for some tests

Fig A2.4 Pattern of rough ness used on flood plains

1 .25

1 .20

1 .1 5

c
() 1 .10
�

1 .05

1 .00

0.95
25 50 75 1 00 X 1 03
Re, channel Reynolds number

JBW/63/1 0-91130

Fig A2.5 Drag coefficient as function of channel Reynolds Number

 1 .25 ,------.

x -
x - x -

T
""*"--

Biockage, 13
0 +
� 1 .10
+
+
0 Data
+ Formula
1 .00

0.95
0 2 3 4 5 6
z. = flow depth/rod dia

Fig A2.6 Drag coefficient and blockage coefficient as functions of ratio

of flow depth to rod diameter, Z*

10
8
6
4
Rat plate B
2

1
0.8
o0 0.6
0.4
1 :3 Elliptical cylinder �
(Zahm, Smith & Louden)�
0.2

0.1
0.08 1------�.,rrr=,r..;..,.r-:;;..::_;_--\,.
[Link] r-------�--�

O.M L------L--�--��-
1 0° 1 01 1 02 1 03 1 04 1 05 1 06
R
JBW/6411 0-91130

Fig A2.7 Coefficients of drag for two dimensional forms (Rouse, ed., 1 950)
APPENDIX 3

CHANNEL COHERENCE

1. The influence on the discharge of the interaction between main channe l

and flood plain flows depends on how comparab le the hydraulic conditions in
these zones are : if velocities and depths are very similar , then we can
expect interaction e f fects to be smal l ; i f they are very dissimi lar , then
maj or e f fects are to be expected . The degree to which the dif ferent zones
exhibit flow s imilarity wil l be re ferred to as their " coherence" : the
greater their coherence the more l ikely is the hydraulics of the section to
approach s impl e channel ( neg l ig ib l e interaction) conditions .

2. Channel conveyance is a use ful parameter in developing the concept o f

coherence . Conveyance , K} V I , was de fined b y Ven T e Chow ( 19 5 9 ) as :

Ky = Q /{S • • • A3 . 1

but it is pre ferab le to rede fine i t to be consis tent with dimens ional
analysis , as

IS = Q/{ ( 8gS ) A{ (A/ fP) • • • A3 . 2

Thus the conveyance o f a s imp l e channe l can be represented by the cross­

section area , wetted perimeter and frict ion factor . For a compound
section , the theoretical conveyance be fore a llowing for any interaction
effects is given by the sum o f the conveyances o f the main channel and
flood p l ains :

for the s ituation of two symme trical flood p lains . I f the Manning equation
is appropr iate , then this becomes

• • • A3 . 4
3. This l eads to a parameter for the coherence o f the channel section ,
namely the ratio o f the theoretical conveyance calculated by treating it as
a s ingl e unit to that calculated by summing the conveyances of the separate
zone s . I deal ly , the section c oherence would be defined as :

i=n i=n i=n

L Ai{ [ L Ail L Ail fo L Pi ]
i= l i=l i=l
A3 . 5
i =n
• • •

L [ A . { (A . / f . P . ) )
1 1 1 1
i=l

Note that f . represents the friction factors for the separate zones ,
1
calculated from the appropriate zonal values o f Manning ' s n , Re or relative
roughness according to the resistance function being use d . f , on the other
0
hand , is a g lobal value ca lculated on the bas is of the summe d section
parameters .

4. i s not
There i s a problem with the above definition : in general f
0
known , because in engineering practice it would depend on some whole-section
compendium value of Manning ' s n . It is calculable i f the flood p lain and
main channel fo l l ow identi cal r e s istance functions e . g . both smooth , or
having the same Manning ' s n value . So equation 3 . 5 could be used as the
de f inition o f COH for the smooth laboratory c ase - but could not cover the
rough flood p l ain condition or the general case when different r oughnesses
o r resistance functions apply to dif fe rent zones . A more useful definition
for design purposes is obtained by replac ing f by the perimeter weighted
0
equivalent deduced from the separate ( and calculab l e ) values for the main
channe l and f lood p lains .

i =n i =n i=n
L A.{[ L A. / L(f.P. ) ]
1 1 1 1
i= l i=l i= l
A3 . 6
i =n
. . •

L [A . { (A . / f . P . )
1 1 1 1
i= l

5. This parameter , COH 3 , varies with flow depth in a given channe l , o f

course , and three cases a r e i l lustrated in fig A3 . 1 : the Wal l ingford channel
i l lustrated in F igure 2 . 3 ; the same but with flood p lains r educed to 0 . 25m
wide ; and the Montford Br idge natural river section shown in Figure 2 . 2 .
For the smooth Wa l l ing ford channel , the appropriate friction factors were
us ing ( varying with depth because Re var ies with depth) and for the Mont ford
Bridge sect ion a constant value of Manning ' s n was appl ied for this
i l lustrat ion , with depths re lated to the lower edge of the floo d plains .
The artificial channel with hor izontal f lood p lain and flood plain/main
channel bed width ratio 1 . 5 ( de f ined here as the ratio o f width of each
flood plain to bed width of main channel ) has a very low COH value , below
0 . 3 , when the f lood plains are first inundated, increasing to 0 . 94 when the
flood plain flow depth equal s the depth of main channe l . With narrow flood
plains , width ratio 0 . 167 , COH is less s ens itive to depth and c loser to
unity , lying between about 0 . 5 and 0 . 94 . The natural river sect ion has wide
flood plains with some cross fal l ( note that Figure 2 . 2 has considerab le
vertical exaggeration) with minimum COH value ( 0 . 5 2 ) , not j us t above bank
ful l but when the fu ll width o f flood plain is inundated . Above this the
trend is very s imilar to the wide laboratory channel , whi lst below the trend
is towards unity becaus e its s loping flood plains avoid the discontinuity in
COH at bank ful l . ( For these calculations the main channel zones 3 , 4 and 5
o f Figure 2 . 2 were taken together , as were the remaining f lood plain areas )

6. For a conventional compound cross - section geometry , the coherence o f

the section may be expre s sed in t erms of the geometric ratios : let A* =

N A /A ; p * N P / P ; H* = ( H-h) /H ; and f * f / fc • where N is the number

F F C F F C
=
=
F F
o f flood plains . Then

(1 + A* ) { [ ( 1 + A* ) / ( 1 + f*P* ) ]
• • • A3 . 7
1 + A * ( A* / f * P * )

In this form i t is obvious that as A* becomes large ( deep flow on flood

plain ) then COH approaches unity , for equal roughne ss of main channe l and
3
flood plain (when f * approaches unity as the depth increase s ) . Also when
A* is very sma l l ( fl ood plains j ust inundated) COH approaches 1/{ ( 1 +
3
f * P * ) . As A * and P * depend on H * , then for a given geometry COH also
3
depends on H * .

7. I f the Manning equation applies , and perimeter weight ing o f the

friction factor is applied , then the coherence equation becomes :
 + p 4/3
*
COH • • • A3 . 8
p 2/3
2 *

8. Whether the most general form o f these definitions o f channel

coherence (equation A3 . 7 ) provides a use ful co-ordinating parameter in the
analysi s o f the experimental results remains to be s een . Its potential
bene fit is that it brings together in one parameter mos t of the factors
expected to influence the hydraulics of compound channe ls , and so might take
the place of relative depth as an indicator of how l ike a single channel the
performance might prove . An expected corollary is that the c loser to unity
COH approaches , the more likely it is that the channel can be treated as a
s ingl e unit , using the overall geometry. As f * i s included in the general
definit ion of COH ( s ee e qu A3 . 7 ) , there is some prospect too that
3
diss imilar roughnesses wil l automatica l ly be covered .
APPENDIX 4 : TURBULENCE METHOD , SOLUTION FOR GENERAL CROSS-SECTION SHAPES

The fo llowing theory is given by Shiono and Knight ( 199 1 ) . whose assistance
is gratefully acknowledged . Some small corrections to equations 1 1 in the

pub l i shed vers ion have been provided by the Authors . The method uses the

depth averaged momentum equations and is general in the sense that it can be
app l ied to any cross-s ection which can be des cribed by a series of zones
with l inear cross-fal l . The following text is a direct quotation from the
199 1 paper :

This paper describes an improved analytical solution, developed from the earlier work of Shlono &
K n i g ht ( 1 988), which now includes the effects of seco nd ary flow. Data from the Science a nd
Engi neering Research Council Flood Channel Facility (SERC-FCF) are used to quanHfy the apparent
shear stresses across a two stage channel arising from turbulence and secondary · flow effects. These
apparent shear stresses are then depth averaged to give dimensionless depth averaged eddy viscosity
values. The a nalytical solution is thus capable of reproducing the lateral distributions of depth mean
velocity and boundary. shear stress in compound or two stage channels. lt has been applied to several
natural river channels In the Severn-Trent catchment in order to extend the stage discharge relationship
for overbank flow. See Knig ht, S h lono & Plrt (1 989) and Knight, Samuels & Shlono ( 1 990).
A typical symmetric two stage channel i n which there is no crossfall in regions 1 & 3 is shown in Flg.2.
For a sufficiently wide river channel ( region 1) and flood plain ( region 3), the depth averaged velocity, Ud,
and boundary shear stress , 'tb, will attain co nstant but different values in the two regions, thus creating a
shear layer in the vicinity of region 2. Due to the re-entrant and channel corners in this region the flow is
also strongly affected by secondary flows.

2. ANALYTICAL SOLUTION

In order to predict the lateral variation of depth mean velocity and boundary s hear stress in
open channel flow , the depth mean momentum equation has to be solved for steady u niform turbulent
flow in the streamwise direction . The equation for the longitudinal strearTIIN i se component of mo mentum
on a fluid element may be combined with the continuity equation to give

p
[ auv
ay
+ a uw
az
] . ... (1 )

where x, y, z are st reamwise. lateral and normal directions respectively

O,V,W are temporal mean velocity components corresponding to x ,y,z
u ,v,w are turbu lent perturbations of velocity with respect to the mean
p is the density of water
g is the gravitational acceleration
S0 is the bed slope gradient ( S0 = sine)

The depth mean averaged momentum equation can be obtained by integrating equat ion ( 1 ) over
the water depth, H. Provided W(H) W(O) 0, then S h l o no and Knight (1 988) show that equ ation
= =

( 1 ) becomes

l
I

p gHS Tb/ l +
52
-
0
.... (2)
where is the bed shear stress
s is the side slope (1 :s. vertical:horizontal)

=
l l H --
H
p UVdz and 'y
x = Hl lH
0 0

Analytical solutions have been obtained to equation (2) based on the eddy viscosity approach and by
neglecting the secondary flow contribution i.e. (o(H pUV)cJioy = 0). The eddy viscosity approach has been
adopted because of its common usage by numerical modellers. In t hi s model the depth averaged
transverse shear stress, tyx. is expressed in terms of the lateral gradient of depth mean velocity

T P£
yx yx
. . •. (3)

Since the eddy viscosity has d imensions of m2 s-1 , it is often related to the local shear velocity, U.,and
depth, H, by the dimensionless eddy viscosity coefficient, A, defined by

••. (4)

However as equation (2) shows, the local shear velocity, u . (= v(-tblp)) is affected by the free shear layer
tu rbulence and the secondary flows. I n regions of high lateral shear it might be argued that the u., i n
equation (4) should be replaced by the primary or shear velocity difference between the two regions.
HQwever in the i nterests of simplicity and because of its common usage by hydraulic modellers the form of
equation (4) is retained with A being regarded as a 'catch all' parameter to describe various 3-D effects.
In order to express equation (2) i n terms of one variable only (Ud or 'tb) . the Darcy-Weisbach frictio n , f
( =8'tbf(pUd2)) is used to link U -. and Ud. giving

.••. ( 5)

The depth averaged eddy viscosity in equatio n (4) may t h e n be e x pressed in t h e form

. . . (6)

Su bstituting equat ions (3) & (6) into equation (2) gives

. . . . (7)

In an earlier paper. S h l o n o & Kn ig ht (1 988) assumed that o(Hp UV)dlay 0 and obt a i ned
=
analytical solutions to eq uat i o n (7) for channels of various shape. The experi mental results which are
described in a later section of this paper suggest that for the particular cases considered the shear stress .
due to secondary flow. (pUV)d. decreases approximately linearly either s ide of a maximum value which ·
occu rs at the edge of the flood plain and the main channel. Although this is a first order approximation to
the data, as Flg.7 will later show, it does have the merit that it then a llows equation {7) to be solved
analytically. Further data from a wider range of channel g eometri e s are clearly needed before this
assumption may be generally accepted. H owever, if this is so, then the lateral gradient of the secondary
flow force per u n it length of the channel may be written as

� ( H pUV)"'
ay r or . .. (8)
me

where the subscripts me and fp refer to the main channel and flood plain respectively. The analytical
solution t o equation (7) may then be expressed for a constant depth H domain as

=
..• {9)

and for a linear s id e slope domai n as

+ + w� + 11 }� ... ( 1 0)

wh e r e y

I"T+'S2 f
-
�
s- 8 - s 2
li 8

r
11
.... (1 1)

and � = depth function o n the side slope domain (e.g. � = H - ((y-b)/s) for the main channel side slope).

Equations (9)·(1 1 ) give the lateral variation of depth mean ve locity and boundary shear
stress (via equation (5)) in a channel of any shape provided its geometry can be described by a nu mber
ot linear bou ndary elements. For a constant depth doma in , equation (9) shows that as y -7 with y > oo

0, since the f low must become two d imensional ( Ud {8gS0 H/f} 1 12) in the far field where no secondary
=

!low ex ists (P = 0) , therefore A1 = 0. For a s loping side s lope domai n , eq uati on (10) shows as s -7 oo ,

A3 must be zero i n order that a solution might exist. E quati ons (9) and ( 1 0) a l so require bou ndary
conditions of co ntinuity of H U d and o(HU d )dy across joints of domains, together with the no slip condition,
Ud = 0. a t the remote boundaries. The sub d ivision of the channel cross section into various sub areas
with either constant d e pt h domains or sloping side s lope domains will therefore req u i re suff icient
computer capacity for the matrix inversion of the coeffici ents A1 . . . An. Examples of co mplex nat u ra l
geomet ries mod e l l ed in t h i s way are given in Knig ht, S h lono & Plrt (1 989) a n d K n ig ht, Samuels
& S h i o n o ( 1 99 0 ) .
APPENDIX 5 : DATA ON CHANNEL ROUGHNE SS

The fol l owing information is extracted from Ven Te Chow ( 19 5 9 ) and retains
his classi ficat ion , with the omis s ion of closed conduit data ( c lass A) .
Re fer to Chapter 7 for additional information on channe l roughnes s . The
following table is intended to provide s ome guidance where no other
information may be avai lable , but wherever practicable the roughnes s
coe fficients should b e based on observations from the system under review ,
or from simi lar systems for proposed channe ls . The information in Table
A5 . 1 here is in terms o f Manning ' s n, but this is not to be taken as a
general recommendation for the best resistance function . In many
circumstance s , and e spec ially in lined channel s , the Colebrook-White
function might be more appropriate . S ome information on k values for use
S
in the Colebrook-White equation is g iven in a supplementary tab l e , A5 . 2 , at
the end of the Appendix .
TABLE AS . 1 . RECOMMENDED ROUGHNESS COEFFICIENTS FOR USE IN MANNING EQUATION
AS GIVEN BY VEN TE CHOW . ( "Normal" values are typically used) ,

Type o f channe l and description Min Normal Max

B Lined or buil t-up channe l s

B- 1 Metal

(a) Smooth steel surface

1 . Unpainted 0 . 01 1 0 . 0 12 0 . 014
2 . Painted 0 . 0 12 0 . 013 0 . 0 17
(b) Corrugated 0 . 02 1 0 . 025 0 . 03 0

B-2 Non-metal

(a) Cement
1 . Neat , surface 0 . 010 0 . 01 1 0 . 013
2 . Mortar 0.011 0 . 0 13 0 . 015
(b) Wood
1 . P l anted , untreated 0 . 010 0 . 0 12 0 . 0 14
2 . Planed , creosoted 0.011 0 . 012 0 . 015
3 . Unp lained 0.011 0 . 0 13 0 . 015
4 . Plank with battens 0 . 0 12 0 . 015 0 . 0 18
5 . Lined with roof ing paper 0. 010 0 . 0 14 0 . 0 17
(c) Concrete
1 . Trowel finish 0 . 01 1 0 . 0 13 0 . 0 15
2 . Float finish 0 . 0 13 0 . 0 15 0 . 0 16
3 . Finished , with gravel on
bottom 0. 015 0 . 0 17 0 . 020
4 . Unfinished 0 . 014 0 . 017 0 . 020
5 . Gunite , good section 0 . 016 0 . 0 19 0 . 023
6 . Gunite , wavy section 0 . 0 18 0 . 02 2 0 . 02 5
7 . On good excavated rock 0. 017 0 . 020
8 . On irregular excavated rock 0 . 022 0 . 027
(d) Concrete bottom float finished
with sides o f :
1 . Dressed s tone in mortar 0.015 0 . 017 0 . 02 0
2 . Random stone in mortar 0 . 0 17 0 . 020 0 . 024
3 . Cement rubble masonry ,
plas tered 0. 016 0 . 020 0 . 024
4 . Cement rubbl e masonry 0 . 02 0 0 . 025 0 . 030
5 . Dry rubb le o r r ip-rap 0 . 020 0 . 03 0 0 . 03 5
(e) Gravel bottom with s ides o f :
1 . Formed concrete 0 .017 0 . 020 0 . 02 5
2 . Random s tone in mortar 0 . 02 0 0 . 023 0 . 026
3 . Dry rubble or r ip-rap 0 . 023 0 . 033 0 . 03 6
( f) Brick
1 . Glazed 0 . 01 1 0 . 0 13 0 . 0 15
2 . In cement mortar 0 .012 0 . 015 0 . 0 18
(g) Masonry
1 . Cemented rubble 0 . 017 0 . 025 0 . 03 0
2 . Dry rubb le 0 . 023 0 . 03 2 0 . 03 5
(h) Dres sed Ashlar 0.013 0 . 0 15 0 . 017
(i) Asphalt
1 . Smooth 0 . 013 0 . 013
2 . Rough 0 . 016 0 . 0 16
(j) Vegetal lining 0 . 03 0 0 . 500
DATA ON CHANNEL ROUGHNES S ( cont ' d)

Type of channe l and des c r iption Min Normal Max

C. Excavated or dredged

(a) Earth , s traight and uni form

1 . Cl ean , recently completed 0 . 016 0 . 0 18 0 . 020
2 . Clean , after weathering 0 . 018 0 . 022 0 . 025
3 . Grave l , uni form sect ion ,
clean 0 . 02 2 0 . 025 0 . 03 0
4 . With short gras s , few weeds 0 . 022 0 . 02 7 0 . 03 3
(b) Earth , winding and s lugg ish
1 . No vegetation 0 . 023 0 . 025 0 . 03 0
2 . Gras s , some weeds 0 . 025 0 . 03 0 0 . 03 3
3 . Dense weeks or aquatic
plants in deep channe l s 0 . 03 0 0 . 03 5 0 . 040
4 . Earth bottom and rubb l e
s ides 0 . 028 0 . 03 0 0 . 03 5
5 . Stony bottom and weedy
banks 0 . 025 0 . 03 5 0 . 040
6 . Cobb l e bottom and clean
s ides 0 . 03 0 0 . 040 0 . 05 0
(c) Dragl ine-excavated or dredged
1 . No vegetation 0 . 02 5 0 . 028 0 . 03 3
2 . Light brush on banks 0 . 03 5 0 . 05 0 0 . 060
(d) Rock cuts
1 . Smooth and uni form 0 . 025 0 . 03 5 0 . 040
2 . Jagged and irregular 0 . 03 5 0 . 040 0 . 05 0
(e) Channe l s not maintained ,
weeds and brush uncut
1 . Dense weeds , high as flow
depth 0 . 05 0 0 . 080 0 . 1 20
2 . Cl ean bottom , brush on
s ides 0 . 040 0 . 05 0 0 . 080
3 . Same , highest stage of flow 0 . 045 0 . 070 0 . 1 10
4 . Dense brush , high stage 0 . 080 0 . 1 00 0 . 140

D. Natural s treams

D-1 Minor streams ( top width at flood

stage < 1 00 ft )

(a) Streams on pl ain

1 . Cl ean , stra ight , ful l stage
no r i fts or deep poo ls 0 . 025 0 . 03 0 0 . 03 3
2 . Same as above , but more stones
and weeds 0 . 03 0 0 . 035 0 . 040
3 . Cl ean , winding , some poo ls
and shoals 0 . 03 3 0 . 040 0 . 045
4 . Same as above , but some weeds
and s tones 0 . 03 5 0 . 045 0 . 05 0
5 . Same as above , lower stages
more ine ffective s l opes
and s ect ions 0 . 040 0 . 048 0 . 05 5
6 . Same as 4 , but more s tones 0 . 045 0 . 050 0 . 060
7 . S lug ish reaches , weedy ,
deep poo l s 0 . 05 0 0 . 070 0 . 080
DATA ON CHANNEL ROUGHNES S ( cont ' d)

8. Very weedy reaches , deep

poo ls , or f loodways with
heavy stand of timber and
under brush 0 . 07 5 0 . 100 0 . 150
(b) Mountain streams , no vegetation
in channel , banks usua l ly
steep , trees and brush along
banks submerged at high stages
1 . Bottom : grave l s , cobb l es ,
and few boulders 0 . 03 0 0 . 040 0 . 05 0
2 . Bottom : cobbles with large
boulders 0 . 040 0 . 05 0 0 . 07 0

D-2 Flood plains

(a) Pasture , no brush

1 . Short grass 0 . 02 5 0 . 03 0 0 . 03 5
2 . High glass 0 . 03 0 0 . 03 5 0 . 05 0
(b) Cultivated areas
1 . No crop 0 . 020 0 . 03 0 0 . 040
2 . Mature row crops 0 . 025 0 . 03 5 0 . 045
3 . Mature field crops 0 . 03 0 0 . 040 0 . 050
(c) Brush
1 . S cattered brush , heavy weeds 0 . 03 5 0 . 05 0 0 . 070
2 . Light brush and trees , in
winter 0 . 03 5 0 . 05 0 0 . 060
3 . Light brush and trees , in
summer 0 . 040 0 . 060 0 . 080
4 . Medium to dense brush , in
winter 0 . 045 0 . 07 0 0 . 1 10
5 . Medium to dense brush , in
summer 0 . 07 0 0 . 100 0 . 160
( d) Trees
1 . Dense wi l l ows , summer ,
straight 0 . 1 10 0 . 150 0 . 200
2 . C leared l and with tree
stumps , no spouts 0 . 03 0 0 . 040 0 . 05 0
3 . Same as above , but with
heavy growth of sprouts 0 . 05 0 0 . 060 0 . 080
4 . Heavy stand o f timber , a
few down trees , l ittle under­
growth , flood stage below
branches 0 . 080 0 . 100 0 . 120
5 . Same as above , but with flood
stage reaching branches 0 . 100 0 . 120 0 . 160

D-3 Maj or streams ( top width at flood

stage > 100 ft) . The n value is
less than that for minor streams of
s imi lar description , because banks
o ffer less e f fective resistance .

(a) Regular section with no boulders

or brush 0 . 025 0 . 060
(b) I rregular and rough sect ion 0 . 03 5 0 . 100
 TABLE A5 . 2 . RECOMMENDED ROUGHNESS VALUES , k IN MM , FOR LINED CHANNELS ,
S
FOR USE IN COLEBROKE-WHITE EQUATION

CONDITION

CONCRETE : Good Normal Poor

C lass 4 : Mono lithic construction agains t o iled

steel forms , with no sur face irregularities . 0 . 06 0 . 15
Class 3 : Monol ithic construction agains t s teel
forms , but less perfect sur face . 0 . 15 0.3
C lass 2 : Monol ithic construction against rough
forms ; cement gun sur face ( for very coarse texture
take k
s
= aggregate size in evidence ) 0.6 1.5
C lass 1 : Smooth trowe l led sur faces 0.3 0.6 1.5

BRICKWORK :

We l l pointed brickwork 1.5 3 6

Old brickwork in need of pointing 15 30
APPENDIX 6 : EXAMPLE OF CHANNEL GEOMETRY CONVERSION AND STAGE D I SCHARGE
COMPUTATION .

Channe l geometry conversion

1. As natural channel s , and a l s o many man-made compound channel s , do not

have the "classica l " shape o f a symme tric two-stage trapezoidal
cross-sect ion , some method of working out an e quivalent section is required ,
in order to define the various parameters that appear in the various
formulae for predicting the discharge as a funct ion o f f l ow depth . The
method was explained in Chapter 7 , s ection 7 . 1 , and exampl e s from real
rive r s were i l lustrated in Figures 5 . 12 and 5 . 13 . I t should be appreciated ,
howeve r , that the calculation o f many o f the basic geometric e l ements and
discharges does not require any conversion or approximation : the ful l
detai l o f the surveyed cross sections may be - and ide a l ly shoul d be - used
for calculating the areas , wetted perimeters and hydraulic mean depths of
the main channel and flood p lain , once the vertical divisions at the top
edge o f the channel banks have been determined . An ideal ised section i s ,

however , required to determine bank s l ope s , mean bed leve l , mean l evel o f
the bank tops (hence channel depth ) , channel top width , channel b e d width
and flood p lain width . Thes e geometric parameters of a somewhat idealised
cross-section are required to solve the equations for Region 1 in
particular .

2. The channel considered here i s based on a real river which has been
improved as part of a flood rel i e f proj ect . Thus rea l istic simulated flow
data can be associated with it based on measured data , but the river section
and f l ow data have been scaled and modified so that it becomes an anonymous
case . The upper part o f Figure A6 . 1 shows the section a s it might have been
surveyed , and cros ses have been added to mark the e ight points defining the
ideal ised section in the lower part of the figur e . The x-y co-ordinates o f
the ideal ised section a r e shown be low . The vertical divisions between main
channel and flood plains are now identified , as are bank s l ope s , mean bed
leve l , bed width etc . In what fol lows , two depths o f f low are considered in
detai l , corresponding to stages of 3 . 0m and 4 . 5m , to i l lustrate the detailed
computation procedur e . However , normal ly a computer program woul d be used
for this , and so thes e two sampl e depths are set in the context of tabular
summaries of the results of applying a program written in Basi c . This
program was deve loped for analysing l aboratory data so has some extra
s imp l ificat ions : it converts the idealised section o f figure A6 . 1 into a
comp l et e ly symmetrical section before working out the geometry .

Using within bank data to assess main channel resistance coef ficient

3. It is unnecessary to explain in detail how to work out Manning ' s n for

within bank flows for a given set of s tage discharge data : this is very
conventional hydraulic computation . S imi larly , i f another friction equation
such as the e olebrook-White formula was preferred , the calculations , though
a shade more complex , are straight forward . Tabl e A6 . 1 is computed output ,
showing :

- the geometric details

- the set o f s tage-discharge data availab l e ( z stage relative t o channe l
bed)
the analysis o f these data both in terms of Manning and eolebrook-White
(wide channel ver sion)
- check calculations for ne = 0 . 025 to examine how good a fit to the data it
is .

4. The stage-discharge data here assumes that the hydraulic gradient

matched the channel slope in all case s . This is not neces sarily so , and i f
reliable measurements o f hydraulic gradient are avai lab l e , they should b e
used , o f course . The t emperature is required i n order t o assess viscosity ,
though in practice in real rivers the viscous term in the eo lebrook-white
equation is sma l l , and in some cas es neg ligible . The analysi s o f the data
shows ne ( Man in the t ab l e ) to vary in the range 0 . 022 to 0 . 026 , with
perhaps a marginal trend to increase with depth . The results approaching
bank ful l ( z = 2 . 0m) suggest using ne = 0 . 02 5 , hence the final part of table
A6 . 1 . Here Qex/Qt shows the ratio of observed dis charge to calculated
discharge , with average 1 . 026 and standard deviation 5 . 5 6 % . Much of the
positive discrepancy comes from one result at depth 1 . 5 8m , so the conc lusion
is that a Manning ' s n value for the main channe l o f 0 . 025 is appropriate for
the computations under compound flow . The variability of around 5% , or 10%
at 95% confidence , is fairly typical of field observations . As there are
ten result s , the accuracy of determination of the roughnes s coe fficient is
about 3% ( 2 x s . d . /{no . of observations , at 95%) . I t would have been
equal ly val id to proceed using the Co lebrook-White function , with k having
s
an average value of 64mm .

Detai led calculat ions : sample for two f low depths

5. Geome try : Refer to figure A6 . 1 for co-ordinates , hence dimensions .

Areas 1 to 7 proceed from left to right .

S ide s lopes : s = 0 . 3 0/ 1 . 47 0 . 204 , S = 1 . 66 / 1 . 86 0 . 89 2

FL CL
s
FR
= 2 . 10/ 1 . 68 1 . 25 , S
CR
= 2 . 20/2 . 1 4 1 . 028
s = 0 . 727 s = 0 . 960
Fav Cav

Areas : Stage 3 . 0m : Stage 4 . 5m :

1. 0 . 42 X 0 . 204/2 0 . 018 1 . 92 X 0 . 204/2 0 . 376 m

2. ( 0 . 42 + 1 . 07 ) 1 3 . 26/2 9 . 879 ( 1 . 92 + 2 . 5 7 ) 1 3 . 26/2 29 . 769
3. ( 1 . 07 + 2 . 93 ) 1 . 66/2 3 . 3 20 ( 2 . 5 7 + 4 . 43 ) 1 . 66/2 5 . 810
4. ( 2 . 93 + 3 . 07 ) 22 . 0 3 / 2 66 . 090 ( 4 . 43 + 4 . 5 7 ) 22 . 0 3 / 2 99 . 1 3 5
5. ( 3 . 07 + 0 . 9 3 ) 2 . 20/2 4 . 400 ( 4 . 5 7 + 2 . 43 ) 2 . 20/2 7 . 700
6. ( 0 . 93 + 0 . 3 6 ) 1 3 . 05 / 2 8 . 417 ( 2 . 43 + 1 . 86) 1 3 . 0 5 / 2 27 . 99 2
7. 0 . 36 X 1 . 25/2 0 . 08 1 1 . 86 X 1 . 25/2 2 . 162

Ac 73 . 810 1 12 . 645 m 2
2 x A 1 8 . 395 60 . 299 m 2
F
A.r 9 2 . 205 1 7 2 . 9 44 m 2

Wetted perimeters : {(1 + s ) 1 . 02 1 ; { ( 1 + S ) 1 . 340

FL CL
{(l + s ) 1 . 60 1 ; { ( 1 + S ) 1 . 43 4
FR CR
Stage 3 . 0m : Stage 4 . 5m :

1. 1 . 02 1 X 0 . 42 0 . 429 1 . 02 1 X 1 . 92 1 . 960 m
2. ( 1 3 . 2 6 + 0 . 65 ) 1 3 . 27 6 1 3 . 27 6 m
3. 1 . 340 X 1 . 860 2 . 49 2 2 . 49 2 m
4. ( 22 . 03 + 0 . 1 4 ) 22 . 030 22 . 0 3 0 m
5. 1 . 434 X 2 . 14 3 . 067 3 . 067 m
6. ( 13 . o s + 0 . 5 7 ) 1 3 . 06 2 1 3 . 062 m
7. 1 . 60 1 X 0 . 36 0 . 576 1 . 60 1 X 1 . 86 2 . 978 m

P 27 . 592 27 . 5 9 2 m
c
2 X P 27 . 3 43 3 1 . 27 6 m
F
P 5 4 . 935 5 8 . 868 m
T

Hydrau l i c mean depths :

Stage 3 . 0m Stage 4 . 5m

RC
73 . 8 1 0/27 . 5 92 2 . 67 5 1 1 2 . 645/2 7 . 5 9 2 4 . 083 m

RF
=
18 . 3 9 5 / 27 . 3 43 0 . 672 60 . 299/3 1 . 27 6 1 . 928 m

Ra tios :

A* 18 . 3 9 5 / 7 3 . 8 1 0 0 . 249 60 . 299/ 1 12 . 645 0 . 535

p* 27 . 343/27 . 59 2 0 . 99 1 3 1 . 27 7 /27 . 59 2 1 . 134

=

R*
0 . 67 2/ 2 . 67 5 0 . 25 1 1 . 928/4 . 083 0 . 47 2

6. Basic resis tance cal cul a t i on : The best va lue of the Manning
coefficient for the main channel was estab l ished at 0 . 025 by uti lis ing
s tage-discharge observations at high within-bank flows . There is no direct
way o f estab l i shing the appropriate f lood p l ain value , so there is an
e l ement o f trial and error involve d . O f c ourse through experience and other
sources of information , a reasonabl e first guess may be made . The f lood
p lains here are grass berms , usual ly wel l maintained . Tabl e A5 . 1 o f
Appendix 5 suggests that the f lood p lain roughnes s might b e i n the range
0 . 02 5 to 0 . 03 5 ( Pasture , no brush , short gras s ) so the first assumption is
that nF = 0 . 030 .
s = 0 . 470/ 1 000 so s 0 . 02 168 .

Stage 3 . 0m Stage 4 . 5m

213 21 3
V 2 . 67 5 x 0 . 02 168/0 . 02 5 1 . 67 1 4 . 083 X 0 . 02 168/0 . 025 2 . 2 15 m/s
C
Q 1 . 67 1 x 7 3 . 8 10 123 . 3 4 2 . 27 5 X 1 12 . 645 249 . 5 5m3 /s
C 21 3 213
V 0 . 67 2 x 0 . 02 168/0 . 03 0 0 . 5 5 4 1 . 928 X 0 . 02 168/0 . 03 0 1 . 1 19 m/s
F
2 Q ( sum o f flood p lain f l ows ) =
F
0 . 0 5 5 4 X 18 . 395 10 . 20 1 . 1 19 X 60 . 299 67 . 5 0 m l / s

Q 133 . 5 4 3 1 7 . 05 m 3 / s
Tbasic

Friction factors : f = 8gRS/V 2 ; 8gS = 8 x 9 . 8 1 x 0 . 470/ 1000 = 0 . 03 689

Stage 3 . 0m Stage 4 . 5m

fc o . 03689 x 2 . 67 5 / 1 . 67 1 0 . 03 5 3 5 0 . 03689 X 4 . 083/2 . 2 15 0 . 03068

f 0 . 03689 X 0 . 67 2 / 0 . 5 5 4 0 . 0805 0 0 . 03689 X 1 . 928/ 1 . 1 19 0 . 05682
F
2 . 27 7 1 . 85 2

Coherence : a l l necessary parameters are now avai lab le t o calculate

coherence , COH , using eq 1 3 o f the Summa ry and Design Method .

Stage 3 . 0m Stage 4 . 5m

COH =

( 1+0 . 249 ) { [ ( 1+0 . 249 ) / ( 1+2 . 2 7 7x0 . 99 1 ) ] ( 1+ 0 . 5 35 ) { [ ( 1+0 . 5 3 5 ) / ( 1+ 1 . 85 2x1 . 13 4

1 + 0 . 249{ ( 0 . 249 / 2 . 7 7 7x0 . 99 1 ) 1 + 0 . 0 5 3 5{ ( 0 . 5 3 5 / 1 . 85 2x1 . 1 3 4
= 0 . 7 1 44 = 0 . 8506

7. In e f fect , the calculations for stages o f 3 . 0m and 4 . 5m are examp les o f

what would norma l ly b e a sequential s e t o f calculations for a ful l range o f
stages , progressing i n sufficient ly sma l l depth steps to provide a l l the
geometric information required to estab l i sh a close coverage o f the
stage-discharge function . I t is as sumed that the actual range of depths for
this case goes up to 5m , to cover a rare flood , but when calculating for
Region 2 flows some geometric information is required for greater depths , as
va lues o f coherence , COH , are needed as wi l l be exp lained later . In
consequence , the computer version o f the geometric cal culation has been
taken up to 8m depth , in steps of 0 . 25m , as given in table A6 . 2 . A
cross-check with the detailed calculation at depths , z = H of 3 . 0m and 4 . 5m
shows c lose agreement , though a s mentioned earl ier the particular program
used approximates the idealised section by a ful ly symme tric one , which
marginal ly changes wetted perimeters . This feeds through the remaining
computations to yield sma l l difference s to the values of COH . Note that
thes e coherence values are specific to the assumed roughness coefficients
for main channe l and flood plain : any change in those would require a
recomputation of friction factors etc . With this body of basic information ,
we may proceed to calculate discharges for our s ampl e s tages , for the four
Regions of flow in turn .

8 . Region 1 : Some further parameter values are required : For both sample
depths , the full flood plain width is inundated , ·which therefore provides
the value of 2B . At sha l l ower depths , 2B would be defined by the variable
water surface width , as shown in Table A6 . 2 .

2B = 5 2 . 50 - 0 . 3 0 5 2 . 2 0m
2b = 3 7 . 25 - 1 5 . 22 2 2 . 03m ; b 1 1 . 0 1 5m
2wc = 3 9 . 45 - 1 3 . 5 6 = 25 . 89m
h = ( 1 . 93 + 2 . 0 7 ) / 2 - ( 0 . 07 + ( -0 . 07 ) ) / 2 = 2 . 00m
av
Aspect ratio = 22 . 0 3 / 2 . 00 = 1 1 . 0 1 5

As this is below 20 ( see para 1 0 . 1 . 7 ) . ARF = aspect ratio/ 10 = 1 . 10 .

Region 1 flows are calcul ated from the equations 2 to 9 of the Summary and
Design Method . As s e < 1 , equ . 7 applies :

Stage 3 . 0m : H * = 0 . 3 3 3 3 Stage 4 . 5m : H * = 0 . 5 5 5 6

E q 7 ; G = 10 . 42 + 1 0 . 42 +
0 . 1 7x0 . 960x1 . 85 2 + 0 . 3 4 ( 1-0 . 9 60) 0 . 1 7x0 . 960x2 . 27 7 + 0 . 3 4 ( 1-0 . 9 60)
= 10 . 42 + 0 . 3 7 16 + 0 . 0136 = 1 0 . 805 = 1 0 . 42 + 0 . 3 0 2 2 + 0 . 0 1 36 = 1 0 . 736
Eq 2 ;

Q* 2F = - 1 . 0x0 . 3 3 3 3 / 2 . 2 7 7 = -0 . 1 46 - l . OOx0 . 5 5 5 6/ 1 . 852 = -0 . 300

Eq 3 ;

Q * ZC = 1 . 240 + 0 . 395x5 2 . 20/25 . 89 - 1 . 2 40 + 0 . 395x5 2 . 20/25 . 89

+ 10 . 805x0 . 3 3 3 3 + 10 . 7 3 6x0 . 5 5 5 6

= - 1 . 240 + 0 . 7964 + 3 . 60 1 7 3 . 158 - 1 . 240 + 0 . 7964 + 5 . 9649 5 . 521

Eq 8 ; DI SDEF =

( 3 . 1 5 8- 2x0 . 146 ) ( 1 . 67 1-0 . 5 5 4 ) x ( 5 . 5 2 1-2x0 . 3 00) ( 2 . 2 15 - 1 . 1 19 ) x

3 . 00 x 2 . 00 x1 . 1 0 = 2 1 . 13 m3/s 4 . 5 0 x 2 . 00 x 1 . 1 0 = 53 . 3 9 m 3 / s
Eq 9 ;
Q 1 3 3 . 5 4 - 2 1 . 13 1 12 . 41 m3/s 3 17 . 05 - 5 3 . 39 263 . 66 m3/s
R1

9. Region 2 : This depends on coherence c a l culated with a shift in H* . As

se < 1 , eq 12 app l ies :

shift = -0 . 0 1 + 0 . 05 X 2 + 0 . 06 S e = 0 . 09 + 0 . 06 X 0 . 960 0 . 1476

Stage 3 . 0m ; H * 0 . 3333 Stage 4 . 5m ; H * = 0 . 5556

H* + shift = 0 . 4809 0 . 7032

From the definition of H* , H * = ( H-h) /H and so H

Shi fted value o f H =

2 . 00/ ( 1 - 0 . 4809 ) = 3 . 853 2 . 00/ ( 1-0 . 7 0 3 2 ) 6 . 739

( Extending the depth beyond the real sect ion raises some conceptual probl ems
but as it results from empirical analys i s thes e need not cause concern . In
practice , the flood p lain back s lopes should be extended upwards as
necessary) .

The detail for calculating e oH at thes e values o f H* in essence repeats the

calculations in paragraphs 1 to 6 above , but with the " shifted" hypothetical
depths of flow . I n practice a computer program would be used , a s mentioned
earl ier l eading to Tabl e A6 . 2 . The easy opt ion o f interpolating between the
values in this table wi l l be taken : and this a l s o explains why that table
was continued beyond the f l ow depth o f 5m which the stage/dis charge function
is to cover .

Stage 3 . 0m Stage 4 . 5m

Eq 10 : D I SADF
= eoH for shifted H* 0 . 8 19 0 . 894
Hence Q
R2
0 . 8 19 X 1 3 3 . 5 4 109 . 3 7m 3 / s 0 . 89 4 X 3 1 7 . 05 283 . 44m 3 / s
10. Region 3 : This depends on the value o f COH (without H* shi ft )

Stage 3 . 0m Stage 4 . 5m

Eq 16 : D I SADF
= 1 . 567 - 0 . 667 X 0 . 7 144 1 . 090 1 . 5 6 7 - 0 . 66 7 X 0 . 8506 0 . 9996
Q
R3
= 1 . 09 0 x 133 . 54 = 1 45 . 62 m 3 / s 0 . 99 9 6 x 3 17 . 05 = 3 1 6 . 94 m 3 /s

11. Region 4 :

Eq 18 : D ISADF = COH
Q = 0 . 7 144 X 133 . 54 9 5 . 40 m3 /s 0 . 8506 x 3 17 . 05 269 . 68 m 3 / s
R4

12 . Logic for s e l e ct ion o f region of f l ow : Equations 2 0 , and if necessary

in turn 2 1 and 22 , are app lied to determine whi ch region the flow is in :

Stage 3 . 0m Stage 4 . 5m

From eq 20 ,
Q = 1 12 . 41 , Q = 1 09 . 3 7 , Q = = 283 . 44 ,
2 6 3 . 66 , Q
R1 R2 R1 R2
Q Q , hence : Q < Q ' so region 1 is e l iminated
R1 > R2 R1 R2

REGION 1 : Q 1 1 2 . 4 1 m3/s
PRED

From eq 2 1 : Q = 3 16 . 94 , hence
R3
R 1 < R2 AND R2 < R3
Q Q Q Q so

REGION 2 : Q = 3 1 6 . 94 m 3 / s
PRED

( E q 2 2 only becomes r e levant if the test

of eq 2 1 fai l s ) .

Continuation o f s tage/dis charge assessment

13 . The above as sumed that a Manning ' s n value o f 0 . 03 0 was appropriate for
the flood p lains . However , the hydrauli c engineer shoul d have access to
some above- bank data when carrying out this proj ect , so shoul d compare the
results obtained with this value of n with the avail able observations . If
F
no computer program were availab l e , he/she would have t o go through the
above procedure for each of the observed flow depths , and then compare the
predicted discharges wi th the measured va lues . This is i l lustrated in Table
A6 . 3 , the nine assumed observations being l i s ted at the top of the table .
These go up to a depth o f about 1 . 5m on the flood pl ains , though the
s tage/discharge function requires extending to some 3m depth on the flood
plains .

14. With the bene fit o f a computer program , i t is only a matter of a few
minutes work at the PC to test a range o f assumptions about flood plain
roughness . Table A6 . 3 looks at n
0 . 03 0 , 0 . 025 and 0 . 0275 in turn , the
==
F
co lumn headed Qex/Qpr being the ratio of observed discharge to predicted
flow . The comparison is summarised in terms of the average and standard
deviation of these ratio s , so indicating the goodness of fit . The first
assumption of n = 0 . 03 0 under-predicts by about 3 . 3 percent on average ,
F
with a variabi l ity o f about 5% . This l ed to trying n = 0 . 02 5 , which
F
over-predic t s by about hal f a percent , but reduces the var iab i l ity . S o the
third attempt was with n = 0 . 0275 , giving agreement to within 1 . 6% on
F
average , with some 3 . 6% variabi lity . This may we l l be the pre ferred
assumption with this data set : at high s tages n = 0 . 02 7 5 tends to be a
F
conservative assumption . Note that a Q* l imit of 0 . 5 was applied and
zc
affected stages be low 2 . 2m only.

15 . Having decided upon the best roughnes s coefficients in this way , the
stage/ discharge function over the full range required may be calculated .
Obviously this would be considered tedious i f all calculations were manual ,
but with the bene fit of suitab l e PC s o ftware , is quickly accomplished .
Tab le A6 . 4 provides the extended s tage/dis charge functions for the three
alternat ive flood plain roughnes s e s considered , the final one being the
preferred prediction . Note the transition from Region 1 to Region 2 at a
depth o f 3 . 2 5m , and the trend o f the dis charge adj ustment factor , DISADF .
This drops to about 0 . 85 at the l imit of region 1 , but rises through Region
2 to 0 . 9 3 . I t i s also interesting t o note that with equal roughness Oh

flood p lain as in main channel , the s e cond case , the f l ow reaches Region 3 .
with D ISADF approx . 0 . 9 5 at maximum depth . This i l lustrates how the
progress through the regions depends on the ratio of the roughnesses : high
flood plain roughnes s wi l l delay that progression , and perhaps result in
only Region 1 app lying .
Ta ble A6.1 Use of with i n-bank field observations to asses s rough ness of
m a i n channe l : chan nel geometry; stage/discharge
observations; analysis of indi vidual observati ons to determ i ne
Man n i ng's n and k5 i n Colebrook-Wh ite equatio n (wide channel
version); tabu lar assessment of good ness of fit of Man n i ng
equation with selected ne val ue

geono g bed w FP w No FP chdepth ac , H/V hfp sf ,H/V sfp, VJH S/1000 Aspect
99. 000 9 . 810 22. 030 13. 155 2 .000 2 . 000 0 . 965 2 . 610 0. 727 0 . 046 0.470 1 1 . 015
Experimental data: sa�.cbs Test runbered 1
z exp Q exp S/1000 Tet�p
1 . 060 2 1 . 780 0.470 15. 000
1 . 200 25 . 560 0.470 15.000
1 . 280 29. 690 0 . 470 15.000
1 . 360 33. 180 0.470 15. 000
1 . 580 47.050 0.470 15.000
1 . 720 45 .470 0 . 470 15. 000
1 . 770 48. 900 0.470 1 5 .000
1 . 780 48.170 0.470 15.000
1 . 900 58 . 610 0.470 15. 000
1 . 920 54.290 0.470 15. 000

Pllalysis of experiments as single dlannel

z z/h Q A p R V ReE6 FF V* V/V* Man Kse
1 . 060 0 . 530 2 1 . 780 24.436 24. 976 0.978 0 . 891 3 . 060 0 . 045 0 . 067 13. 271 0 . 024 54 . 205
1 . 200 0 . 600 25.560 27.826 25. 365 1 . 097 0. 919 3. 536 0 . 048 0. 071 12. 916 0 . 025 70.231
1 . 200 0 . 640 29. 690 29. 779 25.588 1 . 164 0 . 997 4. 071 0 . 043 0 . 073 13. 610 0 . 024 56. 159
1 . 360 0 . 680 33 . 1 80 31 . 746 25. 810 1 . 230 1 . 045 4 . 511 0 . 042 0. 075 13.87 9 0 . 024 53.199
1 . 580 0 . 790 47 . 050 37. 216 26.421 1 . 409 1 . 264 6 . 248 0. 033 0. 081 15. 687 0 . 022 29. 162
1 . 720 0 . 860 45.470 40. 746 26. 811 1 . 520 1 . 116 5 . 951 0 . 045 0 . 084 13. 331 0. 026 82. 1 98
1 . 770 o. 885 48. 900 42.016 26. 949 1 . 559 1 . 164 6. 367 0 . 042 0 . 085 1 3 . 727 0 . 025 7 1 . 764
1 . 780 0. 890 48 . 170 42.271 26. 977 1 . 567 1 . 140 6. 265 0 . 045 0. 085 13.407 0 . 026 82 .164
1 . 900 0 . 950 58. 610 45.341 27. 311 1 .660 1 . 293 7 .530 0 . 037 0 . 087 14. 775 0 . 024 49.869
1 . 920 0. 960 54.290 45. 855 27. 366 1 . 676 1 . 184 6. 961 0 . 044 0. 088 13.470 0 . 026 85 . 642

MAIN CHANNEL EQUATION

Manning calculation
Main charnel Man = 0 . 025
Bank full discharge at 15 degC 60. 0457605
fLOOO PLAIN EQUATION
Manning calcula tion
flood plain Mannings n = 0 . 03
z Qex Vm On Vfp Qfp Qt Fn£2 ffp£2 ft£2 Qex/Qt
1 . 060 21. 780 o . 855 20. 884 o. ooo [Link] 20. 884 4 . 941 o. ooo 4. 941 1 . 043
1 . 200 25.560 0 . 922 25. 666 o .ooo 0 . 000 25.666 4 . 756 0 .000 4 . 756 0 . 996
1 . 280 29. 690 0. 959 28. 573 0. 000 0 . 000 28. 573 4.663 0. 000 4.663 1. 039
1 . 360 33. 180 0 . 995 31. 603 o . ooo 0 . 000 31 . 603 4 . 578 0. 000 4 . 578 1 . 050
1 . 580 47.050 1 . 090 40.554 0 . 000 0 . 000 40.554 4 . 376 0.000 4 . 376 1 . 160
1 . 720 45.470 1 . 146 46. 708 0. 000 0 . 000 46.708 4 . 266 0 .000 4. 266 0. 974
1 . 770 48. 900 1 . 166 48.990 0. 000 o . ooo 48.990 4 . 230 0 .000 4. 230 0.998
1 . 780 48. 170 1 . 170 49.451 o . ooo 0 . 000 49.451 4. 223 0 . 000 4 . 223 0. 974
1 . 900 58. 610 1 . 216 55.127 0 . 000 0 . 000 55. 127 4 .142 0. 000 4 . 142 1 . 063
l. 920 54.290 1 . 223 56. 097 o . ooo o . ooo 56.097 4 . 1 30 0 .000 4.130 0 . 968
Average Qrstio = 1 . 02647054 Standard devi ation = 5 . 56235£-02

JBW/66110-01130
Table A6.2 Geometric calculations: x-y coordinates; summary of
idealised geometry (assumed symmetric) ; geometric
parameters and channel coherence

a:tll'ruD TRAPEZOlDAl CHANI£LS: fiElD DATA, versioo TRAPEZ21 , jlne 91

X coord Y coord

[Link] 4.050
0. 300 2. 580
n . )(.O 1 . 9JO
1!>.220 U.U70
37.250 -D. 070
39.450 2. 070
52. 500 2.640
54. 600 4. 320
geono g bed w f1' w No FP et-depth 9C,H/V hfp af ,H/V afp, V,M S/1000 Aspect
99. 000 9. 810 22 . 030 13.155 2.000 2.000 0.965 2.610 o. 727 0.0116 0.470 11.015
MAIN OiAI'H:L E�ATION
Manning calculation
1-Bin chsmel Han = 0.025
Bari< full discharge at 15 degC 60.0457605
fLOOD PLAlN �ATI!J-1
Manning calculation
flood plain t&nrdngs n = 0.03
Geometry of coq:>OUnd c:hamel
z H* fvn Pm Rm Afp Pfp Rfp At Pt Rt P" A* f* 2 Beff Oli3
0 . 250 -7. 000 5.568 22.725 0.245 [Link] 0.000 0.000 5.568 22.725 0.245 [Link] [Link] [Link] 25.890
0.500 -3.000 11.256 23.420 0.481 0.000 0.000 0.000 11.256 23.420 0.481 0.000 0.000 [Link] 25.890
0. 750 -1.667 17.065 24.115 0. 708 0.000 0.000 [Link] 17.065 24.115 o. 708 [Link] [Link] 0 .000 25.890
1. 000 -1.000 22.995 24.809 0.927 [Link] 0.000 0.000 22.995 24.809 0.927 [Link] [Link] 0.000 25.890
1 . 250 -0. 600 29.045 25.504 1 .139 0.000 [Link] 0.000 29.045 25.504 1 .139 [Link] 0.000 0.000 25.890
1 . 500 -o.JJ3 35.216 26.199 1 . 344 [Link] [Link] 0 .000 35.216 26.199 1 . 344 0 .000 [Link] 0.000 25.890
1. 750 -o.143 41.508 26.894 1 . 543 0.000 0.000 0.000 41.508 26.894 1 . 543 0. 000 [Link] 0.000 25.890
2. 000 [Link] 47.920 27.589 1. 737 [Link] [Link] [Link] 47.920 27.589 1. 737 0.000 0.000 0.000 25.890
2.250 0.111 54.393 27.589 1 . 972 0.674 5. 397 0.125 55. 740 38.383 1.452 0.391 0.025 3 . 613 36.673 0. 666
2.500 0 .200 60.865 27.589 2.206 2.696 10.794 0.250 66.256 49.177 1.347 o. 783 0.089 2.977 47.456 0.612
2. 750 0. 273 67.338 27.589 2.441 5.861 13.342 0.439 79.060 54.273 1.457 0.967 0.174 2.550 52.200 0.653
3. ooo o.:m n. 81o 27.589 2.675 9. 198 13.651 o.674 92.206 54. 891 1.680 0;990 0.249 2.280 52.200 o. 715
3 . 250 0.385 80.283 27. 589 2 . 910 12.580 13.960 0.901 105.443 55.510 1 .900 1 .012 o.:m 2.128 52.200 o. 757
3. 500 0 . 429 86.755 27.589 3 . 145 16.000 14.269 1.122 118.771 56. 128 2.116 1.034 0.369 2.030 52.200 0. 788
3. 750 0.467 93.228 27.589 3. 379 19.481 14, 579 1.336 132.190 56.746 2.330 1 . 057 0.418 1 . 962 52.200 0 . 811
4 .000 0.500 99.700 27.589 3 . 61 4 23.(100 14.888 1 . 545 145.700 57.364 2.540 1 . 079 0.461 1.912 52.200 0 . 829
4.250 0 . 529 106.173 27.589 3.848 26.564 15. 197 1. 748 159.301 57.982 2. 747 1 . 102 0. 500 1.873 52.200 0 . 842
4 . 500 0 . 556 112.645 27.589 4.083 30. 174 15.506 1.946 172.993 58.600 2.952 1. 124 0.536 1:844 52.200 0 . 853
4 . 750 0 . 579 119.118 27.589 4 . 318 33.829 15.815 2. 139 186.775 59.219 3.154 1 . 146 0.568 1.820 52. 200 0.862
5 . 000 0.600 125. 590 27.589 4. 552 37. 529 16. 124 2.328 200 . 648 59.837 3. 353 1. 169 0 . 598 1 . 801 52.200 0 . 869
5.250 0. 619 132.063 27.589 4. 787 41.275 16. 433 2. 512 214.613 60.455 3.550 1 . 191 0.625 1. 785 52.200 0. 875
5. 500 0. 636 138. 535 27.589 5 . 021 45. 066 16.742 2.692 228.668 61.073 3. 744 1. 214 0.651 1. 773 52. 200 o. 880
5. 750 0.652 145.008 27.589 5.256 48.903 17.051 2.868 242.814 61. 691 3. 936 1.236 0. 674 1. 762 52.200 0 . 884
6 . 00J o . 667 151.480 27.589 5.491 52.785 17.360 3.041 257.051 62.:no 4.125 1.259 0.697 1. 754 52. 200 0. 887
6 . 250 0 . 680 157.953 27.589 5.725 56.713 17.669 3.210 271.378 62.928 4 . 313 1 . 281 0.718 1 . 746 52.200 0. 890
6. 500 0 . 692 164.425 27.589 5 . 960 60.686 17.979 3 . 375 285.797 63.546 4,497 1 . 303 0.738 1 . 740 52. 200 0. 892
6. 750 0. 704 170.898 27.589 6 . 194 64.705 18.288 3. 538 300. 307 64.164 4.680 1 . }26 o. 757 1. 736 52.200 0. 894
7.000 0 . 714 177. 370 27.589 6.429 68.769 18.597 3. 698 314.907 64.782 4.861 1.348 0. 775 1 . 732 52. 200 0.895
7.250 o. 724 183. 84 3 27. 589 6.664 72. 878 18.906 3.855 329. 598 65.400 5.040 1 . 371 o. 793 1. 728 52. 200 0. 897
7. 500 o . n3 190.315 27.589 6. 898 77.o:n 19.215 4.009 344 . 381 66.019 5. 216 1. 393 o.a1o 1.726 52. 200 o. 89B
7 . 750 0. 742 196. 788 27.589 7.133 81.233 19. 524 4 . 161 359. 254 66. 637 5. 391 1 . 415 0.826 1.723 52.200 0.899
8.00J o. 750 203. 260 27. 589 7. 367 85.479 19.833 4. 310 374. 218 67.255 5 . 564 1 . 438 0.841 1. 722 52.200 0 . 899

JBW/67/10-91130
Ta ble A6.3 Comparison of observations of stage/d ischarge with
predicted flows: l ist of observations ; calcu lations for three
assu med values of n F. (su bscript m = main channel ; fp =
flood p lai n ; Qpred is the predicted discharge)
Experirrent al data: [Link] Test ruroered 2
z exp Q exp S/1000 Tenp
2 . 050 60.750 0.470 15.000
2.060 63. 470 0. 470 15.0Xl
2 . 140 6 3 . 820 0.470 15.000
2. 400 79.490 0.470 15.000
2. 540 84.960 0.470 15.000
2. no 91.460 o.470 15.000
3. 000 ll7. 900 0.470 15.000
3 . 160 140.910 0 . 470 15. 000
3.570 183.670 0 . 470 15. 000

HAlN 01Pl>li£L E�AT!ON

Manning calculation
Main channel Man = 0. 025
Bar* full discharge at 15 degC 60. 0457605
FLOOO PLAIN EQUATION
Hannirg calculation
flood plain Mannlngs n :: 0 . 03
following uses full predictille fU"lCtions. Region 1 incorporates an aspect ratio factor for both Q*2f and Q><2C:
aspect ratio of this geometry is 1 1 . 015
Aspect ratio used in following = 1.1
� H* Qex Vm Q'n Vfp Qfp F* Qpred Qex/Qpr Regioo QtRl QtR2 QtR3 QtR4
2.050 0.024 60. 750 1.276 62. n3 0.062 0 . 002 5.975 60 . 040 1 . 012 1.000 60 . 04 0 39. 079 63. 752 51.903
2.080 0 . 038 63. 470 1. 289 64.433 0.084 0 . 006 5 .135 61. 689 1.029 1. 000 61. 669 39.733 67.317 50.477
2. 140 0 . 065 63.820 1. 315 67.805 0 . 123 0.026 4 . 305 65.049 0. 981 1.000 65.049 41.279 73.486 49.240
2.400 0. 167 79.490 1 . 426 63. 196 0 .247 0 . 426 3 .161 76. 955 1 . 033 1 . 000 76.955 58. 576 96.653 52.550
2 . 540 0.213 64. 960 1 . 486 91.998 0. 302 0 . 949 2 . 918 82. 747 1 . 027 1.000 82. 747 69.295 109.025 57.135
2.770 0.278 9 1 . 460 1. 560 107.216 0 . 430 2 . 632 2 . 521 95.999 0.953 1. 000 95.999 88. 423 126.818 74.119
3. 000 0. 333 117.900 1. 611 123. 352 0 . 555 5. 109 2.280 112.463 1 . 046 1. 000 ll2. 463 109.386 145.645 95.440
3. 160 0 . 367 140.910 1 . 733 135. 105 0.633 7 . 191 2. 115 125.173 1 . 126 1. 000 125. 113 124. 948 160.126 111. 123
3 . 570 0.440 183. 670 1 . 867 167.139 0 . 808 13.118 2.009 168. 336 1.091 2. 000 161. 937 166. 336 201.724 154.666
Average Qratio = 1.03326263 Standard deviation = 4.92696E-D2

Main channel Man = 0.025

floc:d plain Hannlngs n = 0. 025
Following uses ful l prooictille fl.l1ctions. Region 1 incorporates an aspect ratio factor for both Q *2f and Q*2C:
aspect ratio of this geometry is 1 1 . 015
Aspect ratio used in fallowing = 1.1
z H* Qe x Vm Q'n Vfp Qfp F* Qprfrl Qex/Qpr Regioo QtRl QtR2 QtR3 QtR4
2.050 0.024 60. 750 1 . 276 62. 773 0 .074 0 . 002 4 . 149 60 . 06 8 1.011 1.000 60.068 43. 730 61. 934 54.630
2 . 080 0.038 63. 470 1. 289 64.433 0 . 101 0 . 007 3. 566 61. 730 1.028 1. 000 61.730 44 . 569 65.107 53.7%
2 . 140 0 . 065 63. 620 1 . 31 5 67. 805 0. 147 0 .0}1 2 . 990 65.117 0.980 1.000 65.117 46.485 70. 759 53. 355
2.400 0. 167 79.490 1 . 428 83. 196 0.296 0.511 2 . 195 77.857 1.021 1 . 000 77. 857 65.311 92.710 58.862

2 . 540 0 .213 64.960 1 . 486 91. 998 0 . 362 1 . 138 2 . 026 84.292 1.006 1 . 000 84. 292 76.857 104. 650 64.585
2. 770 0 . 278 91.460 1. 580 107. 216 0 . 515 3. 156 1 . 751 99.140 0.923 1.000 99.140 97. 646 122. 042 63. 753
3. 000 0. 333 117.900 1. 611 123. 352 0 . 666 6.130 1 . 584 120. 598 0 . 978 2. 000 117.713 120.598 140. 805 107. 496
3.160 0.367 140.910 1 . 733 135. 105 0 . 760 8.629 1 . 510 137.707 1.023 2. 000 132 . 125 137. 707 155.384 124.989
3.570 0.440 183. 670 1 . 867 167.139 0 . 970 16.462 1. 395 185.651 0 . 989 2. 000 174. 020 185. 651 197. 659 113. 672
Average Qratio = 0 . 995695525 Standard deviation = 3, l2895E-D2

Main channel Man = 0.025

flood plain Hannings n = 0.0275
Following uses ful l prEdictive fli1Ctions. Region 1 incorporates an aspect ratio factor for both Q*2F £Yid Q*ZC :
aspect ratio of this georretry is 11. 015
Aspect ratio used in following : 1.1
z H* Qex Vm Qn Vfp Qfp f* QprOO Qex/Qpr Region QtRl QtR2 QtR3 QtR4
2.050 0.024 60 . 750 1 . 276 62. 77J 0.067 0.002 5.021 60.053 1.012 1 . 000 60.053 41. 315 62. 836 53. 276
2.080 O.OJ8 63. 470 1 .289 64.433 0 . 092 0 . 006 4 . 315 61. 708 1.029 1. 000 61. 708 42.054 66.214 52.133
2 . 140 0 . 065 6}. 620 1.315 67. 805 0. 134 0 . 028 3. 617 65.080 o. 981 1.000 65.060 43. no 72.143 51 . 267
2. 400 0. 167 79. 490 1 . 428 83. 196 0 . 269 0 . 465 2.656 77. 375 1 . 027 1. 000 n . 375 61.829 94. 752 55. 581
2 . 540 0.213 84. 960 1.486 91.998 0.329 1 . 035 2 .452 83.462 1.018 1.000 83.462 72.965 106 . 926 60.687
2. 770 0.278 91. 460 1. 560 107.216 0 .469 2 .871 2 . 118 97.445 0.939 1. 000 97. 445 92. 917 124.50} 78. 714
3. 000 0 . 333 117.900 1 . 671 123. 352 0 . 606 5. 573 1 .916 114. 877 1.026 1. 000 114.8n 114. 855 143. 255 101 . 205
3. 160 0. 367 140. 910 1. 733 135. 105 0 . 691 7.845 1 . 628 131.172 1.074 2. 000 128. 370 131. 172 157.748 117.760
3.570 0.44() 183. 670 1. 887 167. 139 0. 882 14. 965 1 .688 176.770 1.039 2. 000 167.494 176. 770 199. 562 163 . 78 7
Average Qratio l . 01602969 Standard deviation 3. 58727E-D2
JBW/68/10.91/30
= =
Table A6.4 Calculated stage d ischarge functions up to stage of 5.0m :
three assumed values of nF i n tu rn

Main chamel Man = 0 .025

flood plain Mannings n = 0.03
following uses full predictive [Link]::tlons. Region 1 incorporates en aspect ratio fsctOl' for both Q"2f tl"ld Q*2C: aspect ratio of
geonetry is 11.015
Aspect ratio factor wed in following, N?f = 1 . 1
z H* Vm Qn Vfp Qfp f" Qt � Region QtR1 QtR2 QtR3 QtR4 DISPDf
2.000 0.000 1 .253 60.046 [Link] 0.000 0.000 60.046 60.046 0.000 60.046 38.188 94.092 0.000 1. 000
2.250 0 . 111 1.)63 74. 163 0.181 0. 122 3.613 74.406 71. 578 1 .000 71. 578 47.396 83.562 49. 524 0.962
2 . 500 0 .200 1.470 89.447 0 .287 0.773 2.977 90.992 81.013 1 .000 81.013 66.187 105.449 55.677 0.890
2.750 0 . 273 1. 572 105.856 0.418 2.448 2.550 110.751 94.676 1 .000 94. 676 86.680 125.306 72. 326 0 .855
3.000 0. 333 1 . 671 123.352 0.555 5.109 2.280 133.569 112.463 1. 000 112.463 109.386 145. 645 95.440 0 . 842
3.250 0 . 385 1. 768 141.902 0.674 8.482 2 . 128 158.866 1}4.047 2.000 132. 744 134.047 168.724 120.266 0.844
3. 500 0.429 1.861 161.477 0.760 12.490 2.030 186.457 160.582 2.000 155.259 160.562 194.193 146.905 0.861
3. 750 0 .467 1.953 182.052 0.877 17.080 1 . 962 216.211 188.906 2.000 179.822 186.906 221.854 175.335 0. 874
4. 000 0 . 500 2.042 203.601 0. 966 22.212 1 . 912 248.025 218.935 2.000 206.291 216.935 251. 575 205.518 0.883
4.250 0 . 529 2.130 226.104 1.049 27.856 1 . 87} 281.816 250.589 2.000 2}4. 550 250.589 283.256 237.405 0.889
4. 500 0. 556' [Link] 249.540 1. 126 33.987 1 . 844 317.514 283.793 2.000 264.506 283.793 316.822 270.949 0.894
4 . 750 0 . 579 2.299 273.893 1 .200 40.584 1.620 355.061 318.476 2.000 296.082 318.476 352.211 J06. 103 0.897
5. 000 0.600 2.362 299 . 143 1.269 47. 631 1.801 394.406 354.572 2.000 329.211 354.572 389.370 342.825 0.899

Main chamel Man = 0 .025

flood plain Hannings n = 0.025
Following uses full predictive fli'!Ctions. Region 1 incorporates en aspect ratio factOl' for both Q*2f tl"ld Q*2C: aspect ratio of
geonetry is 11 . 015
Aspect ratio factor used in following, N?f = 1 . 1
z H* Vm Qn Vfp Q(p f* Qt Qpred Re![Link] QtRl QtR2 QtRJ QtR4 OISPDf
2.000 0. 000 1 . 253 60.046 0 .000 0.000 1 . 067 60.046 60.046 0.000 60.046 42. 524 94.092 [Link] 1 .000
2. 250 o.n1 1.)63 74.163 0.217 0 . 146 2.509 74.455 71.981 1 .000 71.981 53.270 80.216 54.655 0.967
2. 500 0 .200 1 .470 89.447 0.344 0.927 2.067 91.301 82.356 1 .000 82.356 73.506 101.187 62.792 0.902
2 . 750 0.273 1.572 105. 856 0.501 2.937 1 . 771 lll. 7JO 97.653 1 .000 97.653 95.743 120.549 81.758 0.874
3.000 0.333 1 . 671 123.352 0.666 6.130 1.584 1J5.61J 120. 598 2.000 117. 71J 120.598 140.805 107.496 0.889
3 . 250 0.385 1 . 768 141.902 0.809 10.178 1.478 162.258 147. 7}3 2.000 140.729 147.733 164.081 135. 199 0.910
}. 500 0.429 1.861 161.477 0.936 14.988 1 . 410 191.453 177.059 2.000 166.392 177.059 189.97} 164.968 0.925
}, 750 0 .467 1 . 953 182.052 1.052 20.496 1.362 223.043 208.487 2.000 194.481 208.487 218.252 196. 787 0.935
4.000 0 .500 2.042 203.601 1.159 26.655 1 . 327 256.910 241 .931 2.000 224.826 241.9}1 248.759 230.614 0.942
4.250 0 . 529 2.130 226.104 1.258 33.427 1.301 292.959 277.309 2.000 257.294 277.309 281. 378 266.399 0.947
4. 500 0.556 [Link] 249.540 1 . 352 40.784 1 . 280 331.109 314.547 2.000 291. 774 314.547 316.019 304.091 0.950
4 . 750 0 . 579 2.299 273.893 1.440 48.701 1.264 371.295 352.609 3.000 328.175 353. 571 352.609 343.644 0.950
5.000 0.600 2 . 362 299 . 143 1 . 523 57.156 1 . 251 413.459 391.086 3.000 366 . 418 394.315 391.086 385.013 0.946

Main chamel Man = 0 .025

flood plain Mannings n = 0.0275
following uses full predictive [Link]::ti ons. Region 1 incorporates en aspect ratio factor for both Q*2f tl"ld Q*2C: aspect ratio of
ge(:Mretry is 11 . 015
Aspect ratio factor wed in following, N?f = 1 . 1
z H* Vm Qn Vfp Qfp f* Qt �red Region QtR1 QtR2 QtR3 Qt R4 DISPDf
2. 000 0 .000 1 . 253 60.046 0.000 0.000 0. 741 60.046 60.046 0 .000 60.046 40.281 94.092 0.000 1. 000
2. 250 0 . 111 1 . 363 74.163 0. 197 0. 133 3.036 74. 428 71. 768 1 . 000 71. 766 50.212 81. 934 52.016 0.964
2. 500 0 . 200 1.470 69.447 o.31J 0.843 2.501 91.133 81. 635 l.[XXJ 61 . 635 69.735 103.402 59.075 0. 896
2 . 750 0 . 273 1 . 572 105. 856 0.456 2.670 2 . 143 lll.l96 96.047 l.[XXJ 96.047 91.095 123.004 76.823 0 . 664
3. 000 0. 333 1. 671 123.352 0.606 5. 573 1.916 134.498 114. 877 1. 000 114.877 114. 855 143. 255 101 . 205 ' 0 . 854
3. 250 0 . 385 1. 768 141.902 0. 735 9.253 1. 788 160.408 140.721 2. 000 136.415 140.721 166.371 127.419 0.877
3 . 500 0 . 429 1. 861 161.477 0.851 1J.625 1. 706 188.728 168. 610 2. 000 160. 378 166.610 191.977 155.562 0. 893
3. 750 0.467 1 . 953 182.052 0.956 18.633 1 . 646 219.317 198.434 2. 000 186.564 198.434 219.860 185.621 0.905
4. (XXJ 0. 500 2.042 203.601 1.054 24.232 1.606 252.064 230. 107 2.000 214.816 230. 107 249 .876 217. 553 0. 913
4. 250 0 . 529 2. 130 226. 104 1. 144 30. 389 1. 574 286.881 263.546 2.000 245.012 263. 548 281.919 251. 309 0.919
4 . 500 0. 556 2. 215 249.540 1.229 37. 077 1 . 549 323.694 298.683 2.00J 277.051 296.683 315.906 286.840 0.923
4. 750 0 . 579 2.299 273. 893 1.309 44.274 1. 529 362. 440 335. 439 2. 000 310.646 :m. 439 351. no 324. 099 0.926
5. (XXJ 0. 600 2. 382 299. 143 1.385 51.962 1 . 51} 403.067 373. 748 2. 000 346.332 373. 746 389.455 363.043 0 . 927

JBW/6911 0.91130
 0 0
(") (")

E E

0 0

1 C\1 C\1

gt. ....
o
...
IC) •
,... o

<.0
IC) M
• 0)
(") .

0 0
.... ....
E E
IC) 0
-et cw:i

O CXl
(") IC)
ci ci
0 10
00
ci -<t
0
E E X >.

JBWfl>S/10-91130

Fig A6.1 (a) : river channel as su rveyed.

(b) : idealised form of cross-section, with co-ord inates defining
its shape
APPENDIX 7 ANALYS I S O F OTHER SOURCES O F LABORATORY DATA

1. Asano T , Hashimoto H and Fuj ita K . Characteristics o f variation o f Manning ' s roughnes s
coeffic ient i n a compound c ro s s s ect ion . International a s s oc iation for Hydraulic

Research , Proc . 2 1 s t Congress , Melbourne , Vol 6 , August 1 9 8 5 , pp 3 0- 3 4 .

TABLE A7 . 1 STATI STICAL ANALYS IS OF GOODNESS OF FIT BETWEEN VARIOUS PRED I CTION ASSUMPTIONS
AND ASANO et al RESULTS

Upper Average ratio of experiment a l discharge to predict ion .

Lower S tandard deviation expressed a s percentage variation .

Series B/h B/b Case Case Case Case Case Case Case Case Case Case

1 2 2a 3 4 5 6 6a 7 8

5 10 2 . 50 0 . 99 8 0 . 9 13 0 . 95 5 0 . 99 7 0 . 9 13 0 . 9 13
1 . 77 1 . 46 1.73 1 . 77 1 . 46 1 . 46

Values o f ARF : 1 3 2 3 3 3 1 2 2 1

3 30 2 . 50 1 . 07 2 1 . 137 1 . 108 1 . 00 4 1 . 05 1 1 . 032 0 . 9 42 0 . 97 5 0 . 97 5 0 . 942

5 . 27 4 . 80 4 . 54 3 . 06 4 . 39 2.81 4 . 02 3 . 25 3 . 25 4 , 02

9 30· 3 . 33 1 . 03 1 1 . 08 2 1 . 05 8 1.011 1 . 02 9 1 . 03 0 0 . 96 8 0 . 99 1 0 .99 1 0 . 968

3 . 37 4 . 59 3 . 74 3 . 79 4 . 80 4.61 3 . 43 3 . 42 3 . 42 3 . 43

10 30 2 . 00 0 . 917 0 . 962 0 . 942 1 . 02 1 1 . 06 8 1 . 00 5 0 . 987 1 . 006 1 . 00 6 0 . 987

l. 48 3 . 81 2 .81 2.51 4 . 20 3 . 39 0 . 79 1 . 49 1 . 49 0 . 79

11 30 l. 67 1 . 02 9 1 . 07 4 1 . 05 6 1 . 03 2 1 . 06 0 1 . 03 9 1 . 000 1 . 019 1 . 0 19 1 . 000

2.31 1 . 72 o. 77 2 . 73 3 . 14 3 . 17 2 . 35 l. 87 l. 87 2 . 35

12 30 l . 42 1 .012 1 . 044 l. 03 1 1 . 00 8 1 . 004 1 . 02 2 0 . 98 0 0 . 99 7 0 . 99 7 0 . 98 0

l. 55 3 . 44 2 . 15 3 . 05 3 . 46 3 . 52 1 . 54 1 . 89 1 . 89 1 . 54

13 30 l. 25 0 . 94 5 0 . 980 0 . 983 0 . 99 2 1 . 03 2 0 . 99 4 0 . 967 0 . 9 84 0 . 9 82 0 . 967

1 . 36 1 . 97 5 . 46 l. 55 2 . 24 2 . 03 l. 24 l. 52 l. 57 l. 24

AVER 30 1 . 00 1 1 . 0 46 l. 030 1 . 011 1 . 047 1 . 020 0 . 97 4 0 . 995 0 . 99 5 0 . 974

ONLY 2 . 55 3 . 39 3 . 25 2 . 78 3.71 3 . 25 2 . 23 2 . 24 2 . 25 2 . 23
CASE 1 . Chap 3 predictors with k 0 . 1 5mm on both f lood p l a ins and in main channe l ,
s
us ing wide channel trans ition funct ion for the basic res i s tance

CASE 2 . Q* for region 1 rede fined to depend on main channel bed width rather rather
2
than depth , ie aspect ratio factor , ARF =
aspect ration/ 1 0 , k
s
0 . 1 5mm =
CASE 2a Q* for region 1 rede fined to depend on channel aspect ration , but using an
2
intermediate value o f ARF =
2 for aspect ratio 3 0 , k
s
=
0 . 1 5mm
CASE 3 . Redefined Q * Z ' ARF = aspect ratio/ 1 0 , using Manning equation with the Authors '

values for individual test series for main channel , f lood p l ai n constant at
0 . 0098
CASE 4 . Redefined Q * Z ' ARF aspect ratio/ 1 0 , but with wide-channel trans ition , k
=
s
l
values for individua test series c a l culated from within b ank tests in that

s e r i e s , and appl ied a l s o to f l ood p lain

CASE 5 . As above but some mas s aging o f channe l value s , coupled with k = 0 . 1 5mm on
s
f l oo d p lain
CASE 6 . Rever ting t o Authors ' Mannings n values , origina l definition o f Q * ie ARF 1
2
CASE 6a Authors ' Mannings n , redefined Q * but ARF at inte rmediate value o f 2 for main
2
channel aspect ratio of 3 0
CASE 7 . As 6 a but with the alternative Reg ion 3 formu l a , DI SADF = 0 . 95
CASE 8 . Autho r s ' Mannings n , ARF set at 1 , Region 3 - D I SADF = 0 . 95
2. US WES , Hydrau l i c capaci ty o f meandering channel s in s traight f loodways , Tech . Memo .
2 . 249 , Waterways Experiment s tat ion , Vicksburg , Mis s i s s ipp i , March 1 95 6 .

TABLE A7 . 2 : CALIBRATION DATA FOR THE WES EXPERIMENTAL FACI LITY

Condi t ion WES value o f C al culated value Calculated value

Manning ' s n o f Manning ' s n ; o f k • mm '

s
quoted ; l Q - 3 lQ- 3

Roughnes s case : 0 1 2 0 1 2 0 1 2

l ft bankful l 12 11 . 8 0 . 64

2ft bankful l 12 12.0 0 . 72

0 . 1 ft o n FP 12 25 35 11.3 28 . 1 43 . 4 0 . 39 27 . 9 69 . 4

0 . 2ft on FP 12 25 35 11.7 22.9 33.0 0 . 56 21. 1 62 . 2

0 . 3 ft on FP 12 25 35 12. 3 21.9 30.9 0 . 83 20 . 5 65 . 2

TABLE A7 . 3 US WATERWAYS EXPERIMENT STATION RESEARCH :
COMPARI S ION OF PREDICTION WITH MEASUREMENT FOR VARIOUS ASSUMPTIONS

Case Channel Roughne s s coe f f i c ient s Av . error S E E ( 1)

width Channel Flood p lain for % %
roughne s s 0 , 1, 2

1. Manning , ARF = 1,
o r i g inal d e fn 1 ft 0 . 012 a s Tabl e 7 . 2 6.3 9.4
of 2 ft 0. 012 a s Tabl e 7 . 2 25 . 5 31.7
Q*2

2. Manning , 1 ft 0 . 012 a s Tabl e 7 . 2 6.3 9.4

mod 2ft 0 . 0 12 a s Tabl e 7 . 1 2.4 5.2
Q*2

3. Manning , ARF aspect ratio/ 1 0

lft 0 . 012 a s Tab l e 7 . 2 6.3 9.4
Q*2c & Q*2F
mod i f i ed 2ft 0 . 012 a s Tabl e 7 . 2 2.9 5.0

4. Wide-tran s i t i on , ARF =
1,
o r i g de fn 1 ft 0 . 68mm 0 . 60 , 2 3 . 2 , 6 5 . 6mm 3.6 5.1
of 2 ft 0 . 6 8mm ditto 18 . 8 19 . 6
Q*2

5. Wide-trans i t ion , 1 ft 0 . 6 8mm ditto 3.6 5.0

mod 2 ft 0 . 68mm d i t to 4.6 5.1
Q*2c

6. Wide t rans i t ion , ARF = aspect ratio/ 1 0

1 ft 0 . 6 8mm ditto 3.6 5.0
Q*2c & Q * 2 F
modi fied 2 ft 0 . 68mm ditto 5.2 5.0

7. Wide t r ansit ion ,

ARF = 0 . 8 2 ft 0 . 68mm ditto 16 . 9 15 . 0
0.6 2 ft 0 . 68mm ditto 11.9 7.9

8. Wide t rans i t ion ,

ARF 0.4 2 ft 0 . 5 0mm 0 . 30 , 30 . 0 , 5 0 . 0mm 0.6 3.8

9. Wide trans it ion ,

ARF = 0 . 4 2 ft 0 . 5 0mm 0.30, 30. 0 , 5 0 . 0mm 0.7 3.7

Note (1 ) . S E E, s t andard error o f the e s t imate i s the r . m . s o f the variation about the
mean error value , expr e s s e d here as p e r centag e .
3. Myer s W R C . Momentum t r ans fer in a compound channe l , Journal o f Hydraul i c Research ,
Col 16 , 1978 , No 2 , 1 39 - 1 5 0
Myer s W R C . Frictiona l Res i s tance in channe l s with f l ood p l a i ns : Channe l s and channe l
cont r o l s t ructure s , 1 s t Int . Conf . Southampto n , Eng l and , 1984 , ed . K V H Smith , pub
Spr inger-Ve r l ag , 1984 , p 4 . 7 3-4 . 8 7
Myer s W R C . F l ow r e s i s tance in smooth compound channe l s , exper imental data ,
Unive r s ity o f Ul s t e r , March 1 9 8 5 .

TABLE A7 . 4 : STATISTI CAL ANALYSI S OF GOODNESS OF F I T BETWEEN VARIOUS PREDICTION ASSUMPTIONS

AND MYERS RESULTS

Note : The fo l l owing uti l ises the Reg ion 3 funct ion based on COH .

Geomet ry B /b Us ing C o l eb r ook-White , mean errors and S D ' s , % Mye r ' s e qu .

2b/h ARF =

0 . 13 0 . 20 0 . 40 0 . 60 1 . 00 0 . 60

1 4 . 68 +1.6 * +1.6 * +1.6 * +1.6 * +5 . 7 *

1 . 99 2.9 2.9 2.9 2.9 3.1

2 3 . 21 -1.0 -0 . 6 * -0 . 6 * -0 . 6 * +5 . 5 *
1 . 98 5.7 5.2 5.2 5.2 5 . 3"

3 4 . 74 +1.7 +1.8 * +1.8 * +1.8 * +1.8 * *7 . 3 *

1 . 32 4.5 4.5 4.5 4.5 4.5 4.0

Not e : * denot e s no depths s ha l l ow enough to yie l d Region a .

TABLE A7 . 5 COMPARISON OF ALTERNAT IVE FORMULA FOR REGION 3

Note : Based on C o lebrook-White -funct i on with k = 0

s

Formu l a Geometry Overa l l f i t : Reg ion 3 on l y :

number numb e r

Number Mean SD % Numbe r in Mean so %

o f tests error reg ion error

1 1 42 +0 . 16 2 . 90 5 + 2 . 28 3 . 84
2 49 -0 . 5 8 5 . 20 5 -0 . 06 3 . 98
3 34 + 1 . 84 4 . 19 4 +2 . 9 3 5 . 48
Average : +0 . 4 7 4 . 19 + 1 . 72 4 . 43

2 1 42 +0 . 19 2 . 89 9 +1 . 34 3 . 14
2 49 -0 . 64 5. 17 10 -0 . 07 3 . 16
3 34 +1 .91 4 . 38 15 + 2 . 08 3 . 86
Ave rage : +0 . 49 4. 15 + 1 . 12 3 . 39
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - �-- - - - - - � - - - -

3 1 42 -0 . 7 0 2 . 67 7 +0 . 54 2. 16
2 49 - 1 . 04 5 . 00 6 + 3 . 18 4 . 47
3 34 +0 . 5 6 4 . 32 1 * -2 . 7 3
Average : -0 . 3 9 4 . 00 + 1 . 86 3 . 32

* omi tted from average

FORMULA 1 D I SADF 1 . 5 6 7 - 0 . 6 6 7 COH 3
FORMULA 2 D I SADF 0 . 95
FORMULA 3 D I SADF 1 . 06 0 . 2 4 H*
4. Prinos P and Towns end R D . E s t imating dis charge i n compound open channel s , Canadian
Soc . for Civ i l Engineer ing , 6th Canadian Hydrotechni c a l Con ference , Ottawa , Ontario ,
June 1 9 8 3 , 1 20- 1 46 .
Pr inos P and Towns end R D . Comparison of methods for predict ing d i s charge in compound
open channe l s . Advanc es in Wat e r Resour ces , 1 9 8 4 , Vol 7 , Dec , CML Pub l ications ,
1 80- 1 8 7 . .

TABLE A7 . 6 : STAT I STI CAL ANALYS I S OF FIT BETWEEN VARIOUS PRED I CTION ASSUMPTIONS AND PRINOS
AND TOWNSEND RESULTS

CHANNEL WI DTH , mm 203 305 406 SOB

ARF Mean SD ,% Mean SD , % Mean SD , % Mean SD , %

Manning ' s
n used
for FP

0.011 0 0 . 90 3 6 . 19 0 . 9 10 2 . 07 0 . 899 5 . 95 0 . 89 4 4 . 07
*
0.2 0 . 960 6 . 49 0 . 95 4 2 . 78 0 . 93 3 6 . 88 0 . 92 1 4.51
0.4 0 . 98 5 *
4 . 19 0 . 990 *
2 . 55 0 . 969 7.71 0 . 949 <1> 5 . 10
0.6 0 . 99 7 *
2 . 58 1 . 008 *
1 . 76 0 . 99 0 6 . 95 0 . 974 <1> 5 . 21
0.8 1 . 00 2 *
2 . 30 1 . 017 *
1 . 29 1 . 00 2 <1> 5 . 88 0 . 99 1 *
4 . 44
* * *
1.0 1 . 002 2 . 30 1 . 02 4 1 . 82 1 .010 5 . 04 1 . 003 *
3 . 56

0 . 0 14 0 0 . 84 1 5 . 76 0 . 85 7 6 . 17 0 . 869 6 . 09 0 . 865 5 . 09
0.2 0 . 9 19 7 . 13 0.913 7 . 66 0 . 9 12 7 . 61 0 . 899 5 . 62
0.4 0 . 98 3 6 . 54 0 . 97 8 9 . 59 0 . 96 1 9 . 43 0 . 963 6 . 45
0.6 1 . 009 *
3 . 82 1 . 02 1 8.71 1 .012 1 1 . 06 0 . 97 6 7 . 62
0.8 1 . 02 4 *
1 . 63 1 . 042 6 . 42 1 . 043 1 0 . 46 1 .013 8 . 20
1.0 1 . 03 7 *
2 . 66 1 . 05 7 <1> 4 . 69 1 . 06 8 9 . 35 1 . 03 9 7 . 64

0 . 0 18 0 0. 772 6 . 94 0 . 822 6 . 53 0 . 844 6 . 40 0 . 874 7 . 33

0.2 0 . 86 5 9 . 49 0 . 89 0 8 . 89 0 . 89 6 8 . 47 0 . 87 4 7 . 33
0.4 0 . 97 5 1 1 . 87 0 . 97 1 1 2 . 02 0 . 957 1 1 . 03 0 . 9 18 8 . 74
0.6 1 . 02 7 8. 72 1 . 0 59 1 4 . 82 1 . 027 1 4 . 24 0 . 9 68 1 0 . 65
0.8 1 . 05 4 <l? 5 . 67 1 . 1 09 1 3 . 46 1 . 09 7 16. 72 1 . 02 4 1 3 . 12
1.0 1 . 07 4 3 . 39 1 . 137 1 1 . 24 1 . 139 16 . 17 1 . 07 6 14 . 7 1

0 . 02 2 0 0. 721 8 . 03 0 . 783 6 . 25 0 . 808 7 . 88 0 . 798 6 . 56

0.2 0 . 82 2 1 1 . 48 0 . 85 8 9 . 04 0 . 867 10. 38 0 . 84 1 7 . 95
0.4 0 . 96 1 1 6 . 88 0 . 950 12 . 9 1 0 . 934 13 . 57 0 . 890 9 . 86
0.6 1 . 05 1 15 .95 1 . 06 7 1 8 . 43 1 . 016 17 . 70 0 . 9 46 1 2 . 37
0.8 1 . 097 1 1 . 81 1 . 149 1 8 . 89 1 . 1 16 23 . 19 1 . 0 10 1 5 . 64
l.O 1 . 1 17 8 . 0 27 1 . 197 1 6 . 87 1 . 187 2 4 . 27 1 . 0 86 19 . 9 1

*
The s e resu l t s are within 5% mean error and a l so 5% var i ab i l ity
<l? The s e come c l o s e to those l imits
5. Knight D W , Demetr iou J D and Hamed M E . Stage d i s charge r e l at ions for compound
channe l s , Proc 1 s t Int . conf . Channe l s and Channe l Contro l S tructure s . Apr i l 1 9 84 ,
ed Smith K V H , Spr inger Ver l a g , 1 9 8 4 , 4 . 2 1 -4 . 3 6
Knight D W and Demetriou J D . Flood p l ain and main channel interaction , ASCE ,
J Hudrau l i c Eng . Vo l 109 , No 8 , Aug 1983 , 1 0 7 3 - 1092 .

TABLE A7 . 7 .. STATI STICAL ANALYS I S OF GOODNE S S OF FIT BETWEEN VARIOUS PREDICTION AS SUMPTIONS
AND KNIGHT AND DEMETRIOU RESULTS

Aspect B/b ARF = L O ARF = 0 . 6 ARF = 0 . 4 ARF = 0 . 2 ARF = 0 . 1 ARF = 0

r a t io Mean SO% Mean SO% Mean SO% Mean SO% Mean SO% Mean SO%

2 2 1 . 105 3 . 85 1 . 100 2 . 92 1 . 09 2 1. 82 1 . 07 7 1 . 84 1 . 04 7 1 . 97 1 . 003 1 . 33

2 3 1 . 04 3 3 . 8 1 ± 1 . 043 3 . 8 1 ± 1 . 04 3 3 . 8 1 ± 1 . 03 1 5.31 1 . 0 12 6 . 76 0 . 97 5 6 . 60

2 4 1 . 014 3 . 02 ± 1 . 0 1 4 3 . 02 ± 1 . 0 1 4 3 . 02 ± 1 . 000 2 . 07 0 . 986 2 . 96 0 . 95 1 2 . 95

*
deno tes no r egion 1 flows predicted with thi s va lue o f ARF
The unde r l ined values are those showing least variab i l i ty

6. Kie ly : unpub l i shed the s i s plus per sona l communication

TABLE A7 . 8 STATI STICAL ANALYS I S OF GOODNE S S OF F I T BETWEEN VARI OUS PREDICTION ASSUMPTIONS
AND KIELY ' S RESULTS

SMOOTH FLOOD PLAINS :

Roughnes s used : ARF value s :

Main eh Flood p l

0 . 20 0 . 34 0 . 37 0.5

0.011 0 . 010 Av d i s c% -6 . 0 -5 . 1 *
-5 . 1 *
S D % 3.0 2.5 2.5

ROUGH FLOOD PLAINS :

0.011 0 . 0157 -3 . 0 +1 . 8 +5 . 2
2.5 1.0 4.7

*
No te : deno tes no Region 1 results remained
7. Wormleaton P R , A l len J and Had j ipanos P , Proceedings ASCE , J Hy Div , Vol 1 0 8 , No HY9 ,

Sept 1 9 8 2 , pp 9 7 5-994

TABLE A7 . 9 : STATISTICAL ANALYS I S OF GOODNESS OF F I T BETWEEN VARIOUS PREDICTION ASSUMPTIONS

AND WORMLEATON et a l RESULTS

Upper f i gure : mean discr epancy %

Lower figure : variab i l ity %

Nomina l As sumed value o f ARF :

n value

1.0 0.8 0.6 0.4 0 . 24 0 . 20 0

0 . 01 1 +3 . 5 * +3 . 5 * +3 . 5 *
+2 . 6 +0 . 9 -2 . 0 -9 . 5

6.0 6.0 6.0 6.1 5.7 5.9 6.0

0 . 014 +0 . 3 * +0 . 3 * +0 . 3 * +0 . 3 *
-2 . 5 -3 . 8 - 13 . 6

2. 1 2.1 2. 1 2. 1 4.5 5.2 5.2

0 . 017 +5 . 3 * +5 . 3 * +4 . 5 +.1 . 7 -4 . 4 -7 . 3 - 19 . 8

2.2 2.2 2.6 6.6 9.9 9.5 7.3

0 . 02 1 +2 . 0 * +2 . 0 * -1 . 5 -7 . 2 - 18 . 2 -21 . 3 -34 . 0

6.7 6.7 11.8 16 . 8 16. 4 15 . 3 11.5

Note : * indicates Region 1 i s e l iminated under t he s e condit ions a t minimum dept h test�
O u r Ref : R/S/0203

add -

Dear name -

H R R E P O RT SR 28 1

I enclose details of corrections to t h e above d o c ument which came to light after

·

the document had been printed .

The enclosed material should include :

1. A complete contents list for Vol u m e 1 .

2. A Corrigenda p a g e for Volume 1 .

3. Corrected Figures 3 . 1 8 a n d 3 . 2 1 for Vol u m e 1 .

4. A Corrigenda p a g e for Volume 2 .

5. Figure A3 . 1 , which had been o m itted from Volume 2 .

6. A Corrigenda page supp lied b y D r D W K n i g h t for Appendix 4 i n Volume 2 .

Please include these amendments i n the Volumes that you have.

I apologise for any inconvenience this may c a u s e .

Yours sincerely

JAMES B WARK
Research Department

Encs
•
 DETAILED DEVELOPMENT OF D E SIGN METHOD

CONTENTS

PART 1

Page

l. INTRODUCTION 22

1.1 The importance of compound channel s and over-bank flows 22

1.2 Scope of treatment : straight channel s , the additional
problems posed by skew channel s , meandering and curvature 23
1.3 Approach to design: traditional c a l culation methods for
stage/discharge in compound channels 25

2. FLOW RESISTANCE IN CHANNELS OF COMPLEX CROSS-SECTION 28

2. 1 Resume of resistance for simple open channel s :

available formulae and their relevance and limitations 28
z·. 2 Compound cross sections : variations in hydraulic
parameters , field and laboratory cases ; inappropriate
to treat as single cross-section ( examples) ; the
summation method and variants in the literature ;
choice of vertical division 32
2.3 Allowing for the effects o f interaction: resume o f
approaches in literature and their l imitations ; force
balance and interfacial shear ; experimental studies
and typical results ; more fundamental methods based on
turbulence theory and present limitations . 36
2.4 Features influencing the degree o f interaction: approach
from dimensional analys i s ; dependence on relative
roughnesses of flood plain and main channe l ; on width
ratio ; on depth ratio ; on bank s lope ; cross-section
coherence . 44

3. HYDRAULIC DESIGN BASED ON EXPERIMENTAL ADJUSTMENT FACTORS 49

3.1 Research at Wallingford ; scope o f experiments and

measuremetits made . 49
3.2 Other sources of experimental data 50
3.3 Recommended basic method : separate c a l culations for
channel and flood plain , then summed , and corrected for
interaction effects 52
3.4 Analysis o f experimenta l resul t s : flow regions ; channel
coherence ; influence o f : flow depth flood plain width
channel side slope asymme try with supporting plots and
empirical relations 56
3.5 Separation of main channel and flood plain effects 65
3.6 Influence of flood plain roughnes s : form o f roughness used 68
3.7 Hydraulic design formulae : formulae for interference
effects in different flow regions ; choice of region;
goodness of fit to experimental data 74
CONTENTS ( CONT ' d)

Page

4. SKEW CHANNELS 82

4.1 The importance of momentum transfer with non-aligned flow 82

4.2 Research on skew compound channel s ; increased e ffect of 83
interference due to skewnes s

4.3 Extension o f design method t o skew channels ; limitations 86

5. OTHER SOURCES OF STAGE-DISCHARGE DATA FROM COMPOUND CHANNEL

EXPERIMENTS 88

5.1 Allowance for width/depth ratio in generalised predictive

functions 88
5.2 Other sources of research d at a ; scope and limitations ;
methodology; preliminary analys i s ; difficulties ; data
sources and review 89
5.3 Summary of information from other laboratory reseach 102
5.4 Skew channels 106
5.5 Field information; rivers for which data were available ;
comparison of calculated stage-discharge above-bank with
filed observations . 109
5.6 Conclusions from other data sources 1 15

6. TURBULENCE METHODS 120

6.1 Resume of turbulence theory a s currently applied to

compound cross-sections . Internal fluid mechanics . 120
6.2 Turbulence methods and comparison with FCF data . Review
of two-dimensional methods , non-dimensional eddy
viscosity and its evaluation, comparison with experiment 121
6.3 Application, generality and confirmation of turbulence
methods . Approaches to a s s e s s ing NEV, need for calibration 126
6.4 Comparison o f turbulence method with empirical method o f ·

prediction. Present limitations and potential 128

PART 2 (see VOLUME 2)

7. ANCILLARY TOPICS

7.1 Application to more complex cross sections ; parameter definition

7.2 Shear stress ; experimenta l evidence of variation around perimeter ;
method for assessment o f typical values in channel
7.3 Critical flow, energy and water leve l s ; Froude number
7.4 Sources of basic information on roughness ;

Lined channels
Natural rivers
Gravel bed chann e l s
Sand b e d channel s
Vegetation
CONTENTS ( CONT ' d)

Page

7.4 Sources of basic information on roughnes s ; 8

Lined channel s
Natural rivers
Gravel bed channels
Sand bed channels
Vegetation

7.5 Need for and utilisation o f field dat a ; extrapolation of 16

stage/discharge function
7.6 Incorporation into numerical model s : one-dimensional ; 18
two-dimensional : separate zone procedure vis-a-vis
lumped channel procedure

8. IRREGULAR PLAN FORM 21

8.1 Features of meandering flows in-bank . Characteristic 21

geometry o f natural channe l s , sinuo s i ty , p lanform losse s ,
reduction in conveyance therefrom , secondary currents
8.2 Above-bank flows in meandering channel s . Review of 25
existing information, reversal o f s econdary currents ,
momentum and flow exchange , interference and e f fect
on conveyance
8.3 Flow models for sinuous , meandering and irregular channel s . 3 2
Hydromechanics approach, momentum approach , fine grid
model l ing

9. SEDIMENT TRANSPORT 37

9.1 General aspects of s ediment transp or t . Modes o f transport 37

9.2 Transport process and theory. Dimensional analysis , 38
empirical functions , including Ackers and White ; suspended
load distribution
9.3 The influence of compound flow on bed material transport . 42
Typical river section, method o f calculation for sand and
gravel , loss of transport capacity above bank ful l ,
per formance o f equivalent s imple channel , implications for
fluvial morphology.
9.4 Suspended solids· in compound channel s 47

10. CONCLUDING REMARKS 48

10 . 1 Summary of hydraulic design formulae for the conveyance of 48

straight compound channe ls ; app l i cation logic ; tolerance on
assessment ; limitations
1 0 . 2 The advantages of compound channe l s : environmental ; 55
hydraulic; maintenance
10 . 3 State of knowledge and need for further research 57
CONTENTS (CONT ' d)

Page

11. ACKNOWLEDGEMENTS 60

12. REFERENCES 61

13 . NOMENCLATURE 72

APPENDICES

1. Dimensional analysis applied t o compound channels

2. Resistance functions for the SERC-FCF at Wallingford
3. Coherence
4. Turbulence method: solution for general cross section
shapes
5. Data on channe l roughness
6. Exampl e of channel geometry conversion and stage discharge
computation
7. Analyses of other sources of laboratory data ; tabular
surruna ries
HYDRAULIC DESI G N O F TWO-STAG E C H A N N ELS HR REPORT S R 28 1 ,
DEC 1 99 1

C O R R I G E N DA : to 2 2 Feb 1 9 9 1 , Vol u m e 1

p 9 , line 2; should refer to plate 2

p20, last line levels should read levees

p47, para 2 . 4. 8, 1 3; 0.43 should read 0 . 3

1 6; 0.85 0.5
0.94 0.9
1 8; 0.61 0.52

p 5 3 ; the two sentences at the foot of this page sho uld be at the head of p 5 5 .

p64, para 3 .4 . 20, 1 6; 3.18 should read 3 . 1 1

p64, para 3 . 4 . 2 1 , 1 4; 3.1 2 3.1 3

p 1 1 0, para 5 . 5 .4; The information in the last sentence is based on a

misunderstanding of earlier information, since amended by a personal
commun ication from Dr Myers.

p 1 1 2, 1 1 3, paras 5 . 5 . 9 to 5 . 5 . 1 0; The actual geometry of the R Main cross­

section 1 4 d iffers from that used here, which was based on published information
corrected since the report was writt e n . The reach is now known to b e of irregular
gradient with non-uniform flow, so the hydraulic gradients used in the analysis are
not valid . The information on the R M a i n e i n the text, figs 5 , 9 and 5 . 1 0 and in
table 5 . 3 should be disre g a rded . This reach o f river is no longer considered
suitable for this type of analysis.
 o.5rJ..
X +

0.45

0.40 A X +

0.35

0.30

�
:;:: 0.25
;,:.

0.20

Eq 3.24-->1
0.15 Individual
data

0.1 0 + me
Test - 02
X fp

0.05 "' me
+ Test - 06
A fp

0.00
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1 .0 1 .5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
QSTAR2

Figure 3.1 7 Q.2 against H* for separate zone: symmetric/asymmetric

comparison

0.5

0.4

.,�>
e,<:'
,o
,.,(§'
00
/)
0.3

•
�
0
I
E
rJJ

0.2

0.1

0
0 4 8 12 16 20
JBW13.1Bf4-92/LO

Figure 3.1 8 Variation of flood plain/main channel friction factor ratio for
smooth and rod-rou ghened flood plains, 8/b = 4.2, se = sF = 1
 0.50
"'
+ +
+
"'
0.45 r-

+
!;
!;
0.40 -

+
fr +
!;
0.35 !;
+
++
!;

+
0.30 "'

+
!;
:c !;
? r-
�

0.25 "'
;s !;
"±i-
+
!;
0.20 r- If;.
+
+
!;

0.15 r-
!;

+ + Averaged

+
data

0.10 r- t
!;
Test - 07a

t; Test - 02
1:
0.05 ""

L _[_ I I I I I J
0.00
0.0 0.2 0.4 0.6 0.8 1 .0 1.2 1.4 1.6 1 .8 2.0
DISDEFBF

Figure 3.21 comparison between smooth and

0.50
(1 .3)
!; X
+ +
0.45
!;
!;
X
+
1 7.8

(1.3)!; 1 5.6

+ .,.
0.40
fr + """
!;
00.
X
!;
0.35 �
. 12
.,
(1 .4) !;
00.
!; +
0.30
(1 .45) "' +. 9.3
Averaged
!; data
�

++ +
:2 0.25
i 7.3 Test - 07a

0.20 t; Test - 02

x Test - 07b
0.15

0.10

Added figures show sample

0.05 friction factor ratios

0.00 �----L_ ____ L_ ____ L_ ____ l_ ____ �____J_____J_____�----�--�

0 2 3 4 5 6 7 8 9 10
QSTAR2 JBW/3.22/4-92Jl0

Figure 3.22 0*2 against H*: comparison between smooth and rou g h
flood plains
HYDRAULIC DESIGN O F TWO-STAGE CHANN E LS HR REPORT S R 28 1 ,
DEC 1 99 1

C O R R I G E N DA : to 22 Feb 1 9 9 1 , Volume 2

p28, table 8 . 3 , line 1 3 ; last F 1 should be F2

p41 , eq 9 . 1 4; There should be double height brackets around the main part of the
function on the right

p 5 2, para 1 0. 1 . 1 0, 1 2; reference to e q u ations should be 1 0. 1 to 1 0. 1 2.

Appendix 1

Eq A 1 . 3 should read

Appendix 3

i�n i �n i �n

L A,V[ L_ A;lfo L P; J

Eq A 3 . 5 should be : i-1 i-1 i-1

i �n
L {{A, /f; f P, )
[A,
i�1
 �
7 0.5
:r:

0
0.3 0.5 1 .0

coH3 =
ya
Co nve nce as single channel
Sum o component conveyances

JBW/A3.1/4-921l..O

Figure A3.1 Channel coherence - COH3, as function of ratio of flood

plain flow depth to main cnannel depth : (i) wide horizontal
flood plains; (ii) narrow horizontal flood p lains; (iii) natural
river channel with sloping flood plains
Our Ref: DWK/MJB

Professor J.C.R. Hunt

Editor, Journal of Fluid Mechanics
Department of Applied Mathematics & Theoretical Physics
The University of Cambridge
CAMBRIDGE 7th June 1991

Dear Professor Hunt,

CORRIGENDUM - TURBULENT OPEN-CHANNEL FLOWS WITH VARIABLE

DEPTH ACROSS THE CHANNEL, SHIONO & KNIGHT, J,Fiuid Mechanics
(1991), Vol.222, pp.617-646.

Regrettably there is an error in equation (11) on page 621 of the above paper. I transcribed the
equations from an earlier draft leaving in another 'constant'. The two terms � and 11 should read:

11 = -

(I 2 1!2
+s )
(!:.)
s p 8

I apologise for this oversight and would be grateful if a corrigendum could be published in a future
edition.

Yours sincerely,

Dr. D.W. Knight

Reader in Hydtaulics & Fluid Mechanics

c.c. Dr. K. Shiono, University of Bradford