Wouter Callewaert, Brecht Versteele
characteristics of an open channel confluenceStudy of the influence of bed roughness on the flow
Academic year 2013-2014Faculty of Engineering and ArchitectureChairman: Prof. dr. ir. Peter TrochDepartment of Civil Engineering
Master of Science in Civil EngineeringMaster's dissertation submitted in order to obtain the academic degree of
Counsellors: Ir. Stéphan Creëlle, Ir. Laurent SchindfesselSupervisor: Prof. dr. ir. Tom De Mulder
Wouter Callewaert, Brecht Versteele
characteristics of an open channel confluenceStudy of the influence of bed roughness on the flow
Academic year 2013-2014Faculty of Engineering and ArchitectureChairman: Prof. dr. ir. Peter TrochDepartment of Civil Engineering
Master of Science in Civil EngineeringMaster's dissertation submitted in order to obtain the academic degree of
Counsellors: Ir. Stéphan Creëlle, Ir. Laurent SchindfesselSupervisor: Prof. dr. ir. Tom De Mulder
Foreword
During the first master year of Civil Engineering, our interest in hydraulic engineering grew. This
common interest resulted in the choice of a master dissertation in the field of hydraulics and we were
delighted with the assignment of this subject: “Study of the influence of bed roughness on the flow
characteristics of an open channel confluence”. The balance between performing experiments and
processing of results created an enjoyable working schedule. The difference in interpreting
measurement data and theory gave us a challenging year but also a very interesting one from which we
learned a lot. Getting used to both aspects of experimental research presented itself to be very
enriching.
We would like to express our sincere gratitude to the people who have supported us. First we would
like to thank our promoter, prof. dr. ir. Tom De Mulder, for giving us the opportunity to conduct this
master dissertation and for all his support. Our gratitude also goes to our supervisor ir. Stéphan
Creëlle for his guidance and feedback during the year. His expertise on the subject helped a lot for the
realization of this master dissertation. Furthermore, we would like to thank ir. Laurent Schindfessel
for his practical tips, handy tools, and support during the year.
Finally, we would like to thank our parents for the support and trust they gave us during our study of
Civil Engineering.
Wouter Callewaert & Brecht Versteele
Ghent, June 2014
Deze pagina is niet beschikbaar omdat ze persoonsgegevens bevat.Universiteitsbibliotheek Gent, 2021.
This page is not available because it contains personal information.Ghent University, Library, 2021.
Overview
Study of the influence of bed roughness on the flow characteristics of an
open channel confluence
Wouter Callewaert, Brecht Versteele
Supervisor: Prof. dr. ir. Tom De Mulder
Counsellors: Ir. Stéphan Creëlle, Ir. Laurent Schindfessel
Master's dissertation submitted in order to obtain the academic degree of
Master of Science in Civil Engineering
Department of Civil Engineering
Chairman: Prof. dr. ir. Peter Troch
Faculty of Engineering and Architecture
Academic year 2013-2014
Keywords:
Open channel confluence, flow characteristics, experimental, bed roughness, artificial grass
Study of the influence of bed roughness on
the flow characteristics of an open channel
confluence
Wouter Callewaert and Brecht Versteele
Supervisor: prof. dr. ir. Tom De Mulder
Abstract: An experimental study on a 90° open
channel confluence is performed to investigate the
influence of bed roughness on the flow characteristics.
To this end, artificial grass, which acts as a well-
submerged vegetation, is put on the channel bottom to
increase the bed roughness. Measurement data is
gathered from both configurations and compared. From
this comparison, it was found that the velocity
distribution over the channel depth changes for the grass
configuration. Because of this rearrangement, higher
velocities are present at the water surface and the
dimensions of the separation zone change. Furthermore,
the width of the mixing layer develops in a different way
when going downstream and a different overall flow
pattern is obtained.
Keywords: Open channel confluence, flow
characteristics, experimental, bed roughness, artificial
grass, mixing layer, zone of separation
I. INTRODUCTION
Open channel confluences are very common in
nature whereby every river consists of a main channel
and some tributaries. Since this is a very important
part of the river, it has already led to numerous
studies presented in literature. The main flow features
of a confluence (Figure 1) are a stagnation zone at the
upstream corner, a zone of separation just
downstream of the tributary, a contracted flow region
between the separation zone and the wall opposing
the tributary and a mixing layer.
Figure 1: flow characteristics in an open channel
confluence
The zone of separation is an area just downstream
of the junction where the flow has a small velocity
and where recirculation exists. It is created by the
momentum of the side flow which causes the main
flow to detach at the downstream corner of the
junction. Studies (Shumate, 1998; Best and Reid,
1984) have shown that the separation zone is
significantly longer and wider at the surface level than
at the bottom[1] [2].
L. Crombé (2013) also indicated with numerical simulations that the friction coefficient is a determining parameter [3]. The dimensions of the
zone of separation decrease with increasing bed
friction. Furthermore, the dimensions of this zone
decrease when the flow ratio from the main channel to
the downstream channel increases. The presence of
the separation zone forces the water from both
channels to pass a smaller zone. This contracted
region will have bigger downstream velocities
because the same amount of water has to pass a
smaller zone. The aim of the present paper is to investigate the
influence of bed roughness on these flow
characteristics. To increase the friction factor at the
bottom, the flume is equipped with a cover of
artificial grass. The grass blades of this cover have an
average height of 3cm. In the past, the flume without
the grass cover has already been discussed by
Schindfessel et al. (2014) and will be used as
benchmark [4].
II. LABORATORY EXPERIMENTS
A. Experimental set-up
The experiments were performed at the hydraulics
laboratory of Ghent. The open channel facility
consists of a 90° channel confluence with a chamfered
rectangular cross-section with concrete walls. The
main channel has a length of 33.18m. The tributary is
5.17m long and intersects with the main channel at
13.12m (measured from upstream). Both channels
have a constant width Wd of 0.98m. The coordinate
system originates at the upstream corner of the
junction. The x-axis is orientated along the main
channel towards upstream while the y-axis is pointing
in the downstream direction of the tributary.
The coordinates are always represented
dimensionless by dividing them by the channel width
Wd. The discharge ratio q* is defined as:
(1)
Where Qm and Qt denote the incoming discharge of
the main channel and the tributary respectively, Qd
expresses the downstream discharge and is the sum of
the discharge of both inlet channels. The downstream
discharge Qd of 40 l/s together with a constant
downstream water level hd of 0.415m yields to a
constant downstream Froude number Frd of 0.05. This
value is typical for lowland rivers [5]. The choice for
q* = 0.25 was based on the value of the reference
case.
Schindfessel et al. (2014) carried out their
experiments in the same experimental set-up, with
equal geometrical and flow parameters. As flumes
with a chamfered rectangular cross-section are quite
uncommon in literature, a comparison was made with
the experiments of Shumate (1998). The latter carried
out his experiments in the Iowa Institute of Hydraulic
Research. The flume has similar dimensions but the
cross-section is not chamfered. Moreover, total
discharge and velocity are higher than in the present
study. As a consequence of the lower velocities in the
present experiments, water level differences will be
negligible. According to Schindfessel et al. (2014),
one of the main differences due to the chamfers, is the
depth of the zone of separation. This zone does not
gradually decrease in width towards the bottom, like
observed by Shumate, but has a vertical interface that
ends abruptly at z/Wd = 0.1 (just above the chamfer).
B. Measurement methodology
Flow velocities at the water surface are measured
using large scale surface particle image velocimetry
(LSSPIV). Therefore a 1920x1080 pixel camera is
placed at a fixed height above the water surface,
taking images at a rate of 15Hz. The fluid is seeded
with a tracing material that is recognizable by the
camera. The applied material consists of
polypropylene particles, covered in a white coating.
The duration of the recording is fixed at 180 seconds,
this results in the best equilibrium between quality
and quantity of the measurements. The processing
from frames to velocity data is done with the freeware
PIVlab 1.32.
To create velocity profiles in a cross-section of the
flume the ‘Vectrino II Profiling Velocimeter’ is used.
This device is a high-resolution acoustic Doppler
velocimeter (ADV) that can be used to measure
turbulence and velocities in 3D. Based on
observations from Schindfessel et al. (2014), a
measurement duration of 2min was chosen as
sufficient to obtain time-averaged velocities. To
obtain an accurate velocity profile over the entire
cross-section, nine vertical lines are measured where
each line consists of 15 measurement points. By
interpolation between the measured data, an
approximate plot is obtained with velocities for the
entire cross-section.
Figure 2: Experimental flume with location of the cross-
sections measured by the profiling ADV
Four different cross-sections are measured to obtain
velocity profiles using the profiling ADV (Figure 2).
The sections at x/Wd = 0 and y/Wd = 0 will give
valuable information about the inflow section of the
main channel and the tributary, the sections at
x/Wd = -0.5 and x/Wd = -1.33 will visualize the
complex flow patterns at the confluence.
III. RESULTS
A. Velocity profiles
According to Kleinhans (2008), the vegetation is
well-submerged when the water depth is at least five
times the height of the vegetation [6]. In the present
study, the height of the grass cover is only 3cm while
the water depth is 41.5cm. This means that the flow is
delayed within the vegetation but the flow that goes
over the vegetation is not blocked. As a result, a
logarithmic velocity profile is expected for the flume
covered with grass. This assumption was checked by
measuring a vertical velocity profile with the profiling
ADV (Figure 3). The location of this cross-section is
chosen upstream in the tributary at y/Wd = -1.92, in
the middle of the cross-section (x/Wd = -0.50).
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
Figure 3: Velocity profiles - left: concrete, right: grass
The velocity in the flume with grass is almost zero
when the water touches the grass cover. This indicates
that the above mentioned assumption is valid. The
velocity profile in the concrete flume reaches an
almost constant value closer to the bottom than the
one in the grass section. Since the water level
differences between both configurations are almost
negligible, the water has to pass a smaller area with an
equal flow rate. This causes larger velocities in the
upper part of the water compared with the concrete
flume and changes the velocity profile.
B. General velocity patterns
The main effect of the placement of the grass cover
is a rearrangement of the velocity distribution. This
difference is more pronounced in the tributary
because its discharge is three times larger than the one
of the main channel.
Due to the larger velocities in the tributary, the flow
enters the confluence with a larger force and pushes
the fluid from the main channel further to the outer
side. The separation zone increases in width and
decreases in length (Figure 4). This is partly
contradictory with the literature which predicts that
both the length and the width of the separation zone
decrease with increasing friction coefficient. It can be
concluded that the altered velocity distribution
obtained through the grass cover will dominate the
flow characteristics.
Figure 4: Vector plots obtained with the LSSPIV showing
velocity patterns – top: concrete, bottom: grass (The dotted
contour line indicates points with a longitudinal velocity of
zero).
Cross-sectional plots at x/Wd = -1.33 are shown in
Figure 5. For the flume with the grass cover, the
separation zone extends vertically but ends already at
mid-depth (z/Wd = 0.20). Lateral inflow from the
tributary collides with the flow of the main channel,
bends downward and is redirected towards the left
bank. This is probably the reason for the abrupt end
of the separation zone at a depth of z/Wd = 0.20. This
phenomenon does not appear in the concrete flume
because here, the cross-sectional velocities are
smaller and have not the power to reach the left bank
after the collision with the stream of the main channel,
instead the main flow redirects them downstream. The
highest upstream velocities, in the separation zone,
appear at the water surface and decrease in magnitude
when moving towards the bottom.
Figure 5: Cross-sectional plots at x/Wd = -1.33 with
longitudinal velocity (color scale) and cross-sectional
velocity (arrows) – Top: concrete, bottom: grass
C. Mixing and shear layer
The mixing layer width δ is defined as the ratio
between the outer velocity difference ΔU and the
maximum velocity gradient of the cross-section:
(2)
The width of the mixing layer is shown in Figure 6
and initially increases for both configurations (The
flow direction is from right to left). For the concrete
flume, the width of the mixing layer decreases again
from x/Wd = -0.45 while for the flume with the grass
cover, it increases until a value of x/Wd = -0.6
where it seems to remain constant. These results agree
well with the trend observed by Mignot et al. (2013)
[7]. This plateau is related to the strong lateral
confinement of the flow when it reaches the
confluence.
Figure 6: Mixing layer width
The instability of the shear layers was investigated
by means of vorticity plots and optical observations.
The vorticity of the shear layer starting at the
upstream corner was low in comparison with the shear
layer at the separation zone. These values are
confirmed with the optical observations. A lot of
vortices are shed from the downstream corner while
this is not the case for the upstream corner. As a
consequence only the shear layer at the separation
zone is unstable. The optical observations also
pointed out that the size of these vortices increase
over a longer distance at the grass flume.
IV. CONCLUSIONS
Experiments were performed to investigate the
influence of bed roughness on the flow characteristics
at a 90° confluence. The grass cover that was placed
on the channel bottom was small enough to presume
well-submerged vegetation. By equipping the flume
with this grass cover, the available cross-section for
the fluid to pass decreased. This resulted in an altered
velocity distribution. This difference was more
pronounced in the tributary because its discharge was
three times higher. As a result, the separation zone
could develop more in width and its length decreased
Measurement data from the profiling ADV showed
that the depth of the recirculation zone is limited by
the effect of the chamfers and the return flow to the
left bank located just above the channel bottom.
Furthermore, the mixing layer is wider in the grass
configuration and will extend further downstream
before both streams are mixed completely. At last,
optical observations showed that only the shear layer
starting at the downstream corner was unstable. The
instability arose as vortices that were shed and grew
along the shear plane.
REFERENCES
[1] E.D. Shumate, ‘Experimental description of flow at an
open-channel junction’, Unpublished Master thesis.
University of Iowa, 1998
[2] J.L. Best and I. Reid, ‘Separation zone at open channel
confluences’, Journal of Hydraulic Engineering, Vol.
110,11,1588-1594, 1984
[3] L. Crombe, S. Creelle, T.D.Mulder, Numerieke modellering
van een riviersamenvloeiing: Onderzoek naar de invloed van
vegetatie op de stromingspatronen(thesis).” 2013.
[4] L. Schindfessel, S. Creëlle, T. Boelens and T. De Mulder,
‘Flow patterns in an open channel confluence with a small
ratio of main channel to tributary discharge, Hydraulics
laboratory, Department of Civil Engineering, Ghent
University, Belgium, 2014
[5] Khublaryan, 2009, Surface Waters: Rivers, Streams, Lakes
and Wetlands. In Types and properties of water. Oxford:
EOLSS Publishers Co. Ltd.
[6] M. Kleinhans, Hydraulic roughness, 2008
Last visited December 2, 2013, from Utrecht University,
Faculty of Geosciences.
Web site:
www.geog.uu.nl/fg/mkleinhans/teaching/rivmorrough.pdf
[7] E. Mignot, I. Vinkovic, D. Doppler, N. Riviere, ‘Mixing
layer in open-channel junction flows’, Environmental Fluid
Mech, 2013
Studie van de invloed van bodemruwheid op
de stromingskarakteristieken van een
samenvloeiing van open kanalen
Wouter Callewaert en Brecht Versteele
Promotor: prof. dr. ir. Tom De Mulder
Samenvatting: Om de invloed van bodemruwheid op
de stromingskarakteristieken van een samenvloeiing van
open kanalen te bestuderen, wordt een experimenteel
onderzoek uitgevoerd op twee kanalen die elkaar treffen
onder een hoek van 90°. Teneinde de bodemruwheid te
verhogen wordt een dunne laag kunstgras op de bodem
van het kanaal aangebracht. Daarna wordt meet-data
verzameld van de originele betonnen goot en de goot met
gras en worden de resultaten onderling vergeleken.
Hieruit blijkt dat het snelheidsprofiel bij gras anders is
dan bij beton. Door deze herverdeling bevinden zich
grotere snelheden aan het wateroppervlak en zijn de
dimensies van de separatiezone verschillend. Het verloop
van de breedte van de menglaag is verschillende in
beide configuraties en een veranderd stromingspatroon
wordt verkregen.
Sleutelwoorden: Samenvloeiing van open kanalen,
stromingskarakteristieken, experimenteel,
bodemruwheid, kunstgras, menglaag, separatiezone
I. INLEIDING
Rivieren bestaan gewoonlijk uit een hoofdkanaal
waarop verscheidene zijtakken aansluiten.
Samenvloeiingen van open kanalen komen dan ook
erg vaak voor in de natuur en maken er een belangrijk
deel van uit. In de literatuur zijn dan ook al
verscheidene studies te vinden die dieper op deze
materie ingaan. De belangrijkste componenten van
een samenvloeiing van kanalen zijn (Figuur 1) de
stagnatiezone aan de opwaartse hoek, een
separatiezone gelegen net afwaarts van de zijtak, een
zone waar de stroming wordt samengedrukt tussen de
separatiezone en de overstaande muur en een
menglaag.
De separatiezone is het gebied net afwaarts van de
samenvloeiing waar de stroming een kleine snelheid
heeft en waar recirculatie van de stroming plaatsvindt.
Ze wordt gecreëerd door de kracht waarmee de
stroming uit de zijtak in het hoofdkanaal terecht komt.
Hierdoor wordt de stroming uit de hoofdtak
losgemaakt aan de stroomopwaartse hoek van de
samenvloeiing. Eerdere studies (Shumate, 1998; Best
and Reid, 1984) tonen aan dat de separatiezone
beduidend langer en breder is aan het oppervlak dan
aan de bodem [1] [2]).
Figuur 1: Stromingskarakteristieken in een samenvloeiing
van open kanalen.
L. Crombé (2013) toonde met numerieke
simulaties ook aan dat de wrijvingscoëfficiënt een
bepalende factor is [3]. Hoe hoger de ruwheid, hoe kleiner de separatiezone is. Verder nemen de dimensies van deze zone ook af wanneer de debietsverhouding van het hoofdkanaal tot het zijkanaal toeneemt. De aanwezigheid van deze zone dwingt het water van beide takken om doorheen een kleinere sectie te vloeien. In deze samengedrukte zone zal het water grotere stroomafwaartse snelheden hebben aangezien dezelfde hoeveelheid water een kleinere zone moet passeren.
Het doel van dit onderzoek is om de invloed van de
bodemruwheid op deze stromingscomponenten te
onderzoeken. Om de wrijvingsfactor aan de bodem te
verhogen wordt de testgoot uitgerust met
kunstgrasmatten. De grassprieten van deze matten zijn
gemiddeld 3cm lang. Eerder werden in dezelfde goot
maar dan zonder de grasmatten al experimenten
uitgevoerd door Schindfessel et al. (2014) [4]. Deze
worden gebruikt als referentie.
II. EXPERIMENTEN
A. Experimentele opstelling
De experimenten worden uitgevoerd in het labo
Hydraulica te Gent. Deze accommodatie bevat een
samenvloeiing van open kanalen die elkaar treffen
onder een hoek van 90°. De doorsnede bestaat uit een
afgeschuinde rechthoekige sectie met betonnen
muren. De lengte van het hoofdkanaal is 33.18m. De
zijtak is 5.17m lang en snijdt het hoofdkanaal op een
afstand van 13.12m (van opwaarts gemeten). Beide
kanalen hebben een constante breedte Wd van 0.98m.
De oorsprong van het assenstelsel is gepositioneerd
aan de stroomopwaartse hoek van de samenvloeiing.
De x-as is georiënteerd langs het hoofdkanaal, gericht
naar opwaarts, en de positieve zin van de y-as is
gericht in de afwaartse zin van de zijtak. De
coördinaten worden dimensieloos weergegeven door
ze te delen door de kanaalbreedte. De
debietsverhouding q* wordt gedefinieerd als:
(1)
Met Qd en Qt respectievelijk het instromend debiet
van de hoofdtak en de zijtak. Qd is het afwaartse
debiet wat de som is van de debieten van beide
instroomkanalen. Het afwaartse debiet Qd van 40 l/s
samen met de constante stroomafwaartse waterhoogte
hd van 0.415m leidt tot een constant stroomafwaarts
Froude nummer Frd van 0.05. Deze waarde is
typisch voor laagland rivieren [5]. De keuze voor
q* = 0.25 was gebaseerd op de waarde uit het
referentiegeval.
Schindfessel et al. (2014) voerden experimenten uit
in dezelfde experimentele opstelling, met dezelfde
geometrische - en stromingskarakteristieken.
Aangezien goten met een afgeschuinde doorsnede vrij
ongewoon zijn in de literatuur, wordt een vergelijking
gemaakt met de experimenten van Shumate (1998).
Die laatste voerde zijn experimenten uit in het
hydraulisch onderzoeksinstuut te Iowa. De goot daar
heeft gelijkaardige dimensies maar de doorsnede is
niet afgeschuind. Bovendien zijn het totale debiet en
de snelheden groter dan in de huidige studie. Als
gevolg van de lagere snelheden in de huidige studie
zijn de waterhoogteverschillen verwaarloosbaar.
Volgens Schindfessel et al., is één van de grootste
verschillen, veroorzaakt door de afschuiningen, de
diepte van de separatiezone. De breedte van deze
zone neemt niet geleidelijk af naar de bodem toe,
zoals waargenomen was door Shumate, maar heeft
een verticaal raakvlak dat abrupt stopt op z/Wd = 0.1
(Net boven de afschuining).
B. Meetmethode
Met de LSSPIV (Large scale surface particle image
velocimetry) worden stromingssnelheden aan het
wateroppervlak gemeten. Hiervoor wordt een
1920x1080 pixel camera op een vaste hoogte boven
het wateroppervlak geplaatst die beelden neemt met
een frequentie van 15Hz. Het water oppervlak wordt
bestrooid met een materiaal dat herkenbaar is voor de
camera. Dit materiaal bestaat uit polypropyleen
deeltjes die bedekt zijn met een witte deklaag. De
duur van een opname is 180 seconden, dit resulteert
in het beste evenwicht tussen de kwaliteit en
kwantiteit van de metingen. De verwerking van de
beelden naar snelheidsdata wordt uitgevoerd met de
freeware PIVlab 1.32.
Om een snelheidsprofiel te verkrijgen in een
doorsnede van de goot wordt de ‘Vectrino II Profiling
Velocimeter’ gebruikt. Dit toestel is een hoge-
resolutie akoestische Doppler Velocimeter ADV) die
kan gebruikt worden om turbulentie en snelheden in
3D te meten. Een meetduur van 2min werd, via
observaties van Schindfessel et al. (2014), als
voldoende beschouwd om tijdsgemiddelde snelheden
te verkrijgen. Om een nauwkeurig snelheidsprofiel
over de volledige doorsnede te verkrijgen zijn 9
verticale lijnen opgemeten waarbij elke lijn bestaat uit
15 meetpunten. Via interpolatie tussen de gemeten
data kan een benaderende plot verkregen worden met
snelheden over de volledige doorsnede.
Vier verschillende doorsnedes werden opgemeten
met de ADV om een snelheidsprofiel te verkrijgen.
De secties op
x/Wd = 0 en y/Wd = 0 geven belangrijke informatie
over de instroming uit het hoofd -en zijkanaal. De
secties x/Wd = -0.5 en x/Wd = -1.33 dienen om de
complexe stromingsprocessen te visualiseren die
optreden aan de samenvloeiing.
Figuur 2: Experimentele goot met de locatie van de
doorsneden die opgemeten worden met de ADV
III. RESULTATEN
A. Snelheidsprofielen
Volgens Kleinhans (2008) wordt vegetatie als
“voldoende onder water” beschouwd wanneer de
waterdiepte ten minste vijf keer de vegetatiehoogte is
[6]. In de huidige studie is de hoogte van de
grasbedekking slechts 3cm terwijl de waterhoogte
41.5cm is. Hierdoor wordt de stroming in de vegetatie
vertraagd terwijl de stroming die over de vegetatie
gaat niet belemmerd wordt. Bijgevolg wordt een meer
logaritmisch snelheidsprofiel verwacht in de grasgoot.
Deze aanname wordt gecontroleerd door een verticaal
snelheidsprofiel op te meten met de ADV (Figuur 3).
Dit profiel wordt opwaarts in de zijtak opgemeten op
y/Wd = -1.92 in het midden van de sectie
(x/Wd = -0.50).
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
Figuur 3: Snelheidsprofielen - links: beton / rechts: gras
De snelheid in de grasgoot is bijna nul aan de
bovenkant van de matten. Dit duidt erop dat de
bovenvermelde aanname correct is. Het
snelheidsprofiel in de betonnen goot bereikt een bijna
constante waarde dichter bij de bodem dan het profiel
in gras. Aangezien het waterhoogteverschil tussen
beide configuraties verwaarloosbaar is vloeit het
water door een kleinere sectie met hetzelfde debiet.
Hierdoor verandert het snelheidsprofiel over de diepte
en worden hogere snelheden aan het water oppervlak
verkregen en lagere aan de bodem vergeleken met de
betonnen goot.
B. Algemeen stromingspatroon
Het belangrijkste effect van de plaatsing van de
grasmaten is een herverdeling van de snelheid. Dit
verschil komt meer tot uiting in de zijtak aangezien
het debiet daar drie keer groter is dan in het
hoofdkanaal. Door de grotere snelheden in de zijtak,
komt de stroming de samenvloeiing met een grotere
kracht binnen en duwt de stroming van het
hoofdkanaal verder naar de buitenzijde. De
separatiezone neemt af in lengte en toe in breedte
(Figuur 4).
Figuur 4: Vectorplot van de LSSPIV data die de
snelheidspatronen weergeeft. - boven: beton / onder: gras
(de streeplijn geeft punten aan waar u=0)
Dit is gedeeltelijk tegenstrijdig met de literatuur
waarin voorspeld wordt dat zowel de lengte als de
breedte van de separatiezone afnemen met
toenemende wrijvingscoëfficiënt. Uit het vorige kan
besloten worden dat de hogere snelheden aan het
wateroppervlak in de grasconfiguratie de
stromingskarakteristieken zullen domineren. Figuur 5
toont plots van de snelheden op x/Wd = -1.33.
Figuur 5: Dwarsdoorsnede op x = -1.33 met longitudinale
snelheden (kleur) en laterale snelheden (pijlen) -
Boven: Beton / Onder: Gras
In de grasgoot strekt de separatiezone zich verticaal
uit tot op een diepte van z/Wd = 0.20. Laterale
instroming van de zijtak botst tegen de stroming van
het hoofdkanaal. Hierdoor wordt deze stroming
neerwaarts gebogen en geheroriënteerd naar de linker
zijde. Dit is de reden voor het abrupte einde van de
separatiezone op z/Wd = 0.20. Dit fenomeen komt
niet voor in de betonnen goot omdat de longitudinale
snelheden in de zijtak hier kleiner zijn. Hierdoor
hebben ze niet de kracht om, na de botsing met de
stroming van het hoofdkanaal, nog de linkerzijde te
bereiken. In plaats daarvan worden ze via de stroming
uit het hoofdkanaal naar afwaarts getransporteerd. In
beide gevallen bevinden de hoogste stroomopwaartse
snelheden in de separatiezone zich aan het
wateroppervlak en nemen ze in grootte af naar
beneden toe.
C. Meng –en Afschuiflaag
De breedte van de menglaag δ wordt gedefinieerd
als de verhouding tussen het uiterste snelheidsverschil
ΔU tot de maximale snelheidsgradient in de
doorsnede:
(2)
De breedte van de menglaag δ wordt weergegeven
in Figuur 6 en neemt initieel toe in beide
configuraties. (De stroming richting is van links naar
rechts). In de betonnen goot neemt de dikte van de
menglaag af vanaf x/Wd =-0.45 terwijl ze in de
grasgoot toeneemt tot een waarde van x/Wd = -0.6
waarna ze constant blijft. Deze resultaten komen goed
overeen met de trend die waargenomen werd door
Mignot et al. (2013) [7]. Dit plateau heeft te maken
met de sterke laterale begrenzing van de stroming
wanneer ze de samenvloeiing bereikt.
Figuur 6: Breedte van de menglaag
Instabiliteit van de afschuiflaag wordt onderzocht
met behulp van vorticiteitplots en visuele observaties.
De vorticiteit van de afschuiflaag, die begint aan de
stroomopwaartse hoek, is klein in vergelijking met de
afschuiflaag aan de separatiezone. Dit wordt
bevestigd door de visuele observaties. Verscheidene
vortexen ontstaan aan de stroomafwaartse hoek
terwijl dit niet het geval is aan de stroomopwaartse
hoek. Bijgevolg is enkel de afschuiflaag aan de
separatiezone onstabiel. De visuele observaties maken
ook duidelijk dat de grootte van de vortexen over een
langere afstand toeneemt in vergelijking met de
grasgoot.
IV. CONCLUSIE
Om de invloed van bodemruwheid op de
stromingskarakteristieken in een samenvloeiing onder
een hoek van 90° te onderzoeken werden
experimenten uitgevoerd. Het gras dat op de bodem
werd geplaatst was klein genoeg om als ‘voldoende
onder water’ beschouwd te worden. Door de goot uit
te rusten met deze grasmatten wordt de beschikbare
sectie, waarlangs het water kan stromen, verkleind.
Dit resulteert in een veranderde snelheidsverdeling.
Dit verschil komt meer tot uiting in de zijtak omdat
het debiet hier drie keer groter is dan in de zijtak. Dit
resulteert in een bredere separatiezone met een
kleinere lengte. Meet-data van de ADV tonen aan dat
de diepte van de recirculatiezone begrensd is door het
effect van de afschuiningen en de terugkerende
stroming naar de linkerzijde van het hoofdkanaal die
zich net boven de kanaalbodem bevindt. Verder is de
menglaag breder in de grasconfiguratie en zal ze zich
verder uitstrekken naar afwaarts toe voor beide
stromingen volledig vermengd zijn. Als laatste
toonden visuele observaties aan dat de afschuiflaag
die start aan de stroomafwaartse hoek onstabiel is.
Deze instabiliteit was merkbaar door het afscheiden
van vortexen die groeien naar afwaarts toe.
REFERENTIES
[1] E.D. Shumate, ‘Experimental description of flow at an
open-channel junction’, Unpublished Master thesis.
University of Iowa, 1998
[2] J.L. Best en I. Reid, ‘Separation zone at open channel
confluences’, Journal of Hydraulic Engineering, Vol.
110,11,1588-1594, 1984
[3] L. Crombe, S. Creelle, T.D.Mulder, Numerieke modellering
van een riviersamenvloeiing: Onderzoek naar de invloed van
vegetatie op de stromingspatronen(thesis).” 2013.
[4] L. Schindfessel, S. Creëlle, T. Boelens en T. De Mulder,
‘Flow patterns in an open channel confluence with a small
ratio of main channel to tributary discharge, Hydraulics
laboratory, Department of Civil Engineering, Ghent
University, Belgium, 2014
[5] Khublaryan, 2009, Surface Waters: Rivers, Streams, Lakes
and Wetlands. In Types and properties of water. Oxford:
EOLSS Publishers Co. Ltd.
[6] M. Kleinhans, Hydraulic roughness, 2008
Last visited December 2, 2013, from Utrecht University,
Faculty of Geosciences.
Web site:
www.geog.uu.nl/fg/mkleinhans/teaching/rivmorrough.pdf
[7] E. Mignot, I. Vinkovic, D. Doppler, N. Riviere, ‘Mixing
layer in open-channel junction flows’, Environmental Fluid
Mech, 2013
Contents
Introduction ............................................................................................................................................... 1
1. Literature review ............................................................................................................................... 3
1.1. General concepts ....................................................................................................................... 3
1.1.1. Manning, Weisbach & Chézy coefficients ........................................................................ 4
1.1.2. Moody diagram ............................................................................................................. 6
1.1.3. Momentum vs energy resistance coefficients ............................................................. 7
1.1.4. Energy Dissipation ....................................................................................................... 8
1.2. Resistance .................................................................................................................................. 9
1.2.1. Boundary Layer Theory ................................................................................................ 9
1.2.2. Nonalluvial channel .................................................................................................... 12
1.2.3. Alluvial channels ......................................................................................................... 13
1.3. Vegetated Channels ................................................................................................................. 15
1.3.1. Velocity Profiles .......................................................................................................... 15
1.3.2. Energy exchange ..........................................................................................................17
1.3.3. Flexibility .................................................................................................................... 20
1.4. Resistance of composite or compound channels .................................................................. 23
1.5. Secondary currents in open channels .................................................................................... 25
1.6. Open channels ......................................................................................................................... 27
1.6.1. Open channel confluence ........................................................................................... 27
1.6.2. Comparison of confluence and diversion flow .......................................................... 30
1.6.3. Flow variables ............................................................................................................. 31
1.7. Flow instability ........................................................................................................................ 37
1.7.1. Kelvin-Helmholtz instability .......................................................................................... 38
1.7.2. Instability at confluences ........................................................................................... 39
1.7.3. Influence of a gradient on shear flow stability .......................................................... 42
1.7.4. Shear flow instability criteria ..................................................................................... 44
1.8. Conclusion ............................................................................................................................... 47
2. Experiments ..................................................................................................................................... 49
2.1. Introduction ............................................................................................................................. 49
2.2. Layout of the flume ................................................................................................................. 49
2.3. Measurement devices .............................................................................................................. 53
2.3.1. Large-scale surface particle image velocimetry ........................................................ 53
2.3.2. Acoustic Doppler Velocimetry.................................................................................... 57
2.4. Preparations ............................................................................................................................ 65
2.5. Results ...................................................................................................................................... 69
2.5.1. Comparison velocity profile concrete and grass ....................................................... 69
2.5.2. Comparison LSSPIV & ADV ....................................................................................... 70
2.5.3. General flow pattern ................................................................................................... 72
2.5.4. separation zone ........................................................................................................... 76
2.5.5. Mixing layer................................................................................................................. 77
2.5.6. Secondary current ....................................................................................................... 81
2.5.7. Shear planes and vorticity .......................................................................................... 85
2.6. Conclusion ............................................................................................................................... 91
Bibliography ................................................................................................................................................. 93
List of symbols
Symbol Unit Explanation
Cross-sectional area
Coefficient of Chézy
- Weisbach friction factor
- Froude number
Gravitational acceleration, taken equal to 9.81m/s²
Water depth
Vegetation height
Characteristic roughness height
Coefficient of Manning
Downstream discharge
Discharge of the main channel
Discharge of the tributary
- Discharge ratio Qm/Qd
- Slope of the hydraulic grade line
- Reynolds number
Hydraulic radius
- Richardson number
Velocity in x-direction
Shear velocity
Double averaged mean velocity
Velocity in y-direction
Total velocity
Velocity in z-direction
Downstream width of the channel
- Dimensionless coordinates
- Dimensionless velocities
Specific weight of the fluid
° Confluence angle
Dynamic viscosity
Eddy viscosity
Density
Shear stress
Kinematic viscosity
Vorticity
- Von Karman constant = 0.41
E Elasticity modulus
I Moment of Inertia
List of abbreviations
Abbreviation Explanation
ADV Acoustic Doppler velocimeter
DCC Direct cross-correlation
FFT Fast Fourier transforms
IA Interrogation area
K-H Kelvin-Helmholtz
LSSPIV Large-scale surface particle image velocimetry
PTV Particle tracking velocimetry
List of figures
Figure 1-1: Components of flow resistance ................................................................................................... 3
Figure 1-2: Moody Diagram for open channels with impervious rigid boundary [Ben Chie Yen (2002)]6
Figure 1-3: Energy cascade [P.A. Davidson (2004)] .................................................................................... 8
Figure 1-4: Boundary layer δ and laminar sublayer ............................................................................... 9
Figure 1-5: Inner and outer law of boundary layer .................................................................................... 10
Figure 1-6: Hydraulically smooth vs. hydraulically rough ......................................................................... 10
Figure 1-7: Channel boundary classification [Ben Chie Yen (2002)] ......................................................... 11
Figure 1-8: Types of rough-surface flows: (a) k-type; (b) d-type; (c) transitional-type [Chow (1959)] .. 12
Figure 1-9: k-type roughness – mechanism [C. Polatel (2006)] ............................................................... 13
Figure 1-10: Possible movements of particles near the bed ....................................................................... 13
Figure 1-11: Forms of bed roughness in sand-bed channels [G.J. Arcement (1989)] .............................. 14
Figure 1-12: Velocity profile - well-submerged vegetation [Galema (2008)] ........................................... 15
Figure 1-13: Velocity profile – submerged vegetation [Galema (2008)] .................................................. 15
Figure 1-14: Three zone sub model [Nezu and Sanjou (2008)] ................................................................. 16
Figure 1-15: Emerged vegetation [Galema (2008)] ................................................................................... 17
Figure 1-16: Energy exchange model – vegetated channels [H.J.S. Fernando (2013)] ........................... 18
Figure 1-17: Effect of submergence depth on Reynolds stresses [Nezu et al. (2008)] ............................. 18
Figure 1-18: Quadrant conditional [Nezu and Nakagawa (1977)] ............................................................. 19
Figure 1-19: Circle C: Sweep – Circle B: Ejection [Nezu and Nakagawa (1977)] ...................................... 19
Figure 1-20: Flexible vegetation compared to rigid vegetation [Carollo et al. (2005)] .......................... 20
Figure 1-21: Flow patterns in dense flexible vegetation. (a) Erect or rigid. (b) Swaying (no organized).
(c) Monami (organized). (d) Prone. [H.J. (2013)] ..................................................................................... 21
Figure 1-22: Subregions of in function of the water depth [Fu-Chun Wu et al. (1999)] .................... 21
Figure 1-23: Composite channel .................................................................................................................. 23
Figure 1-24: Isolines in ducts and open channels (subcritical) - left=smooth/right=rough [Nezu et al.
(1989)] .......................................................................................................................................................... 25
Figure 1-25: Vectorplot in ducts and open channels(subcritical) - left=smooth/right=rough [Nezu et al.
(1989)] .......................................................................................................................................................... 25
Figure 1-26: Classification of secondary currents [Nezu et al. (2005)] .................................................... 26
Figure 1-27: Flow characteristics in an open channel junction [Weber et al. (2001)] ............................. 27
Figure 1-28: Surface velocity pattern (for q*=0.250) from the Iowa experiment [Weber et al. (2001)] 28
Figure 1-29: Schematic of flow structure for q*=0.250 from the Iowa experiment [Shumate (1998)] .. 28
Figure 1-30: Location of the cross-sections and verticals [Weber et al. (2001)] ...................................... 29
Figure 1-31: Secondary flow patterns for q*=0.250: (a) x*=-1.33; (b) x*=-2.00; (c) x*=-5.00, [Shumate
(1998)] .......................................................................................................................................................... 30
Figure 1-32: Diversion flow structure [Barkdoll (2001)] ........................................................................... 30
Figure 1-33: Schematic of flow structure: (a) q*=0.250 en (b) q*=0.750 from the Iowa experiment
[Shumate, 1998] ........................................................................................................................................... 31
Figure 1-34: Water surface mapping for different discharge ratios [Shumate (1998)] ........................... 32
Figure 1-35: Dimensions of the separation zone as a function of the friction coefficient [Creëlle et al.
(2014)] .......................................................................................................................................................... 32
Figure 1-36: Streamlines and water surface mappings for different junction angles [Huang et al.
(2002)] ......................................................................................................................................................... 33
Figure 1-37: Bed discordance (a) 90° Tributary Step; (b) 45° Step Tributary Step [Biron et al. (1996)] 35
Figure 1-38: Principal of flow instability .................................................................................................... 37
Figure 1-39: Location of shear planes at the confluence ........................................................................... 37
Figure 1-40: Kelvin-Helmholtz instability - (a): Situation sketch (b): Initiation of K-H vortex ............. 38
Figure 1-41: Kelvin-Helmholtz instability visible in clouds ....................................................................... 38
Figure 1-42: Curved free shear layer [Liou (1993)] .................................................................................... 39
Figure 1-43: Variation of growth rate α in function of the frequency σ [Liou (1993)] ............................. 40
Figure 1-44: Sketch of the mixing layer in an open channel junction [Mignot et al. (2013)] .................. 41
Figure 1-45: (a) TANH velocity profile; (d) SECH velocity profile; created either by (b)(e) varying
and h = constant or by (c)(f) varying h and = constant [Chu (1991)] .................................................. 43
Figure 1-46: Physical interpretation of Rayleigh’s inflection-point theorem ........................................... 45
Figure 1-47: Velocity and vorticity profiles to demonstrate Fjørtoft criterion ......................................... 45
Figure 2-1: Layout of the test flume ............................................................................................................ 49
Figure 2-2: Filter at inlet channels ............................................................................................................. 50
Figure 2-3: Downstream boundary condition, weir .................................................................................. 50
Figure 2-4: Discharge measurement devices – left: electromagnetic flow meter, right: gauge for the
weir ............................................................................................................................................................... 50
Figure 2-5: Setup of the LSSPIV system ..................................................................................................... 53
Figure 2-6: Principle of LSSPIV – (a): Image recording, (b): First image t=t0, (c): Second image
t=t0+∆t .......................................................................................................................................................... 54
Figure 2-7: LSSPIV coordinate system with the position of the camera .................................................. 55
Figure 2-8: Overlap between subsequent LSSPIV measurements ............................................................ 55
Figure 2-9: Reflection of light during LSSPIV ............................................................................................ 57
Figure 2-10: Close-up of the transmitter and the four beams of the ADV ................................................ 57
Figure 2-11: ADV - positive direction of Bisector (not on scale) ............................................................... 58
Figure 2-12: Measurement range ADV 30-70mm (not on scale) .............................................................. 58
Figure 2-13: Measurement grid - indication of upper boundary of measurement profile ....................... 59
Figure 2-14: Plot of Göring and Nikora Despiking criterion (2002) ........................................................ 60
Figure 2-15: Location of the ADV – cross-sections .................................................................................... 63
Figure 2-16: Inflow section tributary at y*=-1.92 (concrete flume) – Present study ............................... 65
Figure 2-17: Velocity profiles in tributary (y*=-1.92, x*=0.49) – (a): Present study (after), (b): Present
study (before) (c): Schindfessel et al. .......................................................................................................... 66
Figure 2-18: Location grass cover .............................................................................................................. 68
Figure 2-19: Thickness grass cover when compressed by ADV ................................................................ 68
Figure 2-20: Velocity Profile - Sweet spots - Left: Concrete / Right: Grass ............................................. 69
Figure 2-21: Vector plot LSSPIV (concrete) - The dotted line indicates points where u=0..................... 72
Figure 2-22: Vector plot LSSPIV (grass) - The dotted line indicates points where u=0 .......................... 72
Figure 2-23: Contour plot of total velocities (LSSPIV) - left: Concrete / right: Grass ............................. 73
Figure 2-24: Contour plot of the cross-sectional velocities (LSSPIV) - left: Concrete / right: Grass...... 73
Figure 2-25: Cross-section x*=0 – left: Concrete / right: Grass ............................................................... 74
Figure 2-26: Cross-section y*=0 – left: Concrete / right: Grass ............................................................... 74
Figure 2-27: Cross-section x*=-0.5 – left: Concrete / right: Grass ........................................................... 75
Figure 2-28: Cross-section x*=-1.33 – left: Concrete / right: Grass ......................................................... 75
Figure 2-29: Upstream velocities in the recirculation zone – left: concrete / right: grass ...................... 76
Figure 2-30: Velocity Gradient - Mixing layer - Left: Concrete / Right : Grass ....................................... 77
Figure 2-31: Cross-section x*=-0.75 - flume with grass cover ................................................................... 81
Figure 2-32: Cross-section x*=-0.5 - flume with grass cover .................................................................... 81
Figure 2-33: Cross-section x*=-0.25 - flume with grass cover .................................................................. 81
Figure 2-34: Divergence x*=0 - x*=-1.0 - Left: Concrete / Right: Grass ................................................. 83
Figure 2-35: Divergence at separation zone – left: Concrete / right: Grass ............................................ 84
Figure 2-36: Indication of shear planes ...................................................................................................... 85
Figure 2-37: Frame of the existing confluence where the shear planes are visualized with tracers ....... 85
Figure 2-38: Frame of the grass configuration from x*=-0 to x*=-0.66 ................................................. 86
Figure 2-39: Frame of the grass configuration from x*=-1.02 to x*=-1.69 ............................................. 86
Figure 2-40: Frame of the grass configuration from x*=-1.50 to x*=-2.16 ............................................. 86
Figure 2-41: Vorticity - mixing layer – left: concrete / right: grass........................................................... 87
Figure 2-42: Vorticity –Separation zone – Left: Concrete / right: Grass ................................................ 88
List of Tables
Table 1-1: Base values of Manning’s n for natural channels [Arcement et al. (1989)] ............................... 5
Table 1-2: Base values of Manning’s n for artificial channels ..................................................................... 5
Table 1-3: Comparison between present study and the experiments of Shumate (1998) ....................... 29
Table 1-4: Water-surface elevation in function of junction angle θ .......................................................... 34
Table 1-5: Velocity and vorticity profiles to illustrate the Fjørtoft criterion............................................. 46
Table 2-1: Parameters of the experimental setup ...................................................................................... 51
Table 2-2: Locations of the LSSPIV measurements ................................................................................... 56
Table 2-3: Mean and standard deviation of reference measurements - Concrete flume ......................... 62
Table 2-4: Mean and standard deviation of reference measurements - Grass flume .............................. 62
Table 2-5: Location of the ADV – cross-sections ....................................................................................... 63
Table 2-6: Reliability measurements .......................................................................................................... 67
Table 2-7: Location shear layer vs recirculation area ................................................................................ 82
List of Graphs
Graph 2-1: Mean velocities at x*=-1.33 – z=41cm (concrete) ................................................................... 70
Graph 2-2: Standard deviation of the velocities at x*=-1.33 – z=41cm (concrete) .................................. 70
Graph 2-3: Mean velocities at x*=-0.50 (grass) ..........................................................................................71
Graph 2-4: Velocity U1 and U2 - concrete vs grass flume ......................................................................... 78
Graph 2-5: Total Velocity Profile x*=-0.25 - Grass flume ......................................................................... 78
Graph 2-6: ΔU - Concrete vs Grass flume .................................................................................................. 79
Graph 2-7: Maximum Velocity Gradient - Concrete vs Grass flume ........................................................ 79
Graph 2-8: Width δ of the mixing layer - Concrete vs Grass flume ......................................................... 80
Graph 2-9: Total Velocity Profile x*=-1.33 ................................................................................................. 89
1
Introduction
Open channel confluences are frequently encountered in hydraulic structures and in nature
(e.g. a river with a main channel and a lot of tributaries) and form the nodes of the riverine network.
The different flow features often induce important morphodynamic changes that have a significant
effect on the maintenance of these channels. At certain locations sedimentation can occur and other
locations may be affected by scour. Both processes may influence the general flow pattern and alter the
navigability of the confluence.
Despite the fact that a lot of research has been carried out to clarify the flow patterns that occur at a
confluence, the effect of some determining variables is only partially known. On the one hand, the
influence of the geometric characteristics of the channels or the angle at which they meet has already
been investigated. On the other hand, not many studies have been carried out to quantify the influence
of roughness on the flow features at a confluence. In this thesis, experimental research was performed
to clarify some aspects of this influence at a 90° open channel confluence.
Over the past years, a lot of formulas have been derived to describe flow resistance. Some formulas
consider other parameters of the channel in which the fluid flows. Other formulas claim to be more
accurate than some earlier derived equations. Due to all the research that was carried out over the last
decades, it is not clear to see which one is more accurate and where the distinction has to be made
between different situations.
In the following chapter, an attempt will be made to explain some general concepts on roughness and
resistance. After explaining the concept of resistance, the most common equations to express this
resistance will be mentioned. An attempt is made to give some explanation on the resistance in
channels with different roughness. Not all types of surfaces are examined, but the most common
situations are explained. Afterwards, the most important components of a confluence are highlighted
and the possible effect of some variables on the flow pattern will be described. The studies performed
by Shumate et al. (1999) and Schindfessel et al. (2014) are used as a benchmark in the present study.
By comparison with their results, the consistency of the obtained information will be verified.
Furthermore, some general concepts on flow (in)stability will be explained to clarify the occurrence of
instability at the confluence.
With the knowledge of the previous chapter, experimental research is carried out in a test flume at the
Laboratory of Hydraulics in Ghent. Data will be gathered with two different techniques. The main goal
is to verify whether the additional bed roughness influences the flow features at the confluence or not.
With additional measurements the overall result on the flow pattern is estimated and the effect on
different components of the intersection is highlighted. From a comparison of the obtained results
with the data obtained at the concrete set up some distinct changes in flow features are noticed that
partially disagree with some earlier executed studies.
3
1. Literature review
1.1. General concepts
In general, a distinction has to be made between the terms ‘roughness factor’ and ‘resistance
coefficient’. Both terms are not interchangeable, so a clear definition of both words is necessary:
- The resistance coefficient is used to describe and quantify the dynamic behaviour of the
boundary which resists the flow of the fluid. This is expressed in terms of momentum or
energy slopes. This resistance is primarily caused by turbulence developed by surface
properties, geometrical boundaries, obstructions and other factors.
- The roughness factor is a value related to the actual unevenness of the boundary.
In fluid mechanics, hydraulic resistance can be described as the force to overcome the action of the
rigid, flexible or moving boundary on the flow. This can be separated into different components
depending on the writer. In 1964, Leopold et Al. subdivided the resistance into a component due to
skin friction, a component due to internal distortion and one due to spills [1]. Yalin (1977) separated
the resistance into skin roughness, sand wave roughness and resistance due to suspended sediment
[2]. In 1965, Rouse tried to review the hydraulic resistance in open channels [3]. He started by making
a distinction between the following components of flow resistance:
- Surface or skin friction,
- Form resistance or drag,
- Wave resistance from free surface distortion,
- Resistance associated with local accelerations or flow unsteadiness.
From all the research that has been carried out, it appears that there is one basic element of flow
resistance, and that is the one due to wall surface roughness. Depending on the writer, this element
may be completed with the components from Figure 1-1.
Figure 1-1: Components of flow resistance
4
Form drag resistance occurs as a result of the geometry of the channel. Due to the bending of the river,
the flow has a tendency to form vortices which cause a resistance to the flow. Form drag also occurs
due to elements that are present in the flow (e.g. resistance due to surface geometry, etc.).
In this research, the resistance will be mainly attributed to the surface or skin friction. The test flume
will be equipped with a different bottom type which has another value for the roughness coefficient.
Furthermore, the effects on the flow pattern will be described.
Manning, Weisbach & Chézy coefficients 1.1.1.
The most common used formulas quantifying open channel flow resistance were developed by
Manning, Darcy-Weisbach and Chézy, each relating the cross-sectional velocity U to a certain
coefficient.
-
(Manning)
- √
√ (Darcy-Weisbach)
- √ (Chézy)
With:
- = Hydraulic Radius [m]
- S = Slope of the hydraulic grade line [-]
- √
These coefficients are called the Manning, Weisbach and Chézy coefficients n, f and C and have
different dimensions. These equations can be related to each other by equalizing the velocities in the
previous equations. This makes all three coefficients interchangeable.
√
√
√
√
( 1.1 )
This confirms the presumption that there is not one perfect formula to describe flow resistance and
that a deeper look into the matter is necessary.
From a practical point of view, it is clear that every coefficient has its advantages and disadvantages.
For example, for completely developed turbulent flow over a rigid rough surface, the Manning
coefficient has the advantage of being nearly independent over flow depth h , Reynolds’s number or
the relative roughness
. Chézy is the oldest and most simplified formula and Weisbach’s f is directly
related to the development of fluid mechanics. The disadvantage of Chézy is that no clear table or
figure that lists Chézy coefficients exists.
As the Manning formula is most frequently encountered, this one will be used in practice. To have an
idea on the magnitude of this coefficient, Table 1-1 and Table 1-2 give an overview of the values for
natural channels and artificial channels. Adding some material to the bottom can easily alter these
coefficients so the listed values are only approximately correct.
5
Table 1-1: Base values of Manning’s n for natural channels [Arcement et al. (1989)]
Material Median
size [mm]
n- channel
straight/uniform smooth
Sand
0.2 0.012 -
0.3 0.017 -
0.4 0.02 -
0.5 0.022 -
0.6 0.023 -
0.8 0.025 -
1.0 0.026 -
Concrete - 0.012-0.018 0.011
Rock Cut - - 0.025
Firm Soil - 0.025-0.032 0.200
Coarse Sand 1.0 tot 2.0 0.026-0.035 -
Fine Gravel - - 0.240
Gravel 2 tot 64 0.028-0.035 -
Coarse Gravel - - 0.260
Cobble 64-256 0.030-0.050 -
Boulder >256 0.040-0.070 -
Table 1-2: Base values of Manning’s n for artificial channels
Channel Surface n
Asbestos cement 0.011
Asphalt 0.016
Brick 0.015
Cast-iron 0.012
Concrete 0.012
Copper 0.011
Masonry 0.025
Metal – corrugated 0.016
Glass 0.010
Plastic 0.009
Lead 0.011
Wood 0.012
6
Moody diagram 1.1.2.
The first concepts about open channel flow resistance were based on knowledge of the resistance in
straight axisymmetric circular pipes of steady uniform flow. This resistance can be derived from the
Moody diagram for rigid surfaces. This diagram is composed of three linear curves (on a logarithmic
scale) which depend on the Reynolds number where each line is only valid between certain values of
this number. For closed circular pipes, there are the following three known equations to calculate the
resistance in a steady uniform flow.
-
- :
Blasius
- :
√
√ White Colebrook
These equations can also be used for open channel flow resistance with different geometries of the
channel section. A provisional plot of the equations for a certain channel shape is shown in Figure 1-2.
With a different geometry of the channel section comes another location of the curves. The values of
K1, K2, K3 and KL depend on this geometry and the writer.
For large boundary roughness (
), a deformation of the free surface and modified flow patterns
occurs. To this end, the relative roughness of the channel needs to be taken into account. This will have
a higher influence on narrow channels than on wide channels.
Figure 1-2: Moody Diagram for open channels with impervious rigid boundary [Ben Chie Yen (2002)]
7
Momentum vs energy resistance coefficients 1.1.3.
In §1.1.2, the flow resistance was expressed in function of the energy slope S as is often done in open
channels. With this parameter, flow resistance can be reviewed in two ways. The momentum concept
considers the resistance as the resultant of the forces against a flow. This flow acts on the boundary of
a control volume. The momentum slope then results from external forces that are working on the
boundary of the control volume and is not directly related to the flow inside the control volume/cross-
section. The slope of the momentum resistance along the direction of a channel cross-section is
expressed as:
∫
( 1.2 )
With:
- = specific weight of the fluid
- σ = surface of the boundary
- A = flow cross-sectional area normal to direction bounded by σ
- = directional normal of σ along
- (
)
= shear stress acting on σ
- = turbulence fluctuation with respect to the local mean velocity component
In this approach, the wall surface resistance is reviewed by using the boundary layer theory and the
velocity distribution which will be clarified later. Another way of considering the flow resistance is by
using the energy concept. The energy slope is the energy lost as the fluid moves across the boundary
and can be expressed as:
∫
( 1.3 )
This actually represents the work done by the flow against internal forces. These forces are created by
molecular viscosity and eddy viscosity which overcomes the flow velocity gradient.
Both approaches differ a lot. The energy slope is a scalar quantity while the momentum slope is a
vector quantity. Furthermore, momentum slope resistance is a surface integral or a line integral while
the energy slope resistance is an area integral or a volume integral. So both terms cannot be mixed up.
The exception is steady uniform flow in a straight prismatic channel with a rigid impervious wall and
without lateral flow. Here the channel slope, so both methods lead to the same value
for the resistance. Based on the previous approaches, also the Weisbach, Manning and Chézy
coefficients can be expressed by using these concepts as , in case of the energy concept and
in case of the momentum concept.
In general it is almost impossible to obtain accurate information on the momentum and energy
resistance coefficients except for uniform flow in pipes or 2D channels where the Moody diagram is
available. So the formulas above are not frequently used, but the concepts remain important.
8
Energy Dissipation 1.1.4.
The main question in this paragraph will be: “What happens with the energy?”. When the flow moves
across the channel, it has a certain kinetic energy and a certain velocity. But due to the surface
roughness, the obstacles in the flow, etc., this energy is dissipated and transformed into heat.
At certain locations in the flow, depending on the characteristics of the channel, the flow passes
irregularities. Because of this, turbulence is created in the flow. According to Kolmogorov’s theory of
1941, these vortices are not stable and split up in different smaller vortices [4]. These smaller vortices
are not stable either and split up again. This process continues until these vortices are small enough to
dissipate their energy through friction caused by their viscosity. This is the principle of the energy
cascade which is presented in Figure 1-3.
Figure 1-3: Energy cascade [P.A. Davidson (2004)]
At each moment, there is a continuous cascade of energy from large vortices to smaller ones. However,
the viscosity does not interfere so the whole process is essentially driven by inertial forces. The process
continues until the Reynolds number, which is based on the size of the smallest vortices, is in the order
of unity. At this point, the viscous forces become important and the dissipation of energy starts.
Through friction caused by the viscosity, the small vortices transform their energy into heat.
The exchange of momentum caused by turbulent eddies is expressed by the eddy viscosity and takes
the internal flow friction into account. It works in a similar way as the molecular kinematic viscosity
does for laminar flow. Eddy viscosity is often written as and is a function of the flow, not of the fluid.
The bigger the turbulence, the higher the eddy viscosity.
9
1.2. Resistance
Boundary Layer Theory 1.2.1.
The boundary layer theory is an important method for the determination of the surface friction. The
theory suggests that when a flow touches a surface, a certain velocity distribution is generated which is
zero close to the wall. This is due to the surface roughness which slows the water particles along the
surface down. These particles slow the particles next to them down until the effect of the wall
roughness is negligible. The width of this boundary layer is denoted with δ and increases with
increasing viscosity and decreasing Reynolds number (Figure 1-4). So the faster the stream flows, the
larger the stream velocity and the larger the Reynolds number. Outside this layer, the velocity
distribution has an almost constant value over the remaining water depth h. A common definition of
the thickness of this layer is that it is the magnitude of the normal distance from the boundary surface
untill the plane at which the velocity is equal to 90% of the limiting velocity . This velocity is
also the velocity at the surface of the flow since the velocity distribution outside the boundary layer is
uniform.
Figure 1-4: Boundary layer δ and laminar sublayer
This boundary layer can be subdivided in two main regions. One where the viscosity is negligible and
one close to the wall where the viscosity has to be taken into account. Both distributions are described
by two universal laws. The inner law where the viscous effect dominates and the outer law where the
viscosity is less important.
( 1.4 )
(
) ( 1.5 )
With:
- √
-
-
-
-
-
10
Beside these two regions, there also exists an overlap as shown in Figure 1-5. Here, both laws are
applicable and a lot of equations have been developed which satisfy both distribution laws, e.g. a
logarithmic function or a power law distribution. In general the logarithmic function from equation 1.6
will be applied. The validity of this function has already been verified by experimental results.
⟨ ⟩
(
) ( 1.6 )
Figure 1-5: Inner and outer law of boundary layer
Within this boundary layer, there is also a very thin stable film of flow present within this boundary
layer, which is called the laminar sublayer. This layer is denoted with and the flow within this layer
stays laminar. The definition of this layer is important to make a distinction between hydraulically
smooth and rough surfaces. When the surface of a channel is enlarged, it is clear that the surface
consists of an irregular pattern of peaks and valleys. Their height is denoted with the variable k and is
called the roughness height and
the relative roughness.
Figure 1-6: Hydraulically smooth vs. hydraulically rough
When the height of the peaks is not larger than the thickness of the laminar sublayer , the surface is
called hydraulically smooth. The height of the peaks has no effect on the flow outside the laminar
sublayer. When the peaks become larger, the flow outside this layer will feel these peaks and will have
an influence on this flow. In this case, the surface is called hydraulically rough. In these channels the
rough surface will disturb the flow and a stable laminar sublayer can no longer be formed.
11
The velocity distribution over the boundary layer is related to the flow resistance by using equation 1.7
proposed by Saint-Venant. Stokes (1845) already suggested a similar equation which links the
tangential shear stress to the molecular dynamic viscosity µ and the velocity gradient but this
equation is more general.
(
) ( 1.7 )
With:
-
This equation is applicable to laminar flows when ε equals the dynamic viscosity µ. In this case, the
equation of Stokes equals the proposed equation of Saint-Venant. It can be used to describe turbulent
flow when ε includes the molecular dynamic viscosity and the turbulent viscosity.
The previous explained boundary layer theory gives a clear relation between the velocity distribution
and the shear stress of the wall. However, there are not only rigid impervious surfaces present on the
boundaries of an open channel. Figure 1-7 shows a possible classification of channel boundaries for
steady flows. A different boundary also involves another velocity distribution and another shear stress
on the wall.
Figure 1-7: Channel boundary classification [Ben Chie Yen (2002)]
12
Nonalluvial channel 1.2.2.
The nonalluvial boundary is a surface that cannot be moved or detached from its place. It can be rigid
like for example concrete or it can be flexible like some channels with flexible synthetic fabrics. Both
boundaries can be impervious or pervious. This is important because the porosity of the boundary also
determines the velocity profile at the bottom. The difference between hydraulically smooth and rough
surfaces has already been explained in the previous section. When the roughness becomes larger, for
example with increasing rock size in the water, the resistance cannot be explained anymore with the
assumption of a hydraulically rough or smooth boundary. In this case, other phenomena are present.
Chow (1959) uses the concepts defined by Morris to give an explanation for the loss of energy over
rough elements [5]. He assumes that the loss of energy in turbulent flow over a rough surface is largely
due to the formation of wakes behind each roughness element. The amount of vortices that are present
in the direction of the flow will determine the turbulence and energy-dissipation phenomena in the
flow. To this end, the spacing between these elements is important. By using this concept, flow over a
rough surface can be classified in three basic types which are shown in Figure 1-8, the k-type, d-type
and transitional type. This concept can also be reduced to rough surfaces when the dimensions of the
roughness elements are taken infinitesimally small.
Figure 1-8: Types of rough-surface flows: (a) k-type; (b) d-type; (c) transitional-type [Chow (1959)]
The k-type corresponds to an isolated element in the flow because the distance between the elements is
so large that there will not be any interaction between them. The turbulence over these elements is
dominated by the influence of the separation zone and vorticity generated along the boundary shear
layer. The main flow detaches from these elements and reattaches further. From this point, an internal
boundary layer develops (Figure 1-9).
13
Figure 1-9: k-type roughness – mechanism [C. Polatel (2006)]
For the d-type, the elements are so close together that the flow passes the top of the elements. The
holes between the elements contain stable eddies with dead water. The flow will not feel the grooves
and the roughness of the elements, so the surface will be almost hydraulically smooth.
Chow (1959) calls this a quasi-smooth flow. However, a distinction has to be made with hydraulically
smooth surfaces because the eddies in between will still consume a certain amount of energy.
The transitional flow is more difficult. The roughness elements are positioned with a smaller spacing
than k-type roughness but with a larger one than d-type roughness. As a consequence, the wake and
vortex of each element will interfere with those of the following element so a complex interaction takes
place. Here, the height of the elements will be of minor importance.
Alluvial channels 1.2.3.
Alluvial channels have a surface that can be moved due to the flow. The sediment particles have the
ability to be set in motion. Water can then move through the voids between the particles of the bed.
These sediments can cause resistance to the flow because the water puts energy in the movement of
these particles. The latter can occur by rolling or saltation (Figure 1-10).
Figure 1-10: Possible movements of particles near the bed
The resistance of alluvial channels can be split up in two parts. A plane bed that has no geometrical
elements on the bed and a bed form part. The bed form can consist of four important forms. The plane
bed resistance is almost similar to a rigid channel because the source of the resistance is again surface
resistance. However, there are two differences. Firstly, the water has to spend energy in alluvial
channels on rolling and picking up of particles. Secondly the flow, contrary to an impervious boundary,
will move through the voids between the particles. So the boundary is not impermeable. When the
velocity increases or the depth of the flow decreases, the Froude number increases and ripples can be
formed. The sources of this resistance are surface resistance and form resistance so an increase in
resistance coefficient can be seen. For channels with dunes and antidunes also wave resistance has to
be taken into account so an extra resistance is generated.
14
Figure 1-11: Forms of bed roughness in sand-bed channels [G.J. Arcement (1989)]
Due to the presence of these bedforms, turbulence will occur between the crests of each dune. The
same principle can be used as in non-alluvial boundaries, only the square-shaped elements are
replaced by dunes, ripples or antidunes. The bed form can change in an unsteady flow so the resistance
of the bed might also change in time.
When the channel is sufficiently wide and the sediment transport is in equilibrium, the resistance
coefficient f can be simplified so it only depends on four independent variables.
(
)
With:
- = representative size measure of sediment
The alluvial channel resistance can be separated, in a linear way, in two parts so that every part only
depends on two or three of the previous variables.
( 1.8 )
With:
-
-
In this way, the problem of determining f with four independent variables is simplified. This
superposition method was also used for the determination of the bed shear .
( 1.9 )
In the nonlinear approaches, the resistance coefficient is kept as a single factor.
15
1.3. Vegetated Channels
Velocity Profiles 1.3.1.
A lot of variation is possible in vegetated channels. It can be submerged or unsubmerged, flexible or
rigid, etc. All these different configurations with their corresponding surface resistance will modify the
velocity profile of the flow. Depending on the height of the vegetation and of the flow, three different
cases can be distinguished: emergent, submerged and well-submerged vegetation.
Flow over well-submerged vegetation occurs when the water depth h is much higher than the
vegetation height k. The velocity of the flow is delayed within the vegetation, but the flow that goes
over the vegetation is not blocked. When the water depth is high enough, the velocity profile becomes
logarithmic as was the case for a non-alluvial boundary. According to Kleinhans (2008) the vegetation
is well submerged when the water depth is a lot higher than five times the vegetation height [6]. The
situation of well-submerged vegetation is shown in Figure 1-12.
( 1.10 )
Figure 1-12: Velocity profile - well-submerged vegetation [Galema (2008)]
When the water depth is less than five times the vegetation height, but still larger than its height, the
vegetation is submerged (Figure 1-13). In this case the logarithmic velocity profile disappears. The flow
begins to feel the vegetation and the following velocity profile occurs:
Figure 1-13: Velocity profile – submerged vegetation [Galema (2008)]
Previous studies claimed that a two zone sub model is valid. In these models, the velocity profile above
the vegetation had a different path than the one in it. Above the vegetation, the well-known
logarithmic distribution is valid. Within the vegetation, an exponential velocity profile is present.
However, Sanjou and Nezu (2008) proved that this model was not valid [7]. Instead, a three zone sub
16
model with an inflection point has to be applied in vegetated channels. Figure 1-14 illustrates this
concept.
Figure 1-14: Three zone sub model [Nezu and Sanjou (2008)]
In this model, three subzones can be distinguished, the wake zone, the mixing-layer zone and the log-
law zone. The wake zone extends from the bottom of the channel to the penetration height . The
location of this point depends on the way the energy is dissipated in this zone, but this will be
discussed later. Within this zone, the velocity profile is almost constant over the depth.
⟨ ⟩ ( 1.11 )
From till the mixing-layer zone is defined. In this zone the velocity profile is a tanh-function
and is defined by the following equation:
⟨ ⟩
(
) ( 1.12 )
With:
∫ (
(
)
)
( 1.13 )
From to H the velocity profile follows again the logarithmic function that was already defined for
non-alluvial channels. These three profiles are chosen in a way that they all converge into one velocity
profile over the water depth.
Nezu and Sanjou (2008) discovered that this last zone disappears when the submergence depth
becomes smaller than a certain critical depth . When the flow depth is smaller than , only the
wake zone and the mixing layer zone remain present. From previous research, it seems that for rigid
vegetation the value
lies between 1.5 and 2. Simulations also indicated that when the submergence
depth decreases towards the critical height , values of U in the mixing layer deviate more from the
suggested profile. This implies that the coherent eddies generated in this layer are also influenced by
this decrease, especially in shallower flows where .
When the water depth is reduced, the vegetation will not be completely submerged, it will become
emerged for h<k. In the previous case of unsubmerged vegetation, it was shown that the velocity
profile was uniform over the biggest part of the vegetation. However, close to the bed, the velocity
becomes smaller due to the bottom roughness. The emergent case can be seen as a simplified case of
17
the unsubmerged vegetation. The water flow over the vegetation is diminished so only the flow
through the vegetation remains. From the previous part, it can be seen that the most efficient way to
describe the flow resistance in vegetated channels is by investigating the submerged case.
Figure 1-15: Emerged vegetation [Galema (2008)]
Energy exchange 1.3.2.
By investigating the energy dissipation in the submerged case, a decent explanation for the velocity
profile can be obtained. The part below the penetration depth is called the wake zone. The reason
for this name is that the turbulent transport in this zone is mainly due to the strong wake effects
behind the vegetation stems. This means that the Karman vortex appears significantly in the wake
zone and governs the longitudinal transport of momentum. The turbulence created by these stem
wakes scales significantly with the plant morphology, this is an important factor to consider. Nepf and
Vivoni (2000) called this zone the longitudinal exchange zone [8]. The vertical transport of
momentum or energy is negligible in this zone. The latter is comparable to the emergent vegetated
open channel flow. Within this zone, the flow depends on a balance between vegetative drag and the
pressure gradient.
The value of depends on the author, e.g. Nepf and Vivoni (2000) defined it as the depth at which
the Reynolds stress has decayed to 10% of its maximum value. On the other hand, Ghisalberti and
Nepf (2006) said that the value at which has decayed to zero, is the penetration depth [9].
Hence different definitions of this zone remain available. They also pointed out that the value of the
penetration depth changes when the water depth increases but becomes a constant when the critical
water depth is reached. The contribution from the bottom roughness to the total resistance in
vegetated channels is negligible because the viscous stress and the bed drag appear to be almost
negligible compared to the vegetative drag.
18
Figure 1-16: Energy exchange model – vegetated channels [H.J.S. Fernando (2013)]
The zone above is called the vertical exchange zone. Within this zone, the vertical exchange is more
pronounced than the horizontal one. The vertical discontinuity of the drag force results in a strong
velocity shear in the mixing layer zone. Raupach et al. (1996) discovered that due to inflection point
instability or Kelvin-Helmholtz (K-H) instability, coherent vortices such as sweeps and ejections are
generated as shown in Figure 1-17 [10]. This instability governs the mass and energy transport between
the zone above the vegetation and the zone between the vegetation. The shear layer in the mixing-layer
zone can be illustrated by plotting the shear stresses over the depth as shown on Figure 1-17.
Figure 1-17: Effect of submergence depth on Reynolds stresses [Nezu et al. (2008)]
Due to the large velocity gradient in this layer, there is a significant increase in shear stresses. For the
emergent case the shear stresses are much smaller because there is no shear layer present and no
vertical transport of momentum can occur. This figure indicates that larger shear stresses at the
boundary of the vegetation layer penetrate deeper in this layer so the vertical exchange of momentum
occurs closer to the bottom. This explains also the name of the penetration depth, the depth until
where a significant vertical exchange of momentum takes place.
19
In the mixing layer zone coherent vortices appear due to K-H instability. An explanation for the
exchange of momentum that goes on in these vortices can be given by using the Quadrant Conditional
Analysis of Nakagawa and Nezu (1977) [11]. They made an analysis of the instantaneous Reynolds
stresses. The quadrant Reynolds stress was defined as follows:
∫
( 1.14 )
When u(t) and v(t) exist in a certain quadrant, I(t) equals unity. Otherwise it is zero. Each quadrant
corresponds to a certain event.
- -
-
-
The sum of the Quadrant Reynolds stresses of all quadrants equals unity. These events are illustrated
in Figure 1-18.
Figure 1-18: Quadrant conditional [Nezu and Nakagawa (1977)]
By plotting the contribution of each of these motions over the depth of the flume, a plot similar to
Figure 1-19 can be obtained. This indicates which mechanism is more important than the other over
the depth of the flume. Ejection motions are motions in which low speed fluid is lifted up towards the
free surface and sweep motions are motions in which high speed fluid is inrushed into the vegetation
layer. From simulations, it appeared that both motions occur periodically. Each sweep is followed by
an ejection and these coherent motions are associated with large-scale secondary flows.
Figure 1-19: Circle C: Sweep – Circle B: Ejection [Nezu and Nakagawa (1977)]
20
Flexibility 1.3.3.
In the previous paragraphs, the flexibility of vegetation was not included. This has a large impact on
the resistance in the channel because the introduction of flexibility results in a reduced vegetation drag
coefficient. This is because the deflected height of the plant is smaller than the undeformed height.
Figure 1-20: Flexible vegetation compared to rigid vegetation [Carollo et al. (2005)]
In 1973, Kouwen and Unny proposed a resistance law for flexible vegetation [12]:
√
(
)
( 1.15 )
( 1.16 )
With:
- = mean velocity,
- = friction velocity,
- h = undeflected height,
- = time-averaged deflected height,
- m = number of stems per unit area,
In equation 1.16, the aggregate stiffness ‘mEI’ is the most suitable parameter to describe flexibility of
vegetation. On the basis of this property, vegetated flexible open channels are classified into the
following four types:
- Erect or Rigid type: in this case, the vegetation elements are erect and do not change their
tip position in time.
- Swaying type: here, the vegetation elements are independently waving without organized
plant motion.
- Monami type: the vegetation elements wave in an organized way in response to a coherent
vortex.
- Prone type: the vegetation elements are prone due to a large drag force.
21
Figure 1-21: Flow patterns in dense flexible vegetation. (a) Erect or rigid. (b) Swaying (no organized). (c)
Monami (organized). (d) Prone. [H.J. (2013)]
The most important one of the previous types is the Monami type. The latter should be distinguished
from the non-organized motion according to Okamoto and Nezu (2009). They pointed out that with
increasing vegetation density, a K-H instability occurs near the vegetation edge and a large-scale
coherent vortex is generated in the mixing layer. For aquatic flexible vegetation, this coherent vortex
will be the cause of the organized plant motion called Monami. Furthermore, due to bending of the
vegetation, the penetration thickness for rigid is larger than for flexible vegetation. As a result, the
transfer of momentum will be less pronounced in the case of flexible vegetation. Wilson et al. (2003)
suggested that the bending of the vegetation surface was the cause of less momentum exchange [13].
The evolution of the resistance coefficient from unsubmerged to submerged vegetation is illustrated in
Figure 1-22. Here, the results of a simplified experimental test that was carried out by Fu-Chun Wu et
al. (1999) is shown [14]. In this analytical model, the flume was covered with separated stems. The
model is a rough approximation so the actual values of the resistance coefficient cannot be applied, but
the evolution of the coefficient stays the same. This is presented in Figure 1-22.
Figure 1-22: Subregions of in function of the water depth [Fu-Chun Wu et al. (1999)]
In the first subregion, the coefficient decreases since the velocity in the flow increases with increasing
depth. In the second subregion, a boundary shear is present from D =0.030 to 0.045 with D the depth
of the flow and T the thickness of the vegetation layer. This boundary shear acts as a resistance to the
free flow over the vegetation so Manning’s n tends to increase. The effect of this layer becomes less
important when the vegetation becomes more submerged so the resistance coefficient decreases again.
This model is only a simplified model but gives a general view on the processes and forces that are
present in submerged and unsubmerged vegetated channels.
23
1.4. Resistance of composite or compound channels
Composite and compound channels are very common in nature. Their roughness coefficient and the
geometric shapes of their cross-sections vary over the entire channel. A composite channel can be
defined as a channel whose wall roughness changes along the wetted perimeter of the cross-section. A
compound channel is a channel that contains different subsections with different geometric shapes.
Usually a compound channel will be a composite channel, e.g. a main channel with floodplains (Figure
1-23). Different subsections and different roughness coefficients are present along this channel.
Figure 1-23: Composite channel
The resistance coefficient of these channels can be expressed by making a weighed sum of the local
resistance coefficient with a weighing function . The cross-sectional value of the resistance
coefficient can then be calculated with the following equation.
∫
( 1.17 )
The fact that the resistance of a composite channel changes across the cross-section also has some
influence on the flow pattern and flow characteristics.
25
1.5. Secondary currents in open channels
Nezu et al. (1989) performed some experiments with regard to the influence of roughness on the
secondary currents in open channels and ducts [15]. Their first discovery was the fact that a dip was
present in the velocity distribution. The maximum velocity appeared to be absent at the free surface
but occurred at a distance of 0.6 times the water height as can be seen on Figure 1-24.
Nezu et al. (1989) proved that the secondary currents were responsible for this phenomenon.
Furthermore, they found that the velocity dip only appeared when the ratio of the channel width to the
water depth was less than five. In this case, the channel was classified as narrow.
Figure 1-24: Isolines in ducts and open channels (subcritical) - left=smooth/right=rough [Nezu et al. (1989)]
The previous figure shows the velocity distribution across half of the channel. It also shows that the
velocity in the rough test flume is more uniformly retarded near the bed. Moreover, the smooth
isolines disappear completely in the rough flume and show a more broken pattern.
Figure 1-25: Vectorplot in ducts and open channels(subcritical) - left=smooth/right=rough [Nezu et al. (1989)]
Figure 1-25 shows vector plots for the smooth and the rough test flume. There exists a strong lateral
component at the free surface in both of them that is directed towards the centre of the channel and is
26
then redirected towards the bottom. Adjacent to this vortex there appears to exist a bottom vortex that
is significantly smaller in size. This vortex is created by the anisotropy of turbulence that is caused by
the free surface effect and the presence of the side wall. With increasing roughness this bottom vortex
disappears while the surface vortex increases in size. These flow characteristics probably will not occur
at the confluence but have to be taken into account when comparing results. The fact that the flow in
one of the branches of the confluence contains a recirculation cell can also slightly be influenced by
these effects.
For curved channels, not only a turbulence-driven force is present but also a centrifugal force. To this
end, a distinction has to be made between the secondary currents that were seen in the straight
channel and the ones encountered in the curved channel. Based on the derivation of the streamwise
component of the channel vorticity, two different types of secondary currents can be described.
Secondary currents in a curved channel are created by the presence of the centrifugal force but also
slightly by the presence of the anisotropic turbulence that was spotted in the straight channel. These
are called secondary currents of the first kind. If the centrifugal force is absent, so the channel is
straight, the currents are from the second kind and the flow characteristics mentioned before are
present. Figure 1-26 points out the differences between both situations.
Figure 1-26: Classification of secondary currents [Nezu et al. (2005)]
27
1.6. Open channels
Open channel confluence 1.6.1.
Open channel confluences are very common in nature where every river consists of a main channel
and some tributaries. Since this is a very important part of the river, it has already led to numerous
studies presented in literature. In this first part, a 90° open channel confluence will be discussed. In
Figure 1-27, the most important flow characteristics of this configuration are shown, with a zone of
separation, a contracted flow region and a shear plane as most important flow regions.
Figure 1-27: Flow characteristics in an open channel junction [Weber et al. (2001)]
The flow features at this junction depend on many geometrical parameters. These include the size and
shape of the channel and also the junction angle. Flow variables such as the Froude number and the
ratio of flow discharge are also of crucial importance. This ratio is defined as:
( 1.18 )
Where and denote the incoming discharge of the main channel and the tributary respectively,
expresses the downstream discharge and is the sum of the discharge of both inlet channels.
The zone of separation is an area just downstream of the tributary, it is a place where the flow has a
small velocity and where recirculation exists. The momentum of the side flow causes the main flow to
detach at the downstream corner of the junction. Studies ([Shumate (1998)], [Best and Reid (1984)])
have shown that the separation zone is significantly longer and wider at the surface level compared to
the bottom of the channel [16] [17]. According to simulations performed by L. Crombé (2013), the
friction coefficient is also a determining parameter [18]. The dimensions of the zone of separation
decreased with increasing roughness. Furthermore, the dimensions of this zone decrease when the
flow ratio q* increases.
The recirculation zone is an area inside the zone of separation where upstream velocities occur. These
velocities decrease towards downstream. Because of the presence of this zone, the water has to pass a
narrower zone. This contracted flow region will have bigger downstream velocities because the same
amount of water has to pass a smaller area.
28
The mixing layer is the area where the fluid of the main channel meets the fluid of the tributary.
Different methods exist to measure and visualize this zone. If different temperatures are assigned to
both flows, the temperature gradient at every location can be seen. This is of course depending on the
flow ratio. Furthermore, the dimensions of the recirculation zone can be determined. Another option is
to give both currents a different concentration of an arbitrary substance. Figure 1-28 shows a surface
velocity plot from the Iowa experiment performed by Shumate (1998). The mixing layer is shown in
the red rectangle. The width of this layer is not constant over the depth of the flume. It decreases in
width from the bottom till the free surface.
Figure 1-28: Surface velocity pattern (for q*=0.250) from the Iowa experiment [Weber et al. (2001)]
Figure 1-29 shows a 3D-configuration of the test setup of the Iowa experiment performed by Shumate.
This 3D-perspective gives more insight in the location and shape of the shear layer and of secondary
currents while a 2D-look focuses more on the general flow patterns of this open channel junction. As
shown in Figure 1-28, the shear plane is not vertical but slightly tilting towards the outer wall.
Figure 1-29: Schematic of flow structure for q*=0.250 from the Iowa experiment [Shumate (1998)]
In the Iowa experiment, measurements were performed at different locations of the flume. The cross-
sections and vertical profiles of these measurements are shown in Figure 1-30. The location of the
cross-sections are dimensionless by dividing the absolute coordinates by the width of the channel
( = 0.914m).
29
Figure 1-30: Location of the cross-sections and verticals [Weber et al. (2001)]
The dimensions of the flume are comparable with the ones of the present study and will give a global
overview of the flow features that can be expected in the present experiment. The most important
characteristics are listed in Table 1-3.
Table 1-3: Comparison between present study and the experiments of Shumate (1998)
Characteristic Present study Shumate (1998)
0.04 0.17
0.104 0.628
0.384 0.271
0.98 0.914
2.36 3.09
0.05 0.37
98000 452000
yes no
In Figure 1-31, vector fields are shown that illustrate the secondary flow patterns in the flume. The flow
at x* = -1.33 (a), just downstream of the channel junction, shows the surface water approaching the
junction-opposite wall (at y* = 1.00). The lateral momentum of the flow is significantly higher near the
surface compared to the bed. Because it is not possible that the flow goes into the wall, the surface flow
is deflected to the bottom. This motion creates a secondary current that is present along the right bank
of the channel (looking downstream). A part of this secondary current, close to the bed, approaches the
junction adjacent wall and is deflected upwards into the zone of separation. This results in a secondary
current further downstream of the confluence (Figure 1-31b). Eventually, at a distance of
approximately 4.00 m downstream, the two secondary currents converge in one large clockwise
rotating current (Figure 1-31c).
30
Figure 1-31: Secondary flow patterns for q*=0.250: (a) x*=-1.33; (b) x*=-2.00; (c) x*=-5.00, [Shumate (1998)]
Comparison of confluence and diversion flow 1.6.2.
Figure 1-32 is a schematic diagram of the flow characteristics in a 90° diverging open channel flow.
When comparing the confluence and the divergence, comparable flow features are observed. The
interaction between the main and the side channel creates a zone of recirculation in both flow
scenarios, but the location changes. In confluence flow, the recirculation zone appears near the main
channel wall, just downstream of the tributary. In diversion flow, this zone occurs on the inside wall of
the tributary. Although the location of the recirculation zone is different, the physics are similar. In
both cases, this zone is wider near the surface than near the bed.
Figure 1-32: Diversion flow structure [Barkdoll (2001)]
Another similarity between both flows is a surface dividing the main channel flow from the side
channel flow. Previously, the dividing surface in confluence flow was called shear plane due to the fact
that two flows are merging, thereby setting up a shearing action as the flow from both channels is
flowing at different velocities. In diversion flow, the surface is called a dividing stream surface because
it cuts the flow in two, entering the side and main channel.
31
Another flow characteristic that is present in both configurations is secondary vortices. Similar to
confluences, the secondary vortices in diversion flow are caused by the non-vertical velocity
distribution where the surface flow strikes the side channel wall and curls down the wall, initiating a
secondary clockwise rotating vortex (looking downstream).
Flow variables 1.6.3.
Flow parameters can be changed to simulate the different scenarios that happen in nature. The
influence of the changing parameter and its size will be discussed in the following paragraphs.
A. Flow rate q*
As defined before, q* is the ratio between the discharge of the main channel to the combining flow
discharge. When this value is close to one, the contribution of the main channel is high. Figure 1-33
shows the configuration for q*=0.250 and q*=0.750. When comparing both flow structures, the
recirculation zone appears to be much larger for the case with q*=0.250. This is caused by the lateral
momentum of the side flow that pushes the shear plane to the right-hand side of the main channel
(looking downstream). Furthermore, the secondary vortices are bigger.
Figure 1-33: Schematic of flow structure: (a) q*=0.250 en (b) q*=0.750 from the Iowa experiment [Shumate
(1998)]
Another method to understand more about the behaviour of an adjustable parameter is by water
surface mapping. This is a measurement technique to visualize the dynamics of the water surface
through the channel junction region. Figure 1-34 represents the contour plots of constant water
surface elevation for two different discharge ratios (q* = 0.250 and q* = 0.750). The water level
decreases from upstream to downstream in both cases. The difference is more pronounced for
q* = 0.250 (z* = 0.034, equivalent to 30.5 mm), than for q* = 0.750 (z* = 0.019, or 17.4 mm). For all
flow conditions, the water surface generally displays a drop when entering the contracted flow region.
This drop disappears again when moving downstream as the flow expands to the entire channel width.
This behaviour is more pronounced for lower values of q*.
32
Figure 1-34: Water surface mapping for different discharge ratios [Shumate (1998)]
B. Roughness
The global roughness of the channel can be adapted by changing the bottom or wall material. Adding
materials like stones, sandpaper or artificial grass will change the friction factor. When the roughness
increases, the overall water level will increase and the sudden water level drop (like mentioned in
§1.7.4.A.) will shift to upstream.
In their paper on the influence of hydraulic resistance on flow features in an open channel confluence,
Creëlle et al. (2014) created a numerical model of a 90° confluence [19]. They adjusted the roughness
characteristics of the channel and looked what the results would be on the flow characteristics.
Figure 1-35: Dimensions of the separation zone as a function of the friction coefficient [Creëlle et al. (2014)]
Figure 1-35 indicates that the length and the width of the separation zone decrease with increasing
friction factor. The ratio of the width of the channel to the length would on the contrary increase. From
their simulations, they came to the conclusion that additional roughness on the bottom should
facilitate the diversion of the lateral inflow. This would be the cause of the decrease in width of the
separation zone. Furthermore, the local depression of the water surface that occurs just downstream of
the confluence (x*=0.0) becomes less pronounced with increasing resistance and the position of this
33
minimum shifts towards the upstream direction. This corresponds well with the observations of
Shumate (1998).
Furthermore, the simulations showed that the increase of the friction factor at the bottom of the flume
altered the velocity profiles in the different cross-sections. The downstream evolution towards a fully
developed flow seems to occur more easily and the fully developed flow is reached faster.
C. Junction angle
In this paragraph, the influence of the junction angle on the flow characteristics will be discussed.
Huang et al. (1999) made a 3-dimensional numerical study of flows based on the experimental data of
Shumate [20]. He used the data with a low discharge ratio (q*=0.250) because this resulted in a larger
separation zone and a more pronounced water-surface elevation. Figure 1-36 shows the calculated
water-surface elevations and streamlines for four different confluence angles θ.
Figure 1-36: Streamlines and water surface mappings for different junction angles
[Huang et al. (2002)]
One clearly distinguishes that increasing the junction angle leads to higher surface-elevations
upstream of the confluence. The results are listed in Table 1-4. These results are consistent with the
findings of Hsu et al. (1998) who found that the ratio of upstream to downstream depth increases with
increasing junction angle [21].
34
Table 1-4: Water-surface elevation in function of junction angle θ
Junction angle Increase of surface elevation
30° 4.3%
45° 5.0%
60° 6.1%
90° 8.8%
Moreover, the dimensions of the separation zone just downstream of the junction increase with
increasing junction angle. Figure 1-36a shows that for the 30° case, the separation zone is very small,
so for even smaller junction angles, the separation zone will almost completely disappear.
Furthermore, the depression within the separation zone increases with increasing junction angle. The
reason of these findings can be explained from the energy loss point of view. Energy loss at the
junction will increase with increasing junction angle because more lateral flow momentum needs to be
turned in the main channel direction. As a result of this bigger energy loss, less energy is available to
limit the size of the separation zone.
D. Sediment transport and erosion
Normally, a river and its tributary will not transport the same amount of sediment, if this difference is
significant, it can cause a disruption of the balance between the rivers capacity to transport sediment
and the sediment supply. The extra input can cause a deposition of sediment over the entire river
width, especially in the recirculation zone. This deposition will bend the flow to the opposing bank.
Together with secondary vortices this can lead to bank erosion.
If the additional amount of sediment from the tributary is larger than the one of the main river,
flooding can occur. Another cause that can trigger flooding is an inflow of larger sediment particles.
Because of this, the roughness will increase and consequently the water depth will decrease. If the
channel depth becomes too small flooding will occur.
E. Scale effects
Most experimental research on open channels is tested in flumes or small rivers. The limit between
small and big rivers is a width to depth ratio of approximately 200. There is not much known of how
the processes observed at these smaller sites scale to larger, wider rivers. To this end, scientists have
two main goals for the next years: in the first place, describing the bed morphology of large river.
Secondly, describing the 3D flow structure within a very large river confluence with a width to depth
ratio approaching 200 and comparing these results with the ones from smaller rivers.
35
F. Bed discordance
Bed discordance between the main and tributary channel is a common phenomenon in most river
confluences, however it is ignored in most models. Biron et al. (1996) used experimental data from a
flume with a junction angle of 30° to investigate this issue. Three different scenarios were investigated
[22]. A situation with concordant beds, one with a 90° tributary step and one with a 45° tributary step.
Those two last scenarios are sketched in Figure 1-37. Their results were focused on four important flow
regions: the zone of separation, the contracted flow region, the zone of deflection and the mixing layer
between both flows.
Figure 1-37: Bed discordance (a) 90° Tributary Step; (b) 45° Step Tributary Step [Biron et al. (1996)]
The experiments pointed out that discordance in bed height between the main channel and tributary
significantly influences the flow dynamics at open channel junctions in three principal ways. At first,
the deflection of the flow from the main channel at the bed may be suppressed. Second, the mixing
layer between the two combining flows is deflected towards the side channel and results in transport of
flow from the deeper main channel into the shallower tributary. Third, water surface differences occur
at the downstream junction corner. In this zone, flow separation is almost absent near the bed,
whereas it is present close to the water surface. The bed height discordance and the presence of a
tributary mouth step results in reduced flow acceleration in the downstream channel. These results
show the importance of bed discordance and it should be taken into account in future models of river
confluence flow dynamics.
37
1.7. Flow instability
In 1981, Drazin & Reid described hydrodynamic (in)stability as “when and how laminar flows break
down, their subsequent development, and their eventual transition to turbulence” [23]. Their
definition gives some first insight in the topic of flow instability. The question if a flow will be stable or
unstable is very important in fluid mechanics since this will determine if perturbations, which are
introduced in the flow, will be amplified or not. Basically, when a perturbation occurs in the flow, there
are two possibilities (Figure 1-38):
- The perturbation diminishes in time which results in a stable flow.
- The perturbation grows in time so instabilities can occur in the flow.
Figure 1-38: Principal of flow instability
Because of the velocity difference at the confluence between the main channel and the tributary, a
velocity gradient will originate at a shear plane. One shear plane extends from the upstream corner to
the reattachment point and a second one from the downstream corner to the end of the separation
zone. These shear planes are the main source of turbulence generation at the open channel confluence.
Figure 1-39 indicates the two shear planes that will be investigated on their stability. In the next
paragraph, the processes that govern this instability will be analysed.
Figure 1-39: Location of shear planes at the confluence
38
Kelvin-Helmholtz instability 1.7.1.
Kelvin-Helmholtz (K-H) instability is widely known and is generated in density stratified shear flows
or within a single continuous fluid where a shear flow with a strong velocity gradient is present. Two
immiscible and inviscid flows with a different density and velocity are shown in
Figure 1-40a. The lighter fluid, which has the highest velocity, is located above the heavier one. Both
streams interact with each other by a single plane, called the shear plane.
Figure 1-40: Kelvin-Helmholtz instability - (a): Situation sketch (b): Initiation of K-H vortex
The velocity and density profiles of both flows are uniform but the interface between the fluid layers is
discontinuous. This discontinuity in the transverse velocity profile is a shear layer and induces larger
waves in the vicinity of the interface. These waves continue to grow until they turn over and the typical
spiral, which characterizes the Kelvin Helmholtz instability, originates as is indicated in Figure 1-40b.
The breaking waves that generate the mixing are called billows.
An environmental example of this instability is the formation of billow clouds or Kelvin-Helmholtz
clouds. If two different air layers are propagating at a different speed they will form a wave structure.
The top of the fastest cloud layer will try to catch up the slower layer and this results in a rolling
structure (Figure 1-41). Other examples are the air-ocean interface where the density difference is of an
order three, or a river entering a colder lake or a river entering a salty ocean.
Figure 1-41: Kelvin-Helmholtz instability visible in clouds
Instability can be indicated with the Richardson number, Ri. This dimensionless number is the ratio
between the potential and kinetic energy.
(1.19 )
39
Instability at confluences 1.7.2.
Kelvin-Helmholtz instability is one of the basic cases of instability that can occur at a confluence. In
the present study, the densities are assumed to be equal so the instability is only caused by the velocity
gradient between both flows. At the confluence, the mixing layer consists of three different types that
can stabilize or destabilize the flow. The basic case is the one where two parallel flows, with different
velocities, meet in a straight flume. Furthermore, the mixing layer can be curved, accelerated or a
combination of both due to the junction geometry. In the following section, the physics of a curved and
spatially accelerated mixing layer will be explained.
A. Straight mixing layers
Straight mixing layers that do not have a pressure gradient are the easiest type of mixing layer
instability. When high velocity gradients occur, the inflection of the velocity profile can lead to Kelvin-
Helmholtz instability. In 1880, Rayleigh proved that the existence of an inflection point is a necessary
condition for Kelvin-Helmholtz instability. Because of this, coherent turbulent structures occur at the
upstream end. At this location, both flows meet and these structures increase in size and time scale.
Previous research indicated that the width of this mixing layer increases towards downstream [Bell et
al. (1990)] [24]. Furthermore, the maximum turbulent intensities and Reynolds shear stresses occur at
the centreline of the mixing layer. Starting from the upstream boundary condition, they have an initial
increase in magnitude but decrease gradually when moving downstream.
B. Curved mixing layers
In this case, Kelvin-Helmholtz instability can occur when the inflection point in the velocity profile is
still present. Liou (1993) mentioned that besides a velocity difference, the streamline curvature is also
a governing parameter of the mixing layer instability [25]. The centrifugal force may either enhance or
stabilize disturbances. This additional instability is called Görtler instability and can be recognized by
vortices that counter-rotate while travelling in the downstream direction. Figure 1-42 illustrates a
curved free shear layer.
Figure 1-42: Curved free shear layer [Liou (1993)]
40
Rayleigh described a stability criterion where the centrifugal force was implemented. In his circulation
criterion, he made a distinction between the situation where the inner stream is the fastest and the one
where the outer stream is the fastest. In the first case, the centrifugal force may destabilize the curved
free mixing layers. As a result, both K-H and centrifugal instability are relevant to the hydrodynamic
stability. In the other situation, the centrifugal force stabilizes the faster flow that streams on the
outside of the slow one. As a consequence, Kelvin-Helmholtz instability should dominate. In this
situation, the flow curvature reduces the growth rate of the instability. A parameter to quantify the
effects of the curvature of the mixing layer on the total instability is the Richardson number. In case of
a plane mixing layer, this number has a value that equals zero.
Liou (1993) performed a numerical calculation of the instability problem for both cases. To solve this
eigenvalue problem they used a fourth-order fixed stepwise Runge-Kutta scheme. With the complex
streamwise wavenumber α as eigenvalue, they calculated this problem for different values of the
curvatures (1/R = 0; 0.025; 0.05) with their corresponding Richardson number (0; 0.0488; 0.0952).
In the first case, where the centrifugal force tries to counteract the growth of disturbances and stabilize
the flow, the results pointed out that the spatial growth rate α decreases with increasing curvature of
the mixing layer. Figure 1-43 shows these results in function of the frequency σ for both cases.
Figure 1-43: Variation of growth rate α in function of the frequency σ [Liou (1993)]
The second case, where the centrifugal force destabilizes the mixing layer, shows the contrary. The
growth rates and the frequency range of the unstable mode increase with increasing curvature.
Furthermore, the instability is highly dependent on the frequency.
C. Accelerated mixing layers
An accelerated mixing layer can be induced by adjusting the section width of a straight mixing layer as
studied by Fiedler et al. (1991) [26]. A gradual symmetric reduction of the section width can lead to the
acceleration of both outer flows. They observed that the Reynolds shear stresses and the turbulent
intensity values strongly increase in magnitude compared to the straight mixing layer configuration.
This increase is mainly pronounced in the initial part of the mixing layer, further on it remains
constant.
41
D. Mixing layer in open channel junctions
The mixing layer at a confluence is a combination of the curved and accelerated type. Although these
individual parts are quite straightforward, the application to an open channel junction has some
difficulties. At first, the radius of curvature of the mixing layer is not uniform because it is not
supported by a solid wall. As a result, the shear plane evolves in time and depends on the flow ratio
and the geometric characteristics of the confluence. Secondly, the reduction of the flow section is not
symmetric around the centerline and does not occur gradually.
Figure 1-44: Sketch of the mixing layer in an open channel junction [Mignot et al. (2013)]
42
Influence of a gradient on shear flow stability 1.7.3.
Next to all the previously mentioned parameters that cause instability, the decrease of roughness over
the length of the flume or the immediate increase of the water depth in the flume can cause instability
as investigated by Chu (1991) [27]. His analysis was based on a shallow open channel with a horizontal
length scale significantly larger than the water depth. The Saint-Venant equations were used for the
calculation. His results were correlated with two dimensionless parameters:
- A bed-friction number defined at the inflection point which compares the stabilizing effect of
the bed-friction to the destabilizing effect of the transverse shear.
- A velocity parameter that characterizes the changes in depth and roughness across the open
channel.
In the limiting case of a weak transverse shear flow, the bed-friction number becomes the only
dimensionless parameter governing their stability.
The two most important parameters of the stability of a transverse shear flow are its magnitude and
the degree of bed-friction influence. The bed-friction has a positive and a negative effect on the
stability. On the one hand the bed friction generates small-scale disturbances, but on the other hand, it
has a stabilizing effect on the large-scale transverse disturbances. If this second factor dominates the
process, a transverse shear flow will become stable. Furthermore, a flow can be unstable because of the
inflections in the transverse velocity profile that are produced by the abrupt change in water depth.
Numerical calculations were executed on four groups of parallel flows. The hyperbolic tangent (Tanh)
and the hyperbolic secant velocity profile (Sech) were obtained either by varying the friction coefficient
while keeping the depth h constant or by varying the depth h while keeping the friction coefficient
constant. The obtained velocity profiles are shown in Figure 1-45 and listed below.
- Constant depth and TANH velocity profile
- Constant depth and SECH velocity profile
- Depth-varying flows with uniform bed-friction coefficient
43
Figure 1-45: (a) TANH velocity profile; (d) SECH velocity profile; created either by (b)(e) varying and
h = constant or by (c)(f) varying h and = constant [Chu (1991)]
They concluded that transverse shear flows in an open channel may not become unstable if the water
depth h and the friction coefficient vary gradually across the open channel. In this way, inflections
will not occur due to the abrupt change in water depth or friction coefficient.
44
Shear flow instability criteria 1.7.4.
In the previous paragraphs, a lot of parameters are mentioned that have an influence on the possible
(in)stability of shear planes. In this paragraph, a possible criterion is mentioned to predict whether
instability occurs or not. This criterion will be used in the present study to verify its validity on the
shear planes of the confluence. The occurrence of hydrodynamic instability can be verified by using the
following procedure:
1. Search for a laminar solution to the hydrodynamic equations (Mass conservation and
Navier-Stokes equations) .
2. Add small disturbances to the laminar solution, usually sinusoidal in space and time.
3. Substitute the disturbed solution into the hydrodynamic equations to derive disturbance
equations. This results in an eigenvalue problem for the wave number and frequency of the
disturbances.
4. Solve the eigenvalue problem to study instability.
When following this scheme, three different types of solutions are possible:
- Absolute instabilities where the disturbances grow in time
- Convective instabilities where the disturbances grow in space
- Stable solutions where the disturbances are damped
Orr-Sommerfeld described this eigenvalue problem for viscous parallel flows. Later, Rayleigh (1880)
adapted this equation for inviscous flows and derived a criterion to check the stability of an inviscid
shear flow. This criterion is better known as the Rayleigh’s inflection-point theorem. He verified that
shear flows can be unstable if an inflection point is present in the velocity profile. This condition is
necessary, but not sufficient in order to have inviscid shear flow instability.
The physical interpretation of this theorem will be discussed by introducing the pseudovector
vorticity . This parameter represents a field that describes the local spinning motion of a fluid. For a
two-dimensional shear flow, the vorticity distribution is given by equation 1.20.
( 1.20 )
For the velocity profile shown in Figure 1-46a, the vorticity can be denoted as equation 1.21. When the
velocity field U has an inflection point,
, the vorticity will have a local maximum.
( 1.21 )
Consider a fluid particle that undergoes a virtual displacement from to a region of lower vorticity
(Figure 1-46b). According to the vorticity transport law in an inviscid fluid, this particle takes its
vorticity along with it. As a result, there is a net excess of vorticity which induces a positive vorticial
flow as shown in Figure 1-46c. This rotating current transports lower vorticity from above to the left
45
side of this vortex particle while higher vorticity from below is transported to right side of this particle
(Figure 1-46d). Both actions induce a downward motion pushing this fluid particle towards its original
state (Figure 1-46e). On the contrary, if a fluid particle moves from a region of high vorticity into
a region of lower vorticity , it is not forced back to its original location but is accelerated away from
it. This leads to instability.
Figure 1-46: Physical interpretation of Rayleigh’s inflection-point theorem
A more accurate criterion for instability was developed by Fjørtoft (1950). This stronger version of
Rayleigh’s criterion needs, besides the existence of an inflection-point, a vorticity maximum inside the
flow region (excluding the boundaries). This criterion is illustrated in Figure 1-47 which shows six
parallel flows with their velocity and vorticity profile. If a vorticity extremum was present in
Figure 1-46, a fluid particle, that arrives at this location, is not forced back to its original position.
Since the vortex at the next location is lower, it will not be transported to another location and the fluid
particle, with its high vorticity, will remain in place.
Figure 1-47: Velocity and vorticity profiles to demonstrate Fjørtoft criterion
46
According to the Fjørtoft, only when the velocity profile has an inflection point and the vorticity has an
extremum inside the flow region, instability will occur [28]. Thus, negative vorticity needs a minimum
at the same level of the inflection point to trigger instability. Similarly, positive vorticity needs a
maximum. Table 1-5 summarizes this and shows whether the conditions for instability are fulfilled or
not. In paragraph §2.5.7, this criterion will be used to check the instability of the shear planes at the
confluence.
Table 1-5: Velocity and vorticity profiles to illustrate the Fjørtoft criterion
Graph Inflection point? Vorticity extremum Stability
a X n/a Stable
b X n/a Stable
c ζ <0 & no minimum Stable
d ζ <0 & minimum Unstable
e ζ >0 & no maximum Stable
f ζ >0 & maximum Unstable
47
1.8. Conclusion
In the first paragraphs, the basic concepts of roughness and resistance were explained. It is important
that a clear distinction is made between both terms since they are not interchangeable. Resistance
quantifies the behaviour of the boundary which resists the flow of the fluid whereas roughness
describes the actual unevenness of the boundary.
Starting from the boundary layer theory, an overview was given of the different types of bed material
and their influence on the bed friction. In the present study, a grass cover was used to increase the bed
roughness, as a result, the part about vegetated channels was described extensively. The different
forms of vegetation were described in function of the degree of submergence. The main types are well-
submerged, submerged and emerged vegetation. In the present study, the water depth is 41.5cm and
the grass cover is 3cm. Since the ratio between both is bigger than five, the well-submerged velocity
profile is valid [6]. Furthermore, the effect of flexibility and the energy exchange within the vegetated
channel was discussed.
Secondary currents in a straight channel originate in another way than the ones in a curved channel. In
a straight channel, the free surface and the presence of the walls will induce turbulent friction. As a
result, secondary currents originate in the channel. On the other hand, also a centrifugal force
influences the flow pattern in a curved channel. The turbulence–driven force that was present in
straight channels still exists in curved channels but has less influence. It is clear that in a 90°
confluence, the centrifugal force will dominate the origin of these secondary cells because of the big
curvature of the flow.
The flow characteristics of an open channel confluence were investigated based on the experiments of
Shumate. His experimental set-up is similar to the one of the present study, thus most of his
observations are expected in this research. The most important flow features of a confluence are flow
separation, flow stagnation, flow contraction and a mixing layer between the combining streams.
Numerical simulations showed that both the length and the width of the separation zone decrease with
increasing friction coefficient, whereas the ratio of width to length increases with increasing resistance
[19]. Furthermore, the dimensions of this zone decrease when the flow ratio q* increases.
The literature review ended with a description of flow instability, in particular shear flow instability.
First a description of the Kelvin-Helmholtz instability was given and was linked to Rayleigh’s
inflection-point theorem (1880). He developed a linear stability theory for inviscid parallel shear flows
and proved that an inflection-point in the velocity profile is a necessary condition to trigger instability.
This condition is necessary, but not sufficient in order to have inviscid shear flow instability. A more
accurate criterion is developed by Fjørtoft (1950). This stronger version of Rayleigh’s theorem needs,
besides the existence of an inflection-point, a vorticity maximum inside the flow region. This criterion
will later be used to check the stability of the shear planes at the confluence. Transverse shear flows in
open channels are induced by gradients (e.g. water depth or bottom friction). These shear flows can
remain stable when this gradient is small enough to prevent inflections in the velocity profile.
49
2. Experiments
2.1. Introduction
In the literature review, the theoretical background of the experimental research was explained. In this
chapter, this information will be used to explain the expected flow features of the experiments and to
interpret the obtained results. At first, the experimental setup with the laboratory flume and the
different measurements devices are explained. Large scale surface particle image velocimetry (LSSPIV)
and measurements with an acoustic Doppler velocimeter (ADV) are performed to obtain experimental
data. The bed material was adjusted to reproduce the effect of changing bed roughness in a confluence.
The basic configuration is the concrete flume while in the second case artificial grass was put on the
bottom to apply another friction factor. Experiments were performed for both configurations with the
same boundary conditions. A constant flow discharge of q*=0.25 was imposed. This value was chosen
based on previous experiments (e.g. Shumate (1998), Schindfessel et al. (2014)) so a comparison can
be made. As mentioned before q* is defined as:
( 2.1 )
Where and denote the incoming discharge of the main channel and the tributary respectively,
expresses the downstream discharge and is the sum of the discharge of both inlet channels.
2.2. Layout of the flume
The experiments were performed at the hydraulics laboratory of Ghent. The open channel facility
consists of a 90° channel confluence with a chamfered rectangular cross-section with concrete walls.
The main channel has a length of 33.18m. The tributary is 5.17m long and intersects with the main
channel at 13.12m (measured from upstream). Both channels have a constant width of 0.98m. The
coordinate system originates at the upstream corner of the junction. The x-axis is orientated along the
main channel towards upstream while the y-axis is pointing in the downstream direction of the
tributary. The coordinates are always represented dimensionless by dividing them by the channel
width . Figure 2-1 gives an overview of the geometrical characteristics of the experimental facility.
Figure 2-1: Layout of the test flume
50
Both inlet channels are connected to the storage tank through pipes. To straighten and stabilize the
flow, filters are installed at the start of both channels. These consist of quarry stones entrapped
between two plates with a honeycomb structure as can be seen on Figure 2-2. As the tributary is only
5.17m long, the flow will probably not be fully developed at the start of the confluence. In the main
channel this is less an issue because its length is 13.16m, so it will be fully developed when it reaches
the confluence. The downstream boundary condition is controlled by a crest weir. Its height is 0.36m
and its width is 0.019m. Some minor leakage is present at both sides of the weir. A picture of the crest
is shown in Figure 2-3.
Figure 2-2: Filter at inlet channels
Figure 2-3: Downstream boundary condition, weir
The flow discharge of the main channel is measured by an electromagnetic flow meter mounted at the
inlet pipe and regulated by an adjustable valve. According to the manufacturer, the accuracy of the
electromagnetic flow meter is . The discharge of the tributary is regulated by a weir that can
be adjusted by a valve. Figure 2-4 shows the discharge measurement devices.
Figure 2-4: Discharge measurement devices – left: electromagnetic flow meter, right: gauge for the weir
51
The downstream discharge of 40 l/s together with a constant downstream water level of 0.415 m
yields to a constant downstream Froude number of 0.05. This number is calculated with equation
2.2 with .
√
( 2.2 )
This value is typical for lowland rivers, (Khublaryan 2009) [29]. The low flow velocities, which are
limited by the robustness of the inlet filter, result in negligible water level differences between the
beginning and the end of the flume. The most important flow parameters are summarized in Table 2-1.
Table 2-1: Parameters of the experimental setup
Flow parameter Value
[m³/s] 0.040
[m/s] 0.104
[m²] 0.384
[m] 0.980
[-] 0.050
[-] 98000
53
2.3. Measurement devices
Large-scale surface particle image velocimetry 2.3.1.
A. Introduction
Large-scale surface particle image velocimetry (LSSPIV) is an optical technique to measure flow
velocities at the free surface. Therefore a 1920x1080 pixel camera is mounted on a horizontal bar and
placed at a fixed height above the water surface taking images at a rate of 15Hz. The setup of the
LSSPIV is shown in Figure 2-5. The fluid is seeded with a tracing material that is recognizable by the
camera. The floating material applied consists of polypropylene particles with a maximum size of
5mm, covered in a white coating. The white coating is used to increase the contrast and makes the
image processing easier. Because the particles are very small, they will not interrupt the average flow
field. This technique results in velocity fields at the free surface.
Figure 2-5: Setup of the LSSPIV system
B. Methodology
The first step in the process is the image recording. The duration of the recording is established on 180
seconds, thus every LSSPIV-measurement captures 2700 frames. This measurement duration results
in a good balance between quality and quantity of the experiments. The obtained frames suffer from
lens distortion but this issue is solved by calibrating them. The processing from frames to velocity data
is done with the freeware PIVlab 1.32. On these frames, the boundary conditions (the walls) have to be
indicated by creating masks on them. When this is done, some pre-processing settings are adjusted to
enhance the results. The next settings can be adapted depending on the resolution and computation
time that is taken as acceptable. Like for every setting, equilibrium needs to be found between the
accuracy of the calculation and the computational time. This was often done by trial and error.
- Contrast-limited adaptive histogram equalization (CLAHE): This setting is enabled by
default and will optimize the contrast of the images and is governed by the size of the
CLAHE filter. With decreasing window size, the contrast will increase but the
computational time will increase as well. A value of 20 pixels seems to give a good balance
between both.
54
- Highpass: When enabled, the images are sharpened and background noise is removed. It
can be adjusted by changing the filter size. In the present, study a value of 25 pixels was
chosen and gave decent results.
- The threshold value for clipping: All values below this threshold are removed, in most
calculations a value of 65 was used.
First, a distinction has to be made between PTV (particle tracking velocimetry) where particle by
particle is followed and PIV (particle image velocimetry) where an area is investigated. In the present
study, PIV is used where a cross-correlation of particles is made. The images are subdivided into
interrogation areas (IA) which contains some particles. The ideal situation is where each IA consists of
a single particle but then it is hard to find this particle back in the second image. Keane and Adrian
(1992) suggest that 8-10 particles within each interrogation area results in the best analysis. An IA is
selected in the first image (Figure 2-6a) and is compared with an area in the second frame [30].
Because the particles have moved over a certain distance D, the area needs to be found in a bigger
searching area (Figure 2-6b). To know the exact displacement of this particle pattern, a cross-
correlation is calculated for all possible displacements within the searching area. Equation 2.3 is used
to calculate the velocity of this area.
( 2.3 )
With:
Figure 2-6: Principle of LSSPIV – (a): Image recording, (b): First image t=t0, (c): Second image t=t0+∆t
PIVlab offers two different LSSPIV algorithms to process the measurements:
- Direct-cross correlation (DCC) where a fixed interrogation area is defined. In this case, the
size of the area and the spacing between different interrogation areas are adjustable.
- FFT window deformation: this algorithm is more accurate because it uses different passes to
arrive at the final interrogation area. First a large interrogation area (e.g. 128 x 64 px) is
chosen. Large areas lead to a better signal-to-noise ratio and a more robust cross correlation.
55
The disadvantage is that it yields a low vector resolution. As a result, the size of the
interrogation area is decreased in the next steps. The writer of the freeware PIVlab
recommends to use three steps because this seems to give an optimal equilibrium between
vector resolution and computation time
In the experiments, the DCC algorithm with an interrogation area of 32 pixels and a step size of 16
pixels was used. This gave decent results and the computation time is smaller than the FFT window
deformation algorithm. After the processing, all velocities are still expressed in pixels/frames. To this
end, a calibration is done to change the velocities from pixel per frame to meter per second. Because of
the fixed height of the camera, a geometrical scale of 1 pixel = 0.00062m is obtained. The y and
z-coordinate of the camera are fixed (y= 0.623m, z=0m) for all measurements, the only geometrical
variable is the x-position of the bar. This is illustrated in Figure 2-7.
Figure 2-7: LSSPIV coordinate system with the position of the camera
With the fixed position of the camera, a rectangle of 0.65 x 1.19m is recorded. To ensure that the whole
area is measured, the borders of the field of view (FOV) between two different measurements need an
overlap. As one measurement covers a length of 0.65m, dx is chosen at 0.5m. In this way, there is an
overlap of 0.15m between two subsequent measurements (Figure 2-8). This overlap is removed in the
post-processing.
Figure 2-8: Overlap between subsequent LSSPIV measurements
56
Table 2-2 summarizes the locations of the bar and the range of the recordings. In the main channel,
measurements were taken at the confluence until 1m downstream of the separation zone.
Table 2-2: Locations of the LSSPIV measurements
X-position
bar [m]
X-Interval
[m]
Me
as
ur
em
en
t
1 -0.5 -0.63/0.02
2 -1 -0.48/-1.13
3 -1.5 -0.98/-1.63
4 -2 -1.48/-2.13
5 -2.5 -1.99/-2.64
6 -3 -2.49/-3.14
7 -3.5 -3.00/-3.65
The last step of the post-processing is to define the location of the downstream corner. Once this
reference is defined, the obtained data points are localised in the overal xyz-coordinate system.
Remarks:
- The seeding of the material is done manually, so attention has to be paid during
measurements. Particles should be seeded equally over the entire zone and if possible in
front of this zone. When tracers are seeded in the field of view, minor disturbances on the
water surface will occur which cause reflections of the light. These reflections will disturb
the quality of the data processing. It is not always possible to avoid this, in particular in
the separation zone where upstream velocities exist. When there is only seeded in front of
the field of view, tracer particles will always float downstream and will never enter the
separation zone. Nevertheless, velocity patters in this zone are necessary so seeding in this
zone, in the field of view, is unavoidable.
- A lamp is used as extern light source to overrule the natural light and to create equal
illumination of all areas. Although, attention has to be paid as too much light will give a lot
of reflection, which has negative effects on the processing of the results. An example of a
frame with too much reflection is presented in Figure 2-9. The optimal position of the
lamp was found by trial and error and is in most cases 4m away from the field of view. The
light of the lamp has an oblique angle with respect to the vertical plane because only the
water surface needs to be illuminated and not the entire channel depth.
- When too many particles are seeded at the same location, they will stick together and
move as a group. The flow velocity of a single tracer can differ from the velocity of a group
of particles which will affect the obtained flow patterns. Particles also stick to walls when
they have a low velocity e.g. in the separation zone. Thus, too much seeding at the same
location should be avoided.
- After a while, the coating of the particles will disappear. As a result, the contrast will
decrease which can affect the results.
57
Figure 2-9: Reflection of light during LSSPIV
Acoustic Doppler Velocimetry 2.3.2.
A. Methodology
To create velocity a profile in a cross-section of the flume, the ‘Vectrino II Profiling Velocimeter’ is
used. This device is a high-resolution acoustic Doppler velocimeter that can be used to measure
turbulence and velocities in 3D. These velocities are obtained by means of the Doppler effect. The
change of frequency is higher when the acoustic source approaches and lower during the recession.
This change indicates how fast a particle is moving and can be calculated with the following equation:
( 2.4 )
With:
In the present study, the ADV exists of one transmitter and four receivers. The transmitter is located in
the centre of the device and sends out a short pulse. This pulse is reflected on dirt particles that are
suspended in the water and is received by the four beams located at each side of the transmitter as can
be seen on Figure 2-10.
Figure 2-10: Close-up of the transmitter and the four beams of the ADV
58
These four beams pick up the pulses and depending on the time between transmission and receiving of
the signal and the change in frequency the velocity components can be obtained. Each
transmit/receive beam pair is sensitive to velocities in the direction of the angular bisector between
both beams. This means that every beam measures velocities that are 15° away from the transmit beam
as illustrated in Figure 2-11.
Figure 2-11: ADV - positive direction of Bisector (not on scale)
The device can only obtain accurate information from the first 40-70mm measured from the
transmitter and all beams are gathering data from the same sampling volume. This 30mm equals the
length of one profile (Figure 2-12). Every profile consists of 15 measurement cells with each cell having
a vertical size of 2mm. All velocities were measured in beam coordinates so every beam records a
velocity. In total, there are four velocities measured, . After the recording, they were
recalculated towards XYZ coordinates.
Figure 2-12: Measurement range ADV 30-70mm (not on scale)
To obtain an accurate velocity profile over the entire cross-section, nine vertical lines are measured
where each line consists of 15 measurement points with an intermediate distance of 3cm. In order to
perform measurements close to the surface, the ADV sensor was turned around so that the
transmitters send out pulses towards the free surface. Otherwise the transmitter will not be submerged
when measuring between 0.38cm and 0.44cm and the correct velocities will not be recorded. By
interpolation between the measured data, an approximate plot is obtained with velocities for the entire
cross-section.
59
Figure 2-13: Measurement grid - indication of upper boundary of measurement profile
B. Settings
Before the measurements can be carried out, some configurations have to be adjusted in order to
perform good measurements.
- A sampling frequency of 30Hz is used and every measurement has a duration of 120 seconds.
Based on the observations of Schindfessel et al., this sampling time resulted in decent time-
averaged velocities and still made it practically suitable to perform as much measurements as
possible on one day. But this will be discussed in section D. This implies that for every
measurement a total of 3600 pulses has been sent out.
- The velocity that is obtained from the ADV is an average of a lot of velocity estimates which are
called pings. The ping algorithm determines whether a long or a small ping interval is chosen
for the measurements. When the flow conditions are considered to be smooth, a ‘max interval’
should be chosen as the manufacturer suggests. However, when the flow conditions are
turbulent, the ‘min interval’ should be taken. In the present study, the ‘max interval’ algorithm
was used. This algorithm selects ping intervals based on the required velocity range. When a
measurement meets this range, it will be used. When it does not meet it, the algorithm chooses
values that meet the range requirement. Since the flow conditions at the confluence are
turbulent, the ‘min interval’ algorithm should have been used. However, because of the danger
of acoustic interference (reflections of previous pulses interfering with the current pulse) the
ping interval was chosen to be long. In this way, the danger for acoustic interference is
minimized and the turbulent velocity component is more difficult to measure. Since only
average velocities need to be obtained from every measurement, this does not give any
problems.
60
- The SNR, signal-to-noise ratio represents the quality of the signal and how large the noise in
the measurement is. If this value is large, the signal will be easy to detect and the amount of
noise will be negligible. If however this ratio is small (<10), the original signal will be very hard
to detect and measures have to be taken to increase this value again. One of the solutions is to
create some suspension in the upstream channels, so more particles become suspended. Low
values of the SNR is the main cause of inaccurate measurements
- The velocity range was always chosen to be 0.5. This gave good results for almost all
measurements. Close to the bottom and in the middle of the cross-section the SNR display
appeared to be stationary at some locations in the profile. It is hypothesized that this is due to
too much reflections at the bottom and due to the geometric characteristics of the cross-
section. This resulted in inaccurate measurements so the velocity range was adjusted towards
values in the range of 0.2-0.4 until the display fluctuated again.
- Also the choice of the location of the measurement profile has to be made. The manufacturer
already suggests a sweet spot exists in the SNR, this spot is located at 5cm below the
transmitter and is the location where the SNR is highest. So the best profiling range is located
around this sweet spot. Because of this, the measurement cell was chosen at 40 to 70mm so
the sweet spot is located in the middle of the profile.
C. Processing
Before the gathered data of the ADV can be used, some post-processing has to be applied. Most of the
time, the ADV does not record the exact 2min so first the records have to be cut off at a duration of 120
seconds.
Figure 2-14: Plot of Göring and Nikora Despiking criterion (2002)
61
Afterwards, the records of every beam were checked on the presence of noise. This noise is represented
by spikes that occur on the record. The cause of these spikes can be high turbulence intensities. To
remove most of the spikes (outliers) from the record, an algorithm has been used that was proposed by
Göring and Nikora with some adjustments proposed by Wahl (2003) [31], [32]. This algorithm plots
the value of the velocity together with its first and second derivatives against each other in a 3D map. A
3D ellipsoid was also plotted based on the universal criterion . This criterion, that is presented in
formula 2.5, determines whether a point will be taken as valid or if it should be retained from the
dataset.
( ) √ ( 2.5 )
With: -
In the 3D map, this means that if a point lies within the ellipsoid it is a valid point, otherwise it is a
spike. This is done using an iteration method in which the size of the ellipsoid reduces until this
reduction has no longer any significant effect on the record. This proceeds until no new data lies
outside the ellipsoid. Only one iteration was done as proposed by Wahl (2003). He also suggests that
the removed points from the record should be replaced with the mean value of all points.
D. Reference measurements
To check the reproducibility of the measured data, the velocity profile at one location was measured
during every measurement session. This location remained the same during the entire research and
measurements during 10min and 2min were performed. With these data, the possible error that could
occur during the placement of the ADV can be calculated. This reference point was chosen downstream
of the confluence at x*=-1.33 in the middle of the cross-section (y*=0.50) between z=0.14-0.17. From
this sampling profile, which consists of 15 measurement cells, the mean of every velocity component
was calculated. This calculation was performed on a total of seven measurements in the concrete
configuration and five in the grass configuration. From this sample distribution, the average and the
standard deviation were calculated for every component. The results are listed in Table 2-3 and Table
2-4. By comparing the values measured during 2 and 10 minutes, the quality of the chosen time frame
can be discussed.
62
Table 2-3: Mean and standard deviation of reference measurements - Concrete flume
10
min
Mean -0.1400 0.0480 -0.0079 -0.0073 0.0156 0.0173 0.0107 0.0103
Standard
deviation 0.0047 0.0086 0.0156 0.0155 0.0014 0.0013 0.0007 0.0007
2
min
Mean -0.1395 0.0476 -0.0070 -0.0063 0.0157 0.0171 0.0108 0.0103
Standard
deviation 0.0055 0.0087 0.0138 0.0136 0.0018 0.0019 0.0015 0.0013
Table 2-4: Mean and standard deviation of reference measurements - Grass flume
10
min
Mean -0.1532 0.0289 -0.0122 -0.0122 0.0171 0.0211 0.0111 0.0108
Standard
deviation 0.0042 0.0034 0.0026 0.0026 0.0013 0.0011 0.0005 0.0005
2
min
Mean -0.1521 0.0282 -0.0121 -0.0121 0.0156 0.0202 0.0103 0.0101
Standard
deviation 0.0019 0.0044 0.0021 0.0021 0.0014 0.0016 0.0009 0.0009
These tables indicate that the overall velocities are larger in the grass flume. This is because of the
altered velocity distribution which will be discussed later. Furthermore, one can see that the possible
error by choosing a time frame of 2min is negligible in the x and y-direction. The mean values are
almost completely the same for both configurations and the same is true for the standard deviations.
In the z-direction, the difference is larger. For the concrete flume, the difference between the mean
values is almost 10% while it is negligible for the grass configuration. The velocities in the z-direction
are more sensitive to variations since their absolute values are a lot smaller compared to the ones in
the x and y-direction.
To obtain the possible error that can occur by repositioning the ADV between subsequent
measurements, the mean values measured during 10min will be considered. Again a distinction has to
be made between the values measured in the grass flume and the ones measured in the concrete flume.
In the grass flume, the possible errors are a lot smaller. In the x-direction, the standard deviation is
0.0047 for the concrete flume and 0.0042 for the grass flume. This means that almost all
measurements lie within a very small confidence interval around the mean value so the possible errors
on this component can be neglected. For the y-component, the standard deviation is very small as well.
So variations on these components can also be neglected.
In the z-direction, the difference is a lot larger. For the concrete flume, the standard deviation reaches
a value that is two times larger than the mean value. This clearly indicates that this velocity component
is not reliable and should be interpreted with care. When comparing these values with the grass flume,
the standard deviation of 0.0026 (in the z-direction) is still large compared to the average value of
-0.0122 but smaller compared to the concrete configuration. Similar to the concrete configuration, the
63
z-component has to be interpreted with care. When both situations are compared, these differences
have to be taken into account. Consistent vertical motions will be reliable and will be present in the
test flume, but individual values will not. The setup of the ADV measurement device is designed so that
the positioning happens as accurately as possible. However, variations are possible since only a visual
control of its perpendicular position was performed. When the comparison is made between the
standard deviations in the concrete and the grass flume, the standard deviations of the samples of the
grass flume are a lot smaller so these data will be much more reliable.
E. Choice cross-sections
To visualize the processes that take place at the confluence, four different cross-sections have been
measured in both situations. These are chosen at x*=0.0, x*=-0.5, x*=-1.33 and y*=0.0 (Figure 2-15).
Figure 2-15: Location of the ADV – cross-sections
The velocities in the cross-section x*=-1.33 give a clear representation of the processes that occur
downstream and the flow features in the separation zone. The cross-section at x*=-0.5 was chosen to
see what processes occur in the confluence before both flows have merged and how the flow from the
tributary responds to the presence of momentum from the main channel. The cross-sections at x*=0.0
and y*=0.0 were chosen so that the inflow of water from both channels could be monitored.
Table 2-5: Location of the ADV – cross-sections
Location ADV – Cross-sections
A = 0
B = 0
C = -0.5
D = -1.33
65
2.4. Preparations
To ensure good measurement results at the confluence, the inflow sections need fully developed flow
conditions. As mentioned before, the water from the inlet branches passes through a filter to
straighten and stabilize the inflow. The length of the main channel between the filter and the
confluence is 13.38Wd (13.115m) and is long enough to assume that the flow is fully developed. On the
contrary, the tributary is much shorter (5.25Wd) and will probably not have fully developed inflow
conditions. This limited length between the filter and the confluence can have a major effect on the
flow features that will be observed in the confluence. To estimate what the inflow velocity profile looks
like and to verify whether or not the flow is fully developed, a vertical cross-section, upstream in the
tributary, is measured. This cross-section is chosen as far as possible upstream in the tributary so the
complex flow pattern of the confluence will not influence the measurements. Practical, the upstream
position of this cross-section is limited by a tunnel covering the tributary from y*=-2.05 until the filter.
As measuring in the tunnel is quite difficult, the cross-section is chosen at y*=-1.92. A cross-sectional
plot is created with the data received from the ADV which resulted in Figure 2-16. The measurements
were performed in the concrete flume and show the longitudinal velocities (color scale) and the cross-
sectional velocities (vector scale).
Figure 2-16: Inflow section tributary at y*=-1.92 (concrete flume) – Present study
One clearly sees that the flow is not fully developed. The longitudinal velocities at the left are much
higher than the ones on the right. Leakage at the sides of the filter could be a reason for this irregular
velocity distribution but since no leakage was spotted during controls of this filter, these were assumed
to be negligible. Another explanation could be the influence of the channel confluence on the upstream
tributary inflow. At the confluence, the flow from the main channel could push the flow from the
tributary to the left creating a zone of larger velocities at the left and lower velocities at the right side of
the section. This zone of low velocities could extend into the tributary until this section. Despite the
fact that the location of this cross-section is quite far upstream, these phenomena were assigned to the
influence of the confluence so this section was assumed to give a decent representation of the actual
flow features.
The concrete configuration has already been investigated by Schindfessel el al. (2014). In their
experiments, they used the same experimental set-up with the same boundary conditions and flow
66
variables (discharge ratio). To limit the amount of measurements, some data, with a flow ratio of
q*=0.25, from Schindfessel et al. (2014) were used. In this way, the measurements in the concrete
flume are limited and more time could be spent on the measurements of the grass configuration. This
resulted in a good time management so data could be obtained from both configurations at all desired
locations.
Schindfessel et al. measured also a tributary inflow section at y*=-1.50. A different velocity pattern,
than the one presented above, was obtained. The zone with large velocities at the left was less
pronounced and the velocity profile was more uniformly distributed over the width and the depth of
the channel. Since the filter of the tributary had been moved between both experiments, these
differences were assumed to be caused by this movement and no further attention had been given
towards them.
To be sure that the measured velocity profile at this tributary was still present at the end of the
experimental research, a vertical velocity profile was measured at y*=-1.92. This profile was chosen in
the middle of the cross-section at x*=-0.50 (Figure 2-17a). It appeared to be completely different from
the one measured at the beginning of the measurements (Figure 2-17b) and the one recorded by
Schindfessel et al. (Figure 2-17c). The profile had a more logarithmic shape and the high velocities at
the bottom had disappeared. An additional measurement over the width of the channel showed that
the zone with large velocities had shifted towards the right side of the channel compared to the ones
performed at the beginning of the research.
Figure 2-17: Velocity profiles in tributary (y*=-1.92, x*=0.49) – (a): Present study (after),
(b): Present study (before) (c): Schindfessel et al. (2014)
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-
]
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
(a) (b) (c)
67
No explanation could be given towards these differences since no relocation of the filter had occurred
and during checks of the filter, no leakage problems were discovered. As a consequence, the results in
the following sections have to be handled with care. All three profiles give differences in shape and
magnitude. The velocity profile from Schindfessel et al. showed a good distribution across the width of
the channel. Despite the larger velocities at the bottom, it is assumed that these measurements are
appropriate and that they can be used for comparison with the grass configuration. Since the flow is
uniformly distributed over the width of the channel, changes of this profile at the confluence will be
caused by the confluence and will not have been present at the tributary.
The data gathered for the present study in the concrete flume will not be as reliable as the ones
measured by Schindfessel et al. The large velocity difference between the left and the right side of the
channel may have had a large influence on the flow features at the confluence. As a consequence these
measurements have to be investigated critically. The measurements at the grass configuration are a lot
better. The LSSPIV data shows that the inflow is uniformly distributed over the width of the channel.
Furthermore, the velocity profile shows no large velocities at the bottom. As a consequence, these
measurements are considered to be reliable.
When comparing measurements data, it will always be mentioned who took the measurements so it is
clear which velocity distribution at the tributary was present. At the confluence of the concrete
configuration, almost only the LSSPIV data of Schindfessel et al. was used, so comparisons with
LSSPIV data between concrete and grass will be reliable. Changes in the flow pattern will be caused by
the roughness at the bottom and not by the change in velocity profile at the tributary.
For the ADV-data, the velocity profile at x*=-1.33 at the concrete configuration was measured by
Schindfessel et al. Since the data at the grass configuration is assumed to be reliable, a comparison of
both cross-sections will be reliable. The velocity profile at x*=0 was measured in this research but
since the tributary has no influence on this section these data will be reliable as well. The ADV-data
gathered at the other cross-sections (x*=-0.5 and y*=0) in the concrete configuration has to be studied
critically to draw no false conclusions.
Because the reliability of these inflow sections is important, all measurement data with their
corresponding parameters are summarized in Table 2-6.
Table 2-6: Reliability measurements
Observer Type of bed material Type of measurement reliability results
Schindfessel el al. Concrete All LSSPIV and ADV (x*=-1.33) Good
Present study Concrete ADV (x*=0, x*=-0.5, y*=0) Maybe
Present study Grass All LSSPIV and all ADV Good
In paragraph §2.3.2.D, the reproducibility of the ADV data was checked. The analysis showed that
every measurement session resulted in approximately the same values. The reproducibility of these
data was good which indicates that the flow features at the confluence are less influenced by the
68
different velocity profiles than initially considered. This increases the reliability of the measurements
performed downstream.
The grass cover consists of several parts of different length and width. The location of these elements
was chosen attentive to create a smooth transition between the different grass covers and to minimize
the influence of the flow features at the confluence. The cover indicated with the number three on
Figure 2-18 has a length of 6m and was placed so that the intersection with the cover upstream in the
main channel occurred at approximately x*=1. To this end, the possible influence of this transition on
the confluence will be negligible. This cover extends 5m downstream so the entire investigated part of
the main channel at the confluence is covered with one grass element. In this way, possible changes in
flow features between both ends of the cover are caused by the processes that occur at the confluence
and not by inaccurate placement of the separate elements.
Figure 2-18: Location grass cover
With the placement of the grass cover comes an extra error. The grass cover is approximately 1.1 cm
thick when compressed by the ADV. In the concrete flume, the ADV was standing on top of the
concrete but in the grass flume the device stands on top of the grass cover. Because of this, all values
have to be displaced 1.1 cm upwards. In this way a comparison of both plots is more easily.
Figure 2-19: Thickness grass cover when compressed by ADV
69
2.5. Results
Comparison velocity profile concrete and grass 2.5.1.
According to Kleinhans, the vegetation is well-submerged when the water depth is at least five times
the height of the vegetation [6]. In the present study the height of the grass cover is only 3cm while the
water depth is 41.5cm. This means that the flow is delayed within the vegetation but the flow that goes
over the vegetation is not blocked. As a result, a logarithmic velocity profile is expected for the flume
covered with grass.
To be able to interpret the velocity profiles that are obtained in the different cross-sections, the velocity
profiles of both concrete and grass have been measured. They were taken in the tributary at y*=-1.92
in the middle of the cross-section (x*=-0.50). This resulted in the following profiles:
Figure 2-20: Velocity Profile - Sweet spots - Left: Concrete / Right: Grass
Both plots show a logarithmic velocity profile as the theory predicts. The velocity in the flume with
grass reaches a value of almost zero when the upper side of the grass is reached. The velocity profile in
the concrete flume reaches faster an almost constant value than the one in the grass section so the
velocities at the top of the grass flume are larger. This can be clarified by looking at the measured water
elevations at the concrete and the grass flume. Since the water level differences between the concrete
and the grass flume are almost negligible, the water has to pass a smaller area with an equal flow rate.
This causes larger velocities in the upper part of the water while the flow at the bottom is more
retarded due to the additional roughness. In the concrete flume, the velocity profile will develop very
fast and the constant velocity zone will be reached almost instantly.
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
0.0 0.4 0.8 1.2
0.0
0.1
0.2
0.3
0.4
v U0 [-]
zW
d [-]
70
Comparison LSSPIV & ADV 2.5.2.
The methodology of both devices is quite different. However, it is important that both devices give
comparable results at the same location in the flume. Their compatibility will be checked in this
paragraph.
Since the LSSPIV can only measure flow velocities at the surface, the highest measurement row of the
ADV grid will be used to compare with the LSSPIV data. Depending on the cross-section that will be
compared, ADV measurements at z=41cm or z=40cm will be used. This is due to the fact that the ADV
data close to the surface is not always accurate. With an average water depth of 41.5cm, these
measurements will give a decent approximation of the velocities at the water plane. The ADV
measurements were performed at four different cross-sections so a comparison of data at some of
these locations will be made. First the section at x* =-1.33 will be analysed. Graph 2-1 shows the mean
velocities of the concrete configuration measured with both devices.
Graph 2-1: Mean velocities at x*=-1.33 – z=41cm (concrete)
Graph 2-1 indicates that both the magnitude and the shape of the mean velocities give quite similar
results. The location of the separation zone is visible in both measurements and the amplitude of the
velocity gradient is comparable. Graph 2-2 shows the standard deviation of the total velocity V.
Graph 2-2: Standard deviation of the velocities at x*=-1.33 – z=41cm (concrete)
-2
-1.5
-1
-0.5
0
0.5
1
1.5
2
0 0.2 0.4 0.6 0.8 1
Ve
loci
ty /
U0
y/Wd [-]
Vx- LSSPIV Vy- LSSPIV V - LSSPIV
Vx - ADV Vy - ADV V - ADV
0
0.1
0.2
0.3
0.4
0.5
0.6
0 0.2 0.4 0.6 0.8 1
σV
/U
0
y/Wd [-]
σV - ADV σV - LSSPIV
71
The global shape for both devices is the same. The location of the separation zone is visualized by
higher standard deviations. The values of the ADV are approximately 40% higher. This is probably
because the ADV is only suitable to measure in the water and not close to the water surface so more
variation can occur on these values. Comparable results at this cross-section were obtained for the data
of the grass flume.
The same comparison is made at x*=-0.5 in the grass flume. Here, the ADV results at a height of
41.0 cm gave fluctuating values that did not match with the LSSPIV. Experience showed that
measurements just below the water surface were difficult to execute. The water surface gives some
minor disturbances and acoustic devices are sensitive to this. As an alternative, the results at a height
of z=40 cm were used. The mean velocities at z=40 cm in the grass flume are given in Graph 2-3.
Similar to the section at x*=-1.33, both devices give comparable results. Again the standard deviations
obtained with the ADV are a lot higher than the ones obtained with the LSSPIV (Graph not included).
It can be concluded that both devices give comparable results, however attention has to be paid to the
ADV measurements just below the water surface since minor disturbances can result in inferior data.
Graph 2-3: Mean velocities at x*=-0.50 (grass)
-1
-0.5
0
0.5
1
1.5
0 0.2 0.4 0.6 0.8 1Ve
loci
ty /
U0
y/Wd [-]
Vx- ADV Vy - ADV V - ADV
Vx - LSSPIV Vy- LSSPIV V - LSSPIV
72
General flow pattern 2.5.3.
In this paragraph, the general flow pattern together with the main differences between both
configurations will be discussed. This comparison involves LSSPIV data which shows the flow features
at the water surface while the ADV data describes the processes that take place below this surface. The
LSSPIV-data of the concrete flume mainly consists of observations performed by Schindfessel et al.
(2014). Vector plots of the LSSPIV-measurements are shown in Figure 2-21 and Figure 2-22. These
indicate that the presence of the grass cover influences the dimensions of the separation zone. The
length of this zone decreases while the width increases. Inside the separation zone, the flow velocities
at the concrete flume have an upstream directed component. In the other configuration, these
velocities consist of an upstream component together with a lateral one directed towards the
boundaries of the separation zone. Downstream of this zone, this lateral component is still present and
diminishes further downstream. A more extensive description will be discussed in paragraph §2.5.4.
Figure 2-21: Vector plot LSSPIV (concrete) - The dotted line indicates points where u=0
Figure 2-22: Vector plot LSSPIV (grass) - The dotted line indicates points where u=0
73
Figure 2-23 shows contour plots of the surface velocities. In the grass configuration, these are higher
which was also indicated in the velocity profiles of §2.5.1. Because of the larger separation zone, the
flow has to pass a narrower channel so its overall surface velocities increase. At the tributary, the total
velocities are larger for the grass configuration which results in a smaller stagnation zone at the
upstream corner. Furthermore, the plot at the grass configuration indicates that the inflow from the
tributary is more uniformly distributed over the width of the channel.
Figure 2-23: Contour plot of total velocities (LSSPIV) - left: Concrete / right: Grass
Figure 2-24 shows the higher velocities in the cross-sectional-direction at the grass configuration. Due
to the larger velocities in the tributary, the flow enters the confluence with a larger force and pushes
the water from the main channel further to the side.
Figure 2-24: Contour plot of the cross-sectional velocities (LSSPIV) - left: Concrete / right: Grass
Four different cross-sections were measured using the ADV in both configurations and are shown in
the figures below. First the two cross-sections just before the confluence (x*=0 in the main channel
and y*=0 in the tributary) will be discussed. Afterwards, the cross-section at x*=-0.5 and the one
downstream of the confluence (x*=-1.33) will be analysed. Attention has to be paid while analysing the
plots at y*=0 and x*=-0.5 at the concrete configuration since these can be influenced by the changed
velocity profile, as mentioned in §2.4.
The ADV-plots at x*=0 (Figure 2-25) indicate that the highest downstream velocities occur at the wall
opposing the tributary (y*=1). This is due to the flow contraction caused by the lateral momentum of
the tributary. The flow on the left bank is decelerated and ends in the stagnation zone. The cross-
sectional motion, shown as arrows, is quite similar in both configurations. Its distribution is uniform
over the width of the channel and small in magnitude.
74
Figure 2-25: Cross-section x*=0 – left: Concrete / right: Grass
The velocities of the inflow section of the tributary (y*=0) are shown in Figure 2-26. Across the width
of this section, the longitudinal velocities decrease from the downstream corner, where the velocity is
maximal, to the upstream corner. The figure also indicates that the dimensions of the stagnation zone
decrease when the flume is equipped with the grass cover as mentioned before. Because of these higher
velocities, the water that was almost standing still in the stagnation zone is also pushed into the main
channel. A second difference between the concrete flume and the flume with the grass cover is seen in
the cross-sectional velocity. For the concrete case, the lateral-component is quite uniform over the
depth of the cross-section whereas, for the case with the grass cover this increases from the water
surface to the channel bottom. Furthermore, the lateral-component is slightly curved and follows the
shape of the chamfers. Close to the bottom of the grass configuration, a small area of upstream
velocities is observed. This area is larger at the upstream corner and disappears at the left bank.
Figure 2-26: Cross-section y*=0 – left: Concrete / right: Grass
Comparing the cross-sections at x*=-0.5, the cross-sectional velocity pattern is very different in both
situations. From y*=0 to y*=0.5, the concrete flume shows larger velocities in the middle compared to
the bottom and the top of the section. This means that the maximum lateral velocity is obtained at
mid-depth. This agrees with the observations at y*=0 and y*=-1.92 in the concrete configuration
(Figure 2-16). The zone with large velocities observed at y*=-1.92 has propagated towards y*=0 and is
the cause of the large velocities at x*=-0.5 at mid-depth. On the other hand, these velocities show a
velocity profile that increases towards the bottom for the grass configuration. Similarly this profile is
still only present between y*=0 and y*=0.5. Another difference is the existence of a secondary current
in the flume with the grass cover. This current rotates clockwise just above the channel bed and
extends from y*=0 until y*=0.5. This secondary current will be explained more extensively in
paragraph §2.5.6. The high longitudinal velocities observed at the grass configuration around y*=0.40,
75
indicate the location of the shear plane. The lateral inflow from the tributary is transformed in a
longitudinal component. This results in a reduction of this lateral component around this location.
Figure 2-27: Cross-section x*=-0.5 – left: Concrete / right: Grass
The last cross-section that will be discussed is located downstream of the confluence at x*=-1.33. The
measurement data of the concrete case is based on the experiments of Schindfessel et al. (2014). The
plots of this cross-section are shown in Figure 2-28 and show a clockwise rotating cell in the
separation zone. The dotted contour line indicates points with a longitudinal velocity of zero. Within
this zone, positive (upstream) velocities are present. For both cases the highest upstream velocities in
the separation zone appear at the water surface and decrease in magnitude when moving towards the
bottom. The secondary cell does not extend until the bottom of the flume but stops somewhere in
between.
For the concrete case, the separation zone is constant in width and ends at z*=0.10. This is contrary to
the results of Shumate who predicts a separation zone that decreases in width towards the bottom [16].
These differences are due to the presence of the chamfers. The larger longitudinal velocities near the
bottom of the left bank are caused by the reduced transfer of lateral momentum from the tributary to
the main channel. These observations were already made by Schindfessel el al. (2014).
For the case with the grass cover, the separation zone still extends vertically but ends already at mid-
depth (z*=0.20). Lateral inflow from the tributary collides on the flow of the main channel, bends
downward and is redirected towards the left bank. This is the reason for the abrupt end of the
separation zone at a depth of z*=0.20. This phenomenon does not appear for the concrete case
because here, the lateral velocities are smaller and do not have the power to reach the left bank after
the collision with the stream of the main channel. The zone with large velocities around y*=0.40 has
increased in comparison with the concrete flume. This will be further explained in §2.5.6.
Figure 2-28: Cross-section x*=-1.33 – left: Concrete / right: Grass
76
separation zone 2.5.4.
The length of the separation zone is clearly visible on the LSSPIV vector plots (Figure 2-21 and Figure
2-22). The dimensions of this zone vary in function of the bottom roughness. The experiments show
that the length decreases and the width increases when the bottom is equipped with a grass cover.
Because of this cover, the length decreases from 2.21 to 1.96 times the channel width . The ADV
plots at x*=-1.33 (Figure 2-28) show that the width of the separation zone increases from 0.21 to
0.28 which is compatible with the vector plots from Figure 2-21 and Figure 2-22. As mentioned
before, the depth of the separation zone extends further to the bottom at the concrete configuration.
These results are partly contradictory with the literature which predicts that both the length and the
width of the separation zone decrease with increasing friction factor. The artificial grass will besides on
the increased bed resistance also have an effect on the velocity distribution like mentioned before. By
equipping the flume with a grass cover, a different velocity profile was obtained. The flow at the
bottom is more retarded which results in higher flow velocities at the water surface. This phenomenon
is more pronounced in the tributary because its discharge is three times larger than the main channel.
Hence, the water coming from the tributary enters the main channel with a larger force and pushes the
water coming from the main channel further away. The flow coming from the main channel is more
compressed so the separation zone can expand more easily in width. Since the inflow of water from the
tributary channel was larger, the return flow towards the separation zone will also be larger so the
separation zone will diminish in length. This process is less pronounced at the concrete flume because
the momentum of the tributary is smaller. As a result, the flow of the tributary can more easily be
transported further downstream.
Another parameter that can be checked is the ratio of the width to length of the separation zone.
According to Creëlle et al. (2014) this ratio increases with increasing resistance. The measurements
confirm this, the ratio increases from 0.095 to 0.143 [19]. The upstream velocities in the separation
zone are shown in Figure 2-29. The plot points out that these upstream velocities are higher for the
concrete case than the case with the grass cover. The highest upstream velocities u/U0 are 0.65
respectively 0.50. This can be explained by means of the power of lateral inflow. The concrete flume
has, relative to the grass flume, smaller surface velocities coming from the tributary so the lateral
inflow cannot penetrate deep into the main channel. The contribution from the main channel is still
high resulting in large streamwise velocities. As a result, the upstream flow will in proportion be high.
The reverse is true for the flume with the grass cover.
Figure 2-29: Upstream velocities in the recirculation zone – left: concrete / right: grass
77
Mixing layer 2.5.5.
At a confluence of two independent streams, a mixing layer originates that separates both streams. In
this mixing layer, a velocity gradient exists. This mixing layer disappears again when both flows have
mixed enough and the velocity gradient is negligible. Due to the geometry of the junction, the flow is
both accelerated and curved in the mixing layer which complicates the development of the mixing
layer and the velocity gradient. Moreover, the grass cover on the bottom of the flume has a significant
impact on this mixing layer. In this paragraph, a short analysis of this influence will be discussed based
on the investigation performed by Mignot et Al. (2013) and the one performed by Chu et al. (1988)
[33][34].
The centreline of the mixing layer is defined as the location of maximum velocity gradient. This line
corresponds very well to the streamline that starts at the upstream corner of the confluence and has a
curved pattern. To visualize this pattern the part of the confluence between x*=0.0 and x*=-1.0 was
subdivided into different cross-sections. The LSSPIV-data was used to calculate the total flow velocity.
By using the smooth robust differentiators as proposed by ‘Holoborodko [35], the total velocity
gradient at each point was calculated and plotted in Figure 2-30.
Figure 2-30: Velocity Gradient - Mixing layer - Left: Concrete / Right : Grass
Due to the larger surface velocities of the grass flume, the mixing layer is pushed further away from the
tributary as can be seen on these figures. The flow in the concrete flume intersects with the cross-
section at x*=-1.0 at a height of approximately 3/5 times the channel width while this is at
approximately 4/5 in the grass flume. Furthermore, the amplitude of the velocity gradient is larger
in the grass flume.
To investigate both configurations, the surface velocity distribution was investigated at different cross-
sections. Starting at x*=-0.20, each 0.05 , a cross-section is made and the velocity distribution is
investigated. The cross-sections between x*=0 and x*=-0.20 were not taken into account. This was
78
because data from the tributary was necessary for this calculation. Since another velocity profile was
present, when these measurements were performed, a comparison is difficult to make. The cross-
sections after x*=-0.75 were neither included because the velocity gradient was hard to recognize. The
velocities in these sections had already converged a lot so a good determination of this layer was hard
to make. The results are shown in Graph 2-4 where x*=0 indicates the upstream corner and negative
values of x* denote locations further downstream.
Graph 2-4: Velocity U1 and U2 - concrete vs grass flume
The velocities U1 and U2 correspond to the velocities at the outer limits of the mixing layer. To
calculate these, the mean has been taken of five measurement points in the y-direction at this border
(Graph 2-5). The border of this layer is defined as the location where the velocity gradient becomes
negligible. The mean velocity near the flow coming from the tributary is U2 and the one close to the
main channel U1. The mean velocity U2 is always larger than U1 which could be expected.
Graph 2-5: Total Velocity Profile x*=-0.25 - Grass flume
Graph 2-4 indicates that the amplitude of both U1 and U2 increases with decreasing x*
(i.e. downstream). This is obvious since the overall velocities increase once both flows have merged.
0
0.2
0.4
0.6
0.8
1
1.2
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
U/U
d[-
]
x/Wd[-]
U1 - Grass U2 - Grass U1 - Concrete U2 - Concrete
79
However, the velocity difference over the mixing layer is not equal for the concrete configuration and
the grass configuration as shown in Graph 2-6.
Graph 2-6: ΔU - Concrete vs Grass flume
The velocity difference decreases in both configurations with decreasing x*. This is obvious since both
flows have to converge towards a uniform flow. The velocity difference is larger in the grass flume than
in the concrete flume. The difference decreases only a little bit in the concrete flume, while this value
decreases in the grass flume with half of its initial value. This can be explained when looking at the
overall flow pattern of the confluence. In the concrete configuration at the upstream corner, the
velocities remain low in both tributary and main channel. So at the first cross-section that was plotted,
only a small velocity variation can be spotted. Further downstream, the flow velocities increase, but
both flows start to merge so the overall effect will stagnate. At the grass configuration, the velocity
difference at the upstream corner is more pronounced. Because of this, the shearing effect will be
larger and the flow velocities will converge at a faster rate.
When looking at Graph 2-7, the velocity gradient decreases when going downstream. Since both flows
converge, this velocity gradient has to decrease so a single flow remains.
Graph 2-7: Maximum Velocity Gradient - Concrete vs Grass flume
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
ΔU
/U0[-
]
x/Wd[-]
Grass Concrete
00.5
11.5
22.5
33.5
44.5
-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
max
(dU
/dy)
[-]
x/Wd [-]
Grass Concrete
80
In the concrete configuration, the gradient remains almost constant from a value of x*=-0.45. In the
grass configuration this point cannot be found and the graph keeps decreasing until a value of x*=-0.8.
Once more, this is due to the overall change in flow pattern. Since the flow in the tributary has smaller
surface velocities at the upstream corner of the confluence, the velocity gradient between the tributary
and the main channel will initially be small. In the grass configuration, the flow from the tributary has
larger velocities so the initial velocity gradient will be larger. Furthermore, the velocity gradient will
decrease faster because of the larger velocity difference between both flows. The larger the velocity
difference between both flows, the faster they will merge.
Graph 2-8: Width δ of the mixing layer - Concrete vs Grass flume
The mixing layer width δ is defined as the ratio between the outer velocity difference ΔU and the
maximum velocity gradient of the cross-section.
| ⁄ |
( 2.6 )
Graph 2-8 shows the width of the mixing layer and indicates that it initially increases in both
configurations. In the concrete configuration, this width decreases from x*=-0.45. In the grass
configuration the width increases until a value of x*=-0.6 where it seems to remain constant. These
results agree well with the trend observed by Mignot et al. (2013). This plateau is related to the strong
lateral confinement of the flow when it reaches the confluence. The mixing layer is compressed and
has reached its maximum compression, as a result a constant width remains. This difference can be
explained by looking at the components of formula 2.6. For the concrete flume, the velocity gradient
remains almost constant from a value of x*=-0.6. As a consequence, the width of the mixing layer is
determined by the magnitude of the velocity difference. This difference decreases with decreasing x so
the width reduces. At the grass flume, the velocity gradient decreases with decreasing x until a value of
x*=-0.8. However, the rate at which it decreases, diminishes. As the velocity difference decreases as
well, the rate at which both values decrease, will become equal so the width of the mixing layer will
become the same. Probably at x*=-1.0 and further, the width of the mixing layer will decrease and the
distinction between both flows will become negligible.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
-0.8 -0.6 -0.4 -0.2 0
δ/W
d[-
]
x/Wd [-]
Grass Concrete
81
Secondary current 2.5.6.
As mentioned in paragraph §2.5.3, a secondary current is present at the bottom of the grass
confluence. In the concrete case, the flow from the tributary has less momentum so it can more easily
be transported downstream by the flow from the main channel. To see how the current develops along
the x-axis, two additional cross-sections are measured. A first section at x*=-0.25 and a second one at
x*=-0.75. These are chosen to check the development of this current before and after x*=-0.5. Since
the secondary current is only 10cm thick at x*=-0.5, only the four lowest measurement rows of the grid
are measured. This reduces the spent time and gives decent results. These are shown in Figure 2-31,
Figure 2-32 and Figure 2-33.
Figure 2-31: Cross-section x*=-0.75 - flume with grass cover
Figure 2-32: Cross-section x*=-0.5 - flume with grass cover
Figure 2-33: Cross-section x*=-0.25 - flume with grass cover
82
These figures show the location of the recirculation cell which has a clockwise rotating motion near the
bottom of the flume. In the upstream section (x*=-0.25), the cell has a significant smaller size
compared to the other ones. The cell size increases when going downstream. This increase can be seen
in both the width and the height of the cell. The width increases from 0.20m to approximately 0.40m
at x*=-0.75 and the height of the cell increases from 0.07 Wd at x*=-0.25 to a height of approximately
0.15Wd at x*=-0.75. Another difference is the larger longitudinal-component of the velocity in the
recirculation area when going further downstream. At x*=-0.25 this component is small and the zone
can hardly be seen. At x*=-0.75, this velocity component has increased towards a value that cannot be
neglected. This shows that an important undercurrent is present at the confluence. The flow that goes
over this recirculation area transforms its lateral component into a longitudinal component and is
transported downstream.
Furthermore, the initiation of this recirculation area starts further into the main channel when moving
downstream. At x*=-0.25, this zone starts at y*=0.05 while at x*=-0.50, the zone starts at y*=0.10 and
at x*=-0.75, this is approximately y*=0.20. A lateral shift has occurred from up to downstream. This
counters the possibility that this recirculation area is initiated by a bed discordance caused by two
grass covers that do not match exactly. Since two grass covers meet at the confluence, a height
difference could be present. When this is the case, the recirculation area should start at y*=0 in every
cross-section and the growth of this cell would not be as large as it is now. This lateral shift is also
present at Figure 2-26 at the bottom. The zone with upstream current at the bottom decreases in size
from the right bank to the left. This represents the lateral shift of this secondary cell.
The location of this recirculation area is approximately equal to the location of the mixing layer. This
mixing layer, as mentioned in §2.5.5 starts at the upstream corner and has a curved pattern towards
the right side of the flume (looking downstream). Table 2-7 shows a comparison of the LSSPIV data
and the ADV data. With the ADV data, the location of the recirculation cell was determined, while the
location of the mixing layer at the water surface was measured with the LSSPIV data. This
measurement is based on the location of the largest velocity gradient in each of these cross-sections.
The location of the recirculation area was guessed from the figures on the previous page.
Table 2-7: Location shear layer vs recirculation area
y* - LSSPIV y* - ADV
x*=-0.25 0.36 0.25
x*=-0.50 0.57 0.50
x*=-0.75 0.71 0.65
A comparison between both data is rather difficult since the LSSPIV-data is monitored at the water
surface and the ADV data at the bottom of the flume. Due to the skewed shape of the shear plane, these
locations can differ. However, the location of the recirculation area is always close to the location of the
mixing layer. The recirculation area is located closer to the left than the maximum velocity gradient.
83
A second comparison has been made to verify this difference. When looking at Figure 2-27, the flow
that goes over the recirculation area has a curved downward pattern until it meets the mixing layer. At
this layer, the flow at the water surface will have a downward directed velocity component. This can be
verified by plotting the divergence in both situations. When the divergence is negative, the flow at the
waterline goes down. When the divergence is positive, the flow goes up.
Figure 2-34: Divergence x*=0 - x*=-1.0 - Left: Concrete / Right: Grass
Figure 2-34 shows the divergence across the confluence for both configurations. In the concrete
configuration, no significant downward motion of the flow at the waterline is spotted. In the grass
configuration, a large downward movement can be seen. This movement corresponds with the location
of the end of the secondary cell. This can be explained by the curved motion of the flow over this cell.
The water has to pass over this zone after which it goes down again and gets a downward velocity
component. When Figure 2-34 is compared with Figure 2-30, it can be verified that the location of the
downward current is located just in front of the location of the maximum velocity gradient. The fact
that the flow goes downward is directly linked to the existence of this shear layer. The reason this does
not occur in the concrete flume is explained by the higher surface velocities in the grassed flume.
Because of these larger velocities, the water in the grass flume enters the confluence with a larger force
where it encounters the flow coming from the main channel. The flow on top gets a velocity component
directed downstream. The flow at the bottom gets entrapped and turns around and the secondary cell
originates. Further downstream, the flow also turns around at the bottom. The flow gets a downstream
component by the current coming from the main channel but is also pushed by the downward velocity
components created by the recirculation areas upstream. This is one of the reasons why the
longitudinal-velocities at x*=-0.75 are larger than the ones at x*=-0.25 and x*=-0.50.
84
Divergence plots of the separation zone are shown in Figure 2-35. One clearly distinguishes a relative
big upward motion (positive values) for the grass configuration while this upward motion is almost
absent at the concrete flume. At the grass flume, the water moves upward near the left bank and goes
down again close the shear layer. Between both actions, the water moves lateral towards the right
bank. These observations were also seen in the LSSPIV-vector plots (Figure 2-21 and Figure 2-22) and
the velocity contour plots in the y-direction (Figure 2-24). These plots indicate a lateral motion away
from the left bank towards the shear layer. This current is absent at the concrete configuration where
the fluid only has an upstream velocity component. These findings were confirmed with visual
observations of the tracer particles. In the grass flume they move under an oblique angle with an
upstream component and a lateral component towards the right bank. The particles of the concrete
case only have an upstream velocity component.
Figure 2-35: Divergence at separation zone – left: Concrete / right: Grass
85
Shear planes and vorticity 2.5.7.
The available data from the LSSPIV can be used to get more information about the instability of the
shear planes. First, the instability of the shear planes will be analysed in an optical way. Two different
shear planes are present, one originating at the upstream corner and ending at the reattachment point
and a second one going from the downstream corner until the end of the separation zone. These shear
planes are visualised in Figure 2-36 and in the real situation (Figure 2-37).
Figure 2-36: Indication of shear planes
Figure 2-37: Frame of the existing confluence where the shear planes are visualized with tracers
LSSPIV-frames from the grass configuration are shown in Figure 2-38 (x*=0 to x*=0.66), Figure 2-39
(x*=-1.02 to x*=-1.69) and Figure 2-40 (x*=-1.50 to x*=-2.16). These frames only give an
instantaneous impression of the situation, but are chosen attentively so they give a good
representation of the global flow pattern. Figure 2-38 shows that the tracers follow a curved track and
do not rotate around their axis. The red line on the figure represents the mixing layer. Because most
tracers do not rotate, no vortices are created along the shear plane.
86
Figure 2-38: Frame of the grass configuration from x*=-0 to x*=-0.66
The other shear plane, starting at the downstream corner, shows a very different pattern. Starting at
the downstream corner, vortices are shed over a limited length of the shear plane. These vortices grow
in size and rotate counter clockwise when moving downstream. This rotating motion confirms the
observations from Liou et al. (1993) for curved free shear layers. The rotating motion is created by the
interaction of the streamwise velocities of the main channel and the upstream velocities in the
separation zone. Figure 2-39 shows the growth of the vortices. Between x*=-1.60 and x*=-1.80 in the
grass configuration, the vortices stop growing and start to decay gradually. Figure 2-40 shows this
decay. In the concrete configuration these vortices start to grow from the downstream corner until
approximately x*=-1.40 where they start to decrease in size.
Figure 2-39: Frame of the grass configuration from x*=-1.02 to x*=-1.69
Figure 2-40: Frame of the grass configuration from x*=-1.50 to x*=-2.16
87
In both configurations, the size and quantity of these vortices was monitored. In the grass
configuration, fewer vortices were present compared to the concrete one but their average size was
larger. In the concrete flume, their dimensions varied more from very small to very large so no clear
trend could be drawn. As mentioned before, they start to decay faster so this layer becomes stable at an
earlier section.
Figure 2-41 shows the vorticity distribution at the mixing layer of the confluence. This shear layer
starts at the upstream corner and proceeds downstream. Calculation of these values was performed
using the velocities calculated with the robust smooth differentiator. The vorticity was then calculated
with the following formula:
( 2.7 )
Figure 2-41: Vorticity - mixing layer – left: concrete / right: grass
Both figures indicate the vorticity at the mixing layer which remains low. The blue regions on top of
both figures do not indicate vortices, but just mark the presence of the border. Because of this border,
the velocity gradient is large at this location which results in large amplitudes of the vorticity.
In both situations, vortices are present but are small and decrease rapidly in size. To check if
instability at the mixing layer can be predicted, the Fjörtoft criterion will be used. To trigger instability,
a vorticity extremum and an inflection point in the velocity profile have to be present at the same
location. The velocity profiles at the mixing layer were already studied in §2.5.5. In this paragraph,
only the sections between x*=-0.20 and x*=-0.75 were examined. Since the inflection point at
locations further downstream was difficult to find, instability will not occur at these locations. An
inflection point at the cross-section between x*=0 and x*=-0.20 is assumed to be present, so these will
also be considered to check instability. One condition to have instability according to Fjörtoft, is that
the vorticity has to reach a maximum or minimum.
88
Figure 2-41 shows that a small minimum is present close to the upstream corner that is more
pronounced in the grass configuration. This extremum has already disappeared at x*=-0.25. Since
almost no vorticity is present in this zone at the concrete configuration, instability will not occur. At
the grass configuration, the small zone that contains a vorticity extremum disappears when moving
downstream. As a result, the flow becomes stable. The optical observations agree with this. In this
zone, the flows encounter each other and vortices originate which do not grow in size and disappear
quickly. This can be explained by the small length of the zone where the vorticity minimum is present.
Because of this, the vortices do not have enough time to increase in size before the shear layer becomes
stable again. So for the mixing layer, there can be concluded that it is almost everywhere stable in both
cases. The criterion seems to be valid although some precautions towards its application have to be
taken into account.
For the shear layer that is present at the separation layer, the same investigation will be performed.
The vorticity distribution for the concrete and grass configuration is shown in the figures below.
Figure 2-42: Vorticity –separation zone – Left: Concrete / right: Grass
Both figures indicate that the vorticity reaches higher values at the separation layer compared to the
ones encountered at the mixing layer. This already indicates that there will be a higher chance on
instability. When both figures are compared, one clearly sees that a higher maximum is reached at the
grass flume. Moreover, the zone with large vorticity extends further downstream compared to the
concrete flume.
89
Graph 2-9: Total Velocity Profile x*=-1.33
For both cases, the velocity profile at x*=-1.33 is shown in Graph 2-9 and indicates that the total
velocity V reaches a minimum value at the border of the separation zone. Within the separation zone,
the total velocity increases again. Because there are upstream and downstream velocities at x*=-1.33,
there are two inflection points present in this profile. The outer inflection point is the one from the
shear plane and is indicated on the graph. Both inflection points are present over the entire length of
the separation zone so the first condition for instability is always met. The second condition (vorticity
extremum) will then determine whether or not instability will occur. From the optical observations in
the concrete configuration, it was observed that the vortices increase until x*=-1.4. Figure 2-42 shows
that the vorticity maximum has already decreased towards a value of approximately 15 at this section.
The maximum is less pronounced than it is upwards so the excitation of instability will not occur easily
at this location. The visual inspection of the grass configuration indicated that the vortices start to
degrade at x*=1.6-1.8. Figure 2-42 shows that the vorticity has a value at this location of approximately
15. Similarly, the maximum vorticity has decreased and is less pronounced so instability will not occur
easily.
From the previous inspection, it is concluded that the proposed criterion can be used to locate
instability but that precautions have to be made towards its application. The initiation of instability at
the separation zone can be derived with fewer difficulties than the end of stability at this location.
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
0 0.2 0.4 0.6 0.8 1
V/U
d [
-]
y/Wd [-]
Grass Concrete Inflection Point
91
2.6. Conclusion
Experiments were done to investigate the influence of a grass cover on the flow characteristics at a
confluence. LSSPIV was used to measure velocity patterns at the water surface, while the ADV was
used to measure the flow features at different cross-sections. In this way, the complex flow pattern at a
confluence was investigated in three dimensions.
The ADV-reference measurements indicated that the possible error by choosing a time frame of 2min
(instead of 10min) is negligible in the x and y-direction. This error is slightly bigger in z-direction
because here, the velocities are much smaller compared to the x and y-direction which results in more
sensitivity to deviations. To conclude, the choice of a time frame of 2min was justified. The
reproducibility of the ADV data was checked by means of confidence intervals. This interval is small
for velocities in the x and y-direction but is large in z-direction. For the concrete flume, the standard
deviation reaches a value that is two times larger than the mean value. As a result, the velocities in
z-direction should be interpreted with care.
Because of the limited length of the tributary, the flow is not fully developed when it reaches the
confluence. Different measurements upstream in the tributary, show altered velocity profiles. This is
probably caused by a shift of the filter at the tributary. As a consequence, attention has to be paid when
comparing ADV-data at y*=0 and x*=-0.5. The surface velocities of the LSSPIV were compared with
the ADV-data just below the water surface in order to check their compatibility. This was done for
every cross-section and gave decent results. Both the shape and the magnitude of the velocity profile
were comparable.
By equipping the flume with the grass cover, a different velocity profile was obtained with lower
velocities close to the bottom. Since the water level differences between the concrete and the grass
flume are almost negligible, the water has to pass a smaller area with an equal flow rate. This causes
larger velocities in the upper part of the cross-section while the flow at the bottom is retarded due to
the additional roughness. Since the discharge of the tributary is three times larger than the one of the
main channel, this change in velocity profile is more pronounced in the tributary. As a result,
momentum from the tributary pushes the shear layer further away from the side channel. The flow
coming from the main channel is more compressed so the separation zone can expand more in width,
whereas it decreases in length. Furthermore, the ratio of the width to length of the separation zone
increases. which agrees with the numerical simulations of Creëlle et al. (2014).
The depth of the recirculation zone is limited by the effect of the chamfer and the return flow to the left
bank located just above the channel bottom. This return flow will bend upwards in the separation zone
resulting in a lateral velocity component towards the right bank. This phenomenon does not appear for
the concrete case because here, the inflow velocities are smaller and do not have the power to reach the
left bank after the collision with the stream of the main channel. As a consequence, the flow in the
separation zone has only an upstream velocity component. These observations were confirmed in
divergence plots of the flow. Another difference is the existence of a secondary current in the flume
92
with the grass cover. This current rotates clockwise just above the channel bed and increases in size
when moving downstream. At x*=-0.25, it is hardly present while it is not negligible for x*=-0.50 and
x*=-0.75. Furthermore the initiation of this recirculation area starts further into the main channel
when moving downstream. . In the concrete case, this cell is absent because the flow from the tributary
has less momentum so it can more easily be transported downstream by the flow from the main
channel.
Visual observations showed that only vortices originate at the shear layer of the separation zone. The
Fjörtoft criterion was used to check the instability of the shear layers and to confirm the visual
observations. For both shear layers, the inflection point is present in the velocity profile but only the
shear layer originating at the downstream corner shows a clear vorticity extremum at the same
location of the inflection point. As a result, only this shear will be unstable and only here vortices will
be shed.
93
Bibliography
[1] L. Leopold, M. Wolman and J. Miller, Fluvial processes in geomorphology, Freeman: San Francisco,
1964.
[2] M. Yalin, Mechanics of sediment transport, Oxford, UK.: Pergamon, 1977.
[3] H. Rouse, “Critical analysis of open-channel resistance,” Journal of the Hydraulics Division, vol. 91, no.
4, pp. 1-15, 1965.
[4] A. Kolmogorov, “Dissipation of energy in the locally isotropic turbulence,” Proceedings of the USSR
Academy of Sciences, vol. 32, pp. 16-18, 1941.
[5] V. Chow, “Chapter 8,” in Theoretical concepts of boundary layer, surface roughness,velocity
distribution and instability of uniform flow, 1959, pp. 192-199.
[6] M. Kleinhans, “Hydraulic roughness,” Utrecht University, [Online]. Available:
www.geog.uu.nl/fg/mkleinhans/teaching/rivmorrough.pdf. [Accessed 27 May 2014].
[7] I. Nezu and M. Sanjou, “Turbulence structure and coherent motion in vegetated canopy open-channel
flows,” Journal of Hydro-environment research 2, 2008.
[8] H. Nepf and E. Vivoni, “Flow structure in depth-limited, vegetated flow,” Journal of geophysical
research vol.105, 2000.
[9] M. Ghisalberti and H. Nepf, “The structure of the shear layer over rigid and flexible canopies,”
Environmental Fluid Mechanics, vol. 6, pp. 277-301, 2006.
[10] M. Raupach, J. Finnigan and Y. Brunet, “Coherent eddies and turbulece in vegetation canopies: the
mixing-layer analogy,” Centre for Environmental Mechanics, 1996.
[11] H. Nakagawa and I. Nezu, “Prediction of the contributions to the Reynolds stress from bursting events
in open-channel flows,” Journal of Fluid Mechanics, vol. 80, pp. 99-128, 1977.
[12] N. Kauwen and T. Unny, “Flexible roughness in open channel,” Journal of the Hydraulics Division, vol.
99, no. 5, pp. 713-728, 1973.
[13] C. Wilson, A. Batemann-Pinzen, P. Bates and T. Stoesser, “Open channel flow through different forms
of submerged flexible vegetation,” Journal of Hydraulic Engineering, vol. 129, no. 11, pp. 847-853,
2003.
[14] F. Wu, H. Sheh and Y. Chou, “Variation of roughness coefficients for unsubmerged and submerged
vegetation,” Journal of hydraulic Engineering, pp. 934-942, 1999.
[15] I. Nezu, H. Nakagawa and W. Rodi, “Significant difference between secondary currents in closed
channels and narrow open channels,” pp. A125-A132, 1989.
94
[16] E. Shumate, Experimental description of flow at an open-channel junction, Unpublished master thesis,
University of Iowa, Iowa, 1998.
[17] J. Best and I. Reid, “Separation zone at open channel junctions,” Journal of Hydraulic Engineering,
vol. 110, no. 11, pp. 1588-1594, 1984.
[18] L. Crombé, “Numerieke modellering van een riviersamenvloeiing: Onderzoek naar de invloed van
vegetatie op de stromingspatronen (Unpublished master thesis),” no. University of Ghent, 2012.
[19] S. Creëlle, T. De Mulder, L. Schindfessel and T. Van Oyen, “Influence of hydraulic resistance on flow
features in an open channel confluence: Accepted in de Almeide(ed.),” in Proc. 3rd IAHR Europe
Congress, Porto, 13-17 April 2014.
[20] J. Huang, L. Weber and Y. Lai, “Three dimensional numerical study of flows in open-channel
junctions,” Journal of hydraulic engineering, pp. 268-280, 2002.
[21] C. Hsu, F. Wu and W. Lee, “Flow at a 90° equal-width open-channel junction,” Journal of Hydraulic
Engineering, vol. 124, no. 2, pp. 186-191, 1998.
[22] P. Biron, J. Best and A. Roy, “Effects of bed discordance on flow dynamics at open channel
confluences,” Journal of Hydralic Engineering vol. 122, pp. 676-682, 1996.
[23] P. Drazin and W. Reid, “Introduction to hydrodynamic stability,” pp. 138-180.
[24] J. Bell and R. Metha, “Development of a two-stream mixing layerfrom tripped and untripped boundary
layers,” AIAA journal, vol. 28, no. 12, pp. 2034-2042, 1990.
[25] W. Liou, “Linear instability of curved free shear layers,” Phys. Fluids 6, pp. 541-548, 1993.
[26] H. Fiedler, J.-H. Kim and N. Köpp, “The spatially accelerated mixing layer in a tailored pressure
gradient,” European Journal of Mechanics, vol. 10, no. 4, pp. 349-376, 1991.
[27] V. Chu, J. Wu and R. Khayat, “Stability of transverse shear flows in shallow open channels,” Journal of
Hydraulic Engineering, vol. 127, pp. 351-368, 1991.
[28] R. Fjortoft, “Application of integral theorems in deriving criteria of stability for laminar flows and for
the baroclinic circular vortex,” Geophys. Publ., vol. 17, pp. 1-52, 1950.
[29] M. Khublaryan, “Surface Waters: Rivers, Streams, Lakes and Wetlands.,” in Types and Properties of
Water, Oxford: EOLSS Publischers Co. Ltd, 2009.
[30] R. Keane and R. Adrian, “Theory of cross-correlation analysis of PIV images,” Applied Scientific
Research, pp. 191-215, 1992.
[31] D. G. Goring and V. I. Nikora, “Discussion of "Despiking Acoustic Doppler Velocimeter Data" by Derek
G. Goring and Vladimir I. Nikora,” Journal of Hydraulic Engineering vol. 129, pp. 484-487, 2003.
95
[32]
T. Wahl, “Despiking acoustic Doppler velocimeter Data. discussion,” Journal of Hydraulic Engineering,
vol. 129, no. 6, pp. 484-487, 2003.
[33] E. Mignot, I. Vinkovic, D. Doppler and N. Riviere, “Mixing layer in open-channel junction flows,”
Environmental Fluid Mechanics, 2013.
[34] V. Chu and S. Babarutsi, “Confinement and bed-friction effects in shallow turbulent mixing layers,”
Journal of hydraulic engineering vol.114, pp. 1257-1273, 1988.
[35] Holoborodko, “www.holoborodko.com,” [Online]. Available:
http://www.holoborodko.com/pavel/numerical-methods/numerical-derivative/smooth-low-noise-
differentiators/. [Accessed 3 Mai 2014].
[36] H. Fernando, “Handbook of environmental fluid dynamics, volume one: overview and fundamentals,”
2013, pp. 289-291.
[37] L. Weber, Schumate, E. Shumate and N. Mawer, “Experiments on flow at 90° open-channel junction,”
Journal of Hydraulic Engineering vol.127, pp. 340-350, 2001.
[38] B. Barkdoll, “Discussion of "experiments on flow at an 90° open channel junction" by Larry J. Weber,
Eric D. Shumate, and Nicola Mawer,” Hydraulic Engineering vol. 127, pp. 340-350, 2001.
[39] B. White and H. Nepf, “A vortex based model of velocity and shear stress in a partially vegetated
shallow channel,” Water resources research vol.44, 2008.
[40] B. Stone and H. Shen, “Hydraulic resistance of flow in channels with cylindrical roughness,” Journal of
Hydraulic Engineering vol. 128, pp. 500-506, 2002.
[41] D. Vermaas, W. Uijttewaal and A. Hoitink, “Lateral transfer of streamwise momentum caused by
roughness transition across a shallow channel,” Water resources research vol. 47, 2011.
[42] L. Schindfessel, S. Creëlle, T. Boelens and T. De Mulder, “Flow patterns in an open channel confluence
with a small ratio of main chanel to tributary discharge,” Ghent University, 2014.
[43] H. Fisher and T. Hanamura, “The effect of roughness strips on transverse mixing in hydraulic models,”
Water resources research vol.11 No 1, pp. 362-364, 1975.
[44] D. Goring and V. Nikora, “Despiking Acoustic Doppler Velocimeter Data,” Journal of Hydraulic
Engineering vol.128, pp. 117-126, 2002.
[45] G. Katul, G. Kuhn, J. Schieldge and C. Hsieh, “The ejection-sweep character of scalar fluxes in the
unstable surface layer,” 1996.
[46] M. Muste, I. Fujita and A. Hauet, “Large-scale particle image velocimetry for measurements in riverine
environments,” Water resources research vol. 44, 2008.
[47] K. Bradbrook, S. Lane, K. Richards, P. Brion and A. Roy, “Role of bed discordance at asymmetrical
river confluences,” Journal of Hydraulic Engineering vol. 127, pp. 351-368, 2001.
96
[48] H. Schlichting and K. Gersten, Boundary layer theory.
[49] S. Leonardi, P. Oriandi and R. Antonia, “Properties of d-type and k-type roughness in turbulent
channel flows,” Physics of fluids, 2007.
[50] T. Saad, “Turbulence modelling for beginners,” University of Tenessee Space institute.
[51] S. Otto, T. Jackson and F. Hu, “On the spatial evolution of centrifugal instabilities with curved
incompressible mixing lyaer,” Journal of Fluid mechanics vol. 315, pp. 85-103, 1996.
[52] J. Jarvela, “Flow resistance of flexible and stiff vegetation: af flume study with natural plants,” Journal
of Hydrology vol.269, pp. 44-54, 2002.
[53] B. Yen, “Open channel flow resistance,” Journal of hydraulic engineering, vol. 128, pp. 20-39, 2002.
[54] H. Nepf, “drag, turbulence and diffusion in flow through emergent vegetatoin,” Water resources
research vol.35, 1999.
[55] J. Green, “Modelling flow resistance in vegetated streams: review and development of new theory,”
Hydrological Processes 19, pp. 1245-1259, 2005.
[56] Environmental fluid mechanics chapter 5: Instabilities, Darthmounth college: Thayer school of
engineering, pp. 103-125.
[57] “National committee for fluid mechanics films,” [Online]. Available:
http://web.mit.edu/hml/ncfmf.html. [Accessed 17 March 2014].
[58] A. Galema, “Vegetation resistance evaluation of vegetation resistance descriptors for flood
management (Unpublished master thesis),” 2009.
[59] G. Arcement and V. Schneider, “Guide for selecting Mannings's roughness coefficients for Natural
channels and flood plaines,” United States Geological Survey Water-Supply, 1989.
[60] F. Carollo, V. Ferro and D. Termini, “ Flow resistance law in channels with flexible submerged
vegetation,” Journal of Hydraulic research, vol. 131, pp. 554-564, 2005.
[61] M. A. I. Nezu, “Open-channel Flow Turbulence and its research prospect in the 21st century,” Journal
of Hydraulic Engineering, pp. 229-246, 2005.
[62] P. Davidson, “Turbulence: An introduction for scientists and engineers,” Oxford University Press,
2004, p. 19.
[63] C. Polatel, “Large-Scale roughness effect on free-surface and bulk flo9w characteristics in open-channel
flows,” Iowa, 2006.
Top Related