fluids

Article
PIV Measurements of Open-Channel Turbulent Flow under
Unconstrained Conditions
James K. Arthur

Department of Mechanical Engineering, Bucknell University, Lewisburg, PA 17837, USA;

[email protected]

Abstract: Many open-channel turbulent flow studies have been focused on highly constrained
conditions. Thus, it is rather conventional to note such flows as being fully developed, fully turbulent,
and unaffected by sidewalls and free surface disturbances. However, many real-life flow phenomena
in natural water bodies and artificially installed drain channels are not as ideal. This work is aimed
at studying some of these unconstrained conditions. This is achieved by using particle image
velocimetry measurements of a developing turbulent open-channel flow over a smooth wall. The
tested flow effects are low values of the Reynolds number based on the momentum thickness Reθ
(ranging from 165 to 930), low aspect ratio AR (ranging from 1.1 to 1.5), and Froude number Fr
(ranging from 0.1 to 0.8). The results show that the mean flow has an inner region with a logarithmic
layer with a von Kármán constant of 0.40–0.41, and a log law constant ranging from 5.0 to 6.0. The
friction velocity and coefficient of skin friction are predictable using the formulations of Fr and Reθ
presented in this work. The outer region is also characterized by a dip location, which is predictable
using an equation associated with Reθ . The higher-order turbulence statistics, on the other hand,
show distinguishing traits, such as correlation coefficients ranging from −0.1 to 0.5. Overall, this
work demonstrates that for the unconstrained conditions studied, friction evaluations associated
with Reynolds shear stress and some notable turbulence modelling functions used in conventional
open-channel flows are inapplicable.

Keywords: open-channel flow; particle image velocimetry; turbulence; developing flow; low
Reynolds number; Froude number; low aspect ratio
Citation: Arthur, J.K. PIV
Measurements of Open-Channel
Turbulent Flow under Unconstrained
Conditions. Fluids 2023, 8, 135.
1. Introduction
https://doi.org/10.3390/
fluids8040135 Turbulent flows in open-channels are observed in many applications, ranging from
natural water bodies to artificially installed drains. While some of the general outlines of
Academic Editors: Martin Skote
the flow may be inferred from a theoretical knowledge of classical turbulent boundary
and D. Andrew S. Rees
layers, the flow is in many ways much more complex. This is especially so, given the
Received: 13 March 2023 nature and prevalence of bed conditions, sidewall confinements, and free surface effects.
Revised: 4 April 2023 Consequently, as noted by Nezu [1], several studies have been conducted over several
Accepted: 12 April 2023 decades to understand the flow phenomena. This work focuses on turbulent open-channel
Published: 18 April 2023 flows over smooth walls.
In view of the shear complexity of the flow, experimental research has been far more
valuable in establishing many of the fundamental features of turbulent flow in open-
channel arrangements. This has been made possible through the use of measurement
Copyright: © 2023 by the author.
techniques, such as total head (or Pitot) tubes, propeller current meters, H2 -bubble, hot-film
Licensee MDPI, Basel, Switzerland.
anemometry, laser Doppler anemometry (LDA), and acoustic Doppler velocimetry (ADV).
This article is an open access article
From such tests, several observations about the turbulent boundary layer structure have
distributed under the terms and
conditions of the Creative Commons
been established [2]. Accordingly, we know that the mean streamwise velocity profile of
Attribution (CC BY) license (https://
the turbulent flow in the mid-span consists of an inner region (close to the bed), and an
creativecommons.org/licenses/by/
outer region (further away from the bed). The inner region is also subdivided into a viscous
4.0/). sublayer directly adjacent to the bed where viscous forces are dominant, an intermediate

Fluids 2023, 8, 135. https://doi.org/10.3390/fluids8040135 https://www.mdpi.com/journal/fluids

Fluids 2023, 8, 135 2 of 19

region of transition called the buffer layer, and a fully developed turbulent region where a
universal log law applies. In the outer region, the effects of the flow depth are prominent.
In this zone, the velocity defect profiles tend to be parabolic, and the maximum observed
value is a “dip” below the free surface. A substantial body of the works in this subject
area have been focused on assessing the bed friction parameters and scaling laws, and
developing predictive formulations to forecast flow parameters and profiles [3–6]. These
often emanate from single-point turbulence statistics.
It is worth noting however that many of the concrete observations associated with
turbulent open-channel flows have been based on highly constrained flow set-ups. As
such, it is customary to see data extracted from fully developed regions of the turbulent
flow field, and for conditions where sidewall effects (in the form of secondary currents)
and surface disturbances are minimal. Nonetheless, it must be conceded that many real-
life flow phenomena are far from such idealized constraints. They may, for example, be
subjected to sudden interruptions or short conduit paths that may lead to an undeveloped
flow condition at the location of interest. They may also be under extremely low currents
such that the local Reynolds number could be considerably low. Additionally, flows may
experience extreme confinements, narrow channeling, and wave velocities arising from
disturbances such that the sidewall and surface effects cannot be ignored.
Taking the aforementioned possibilities into account, some researchers have studied
the flow phenomena under less constrained conditions. Sarma et al. [2], for instance,
carried out a wide-ranged experimental program to characterize the flow when the channel
aspect ratio (AR, which is the ratio of width to flow depth) varied from 1.0 to 8.0 and
the Froude number (Fr) varied from 0.2 to 0.7. As their data were obtained from Pitot-
static tube measurements, they only considered mean (streamwise) velocity distribution.
Subsequently, they concluded that Fr and AR had no significant bearing on the form
of the equation for the fully developed velocity distribution in the outer region of the
sidewall close to the bed. Later, Kirkgöz and Ardiclioglu [7] studied the flow for both
developing and developed flows, while accounting for sidewall effects. From their LDA
measurements, they were able to observe that the inner layer of the developing flow also
conformed to the logarithmic law. However, they pointed out that boundary layer along
the channel mid-span develops up to the free surface if the flow AR was equal to three. In
order to evaluate the lower limit of turbulence, Balachandar and Ramachandran [3] also
conducted an extensive set of LDA measurements for flow conditions of a momentum
Reynolds number (Reθ ) ranging from 180 to 480. Their tests confirmed the existence of a
fully developed turbulent boundary layer even under such a low range of Reθ . This is in
stark contrast with the “often quoted lowest estimate of 320” [3]. More recently, Sarkar [8]
and Mahananda et al. [9,10] used ADV to measure the flow in an open-channel flow of a
low aspect ratio (i.e., AR > 2). Of the studies briefly covered in this paragraph, these works
stand out as they are the only ones that have investigated something more than the mean
velocity data. The detailed assessments of Sarkar [8] shows that the turbulent bursting
within the dip of the outer flow showed that the non-dimensional Reynolds shear stress
changes its sign from positive to negative within the dip. Mahananda et al. [9,10], on the
other hand, discovered that the AR and Reynolds number affect the velocity characteristics
in the developing region. However, these works only provide single-point measurements
and they do not consider the influence of Fr.
Thus, in more general terms, while the current knowledge base is significant, it
is limited in two major ways. Firstly, a large proportion of the literature concerning
open-channel turbulent flow has been based on model conditions that do not account
for the possibility of an undeveloped flow, sidewall effects, surface disturbances, and
a low Reynolds number. Consequently, while the study of flows under unconstrained
conditions exists, it is far from the normal stream of research focus, and it requires further
consideration. Secondly, the current information in the literature is largely backed by data
emanating from single-point measurement techniques, and far less comes from whole flow
Fluids 2023, 8, x FOR PEER REVIEW 3 of 20

Fluids 2023, 8, 135 3 of 19

from single-point measurement techniques, and far less comes from whole flow field
measurements. Thus, many studies lack the full picture of detailed multidirectional field
snapshots of measurement.
field To
measurements.
help furnish theThus, many studies lack
above-mentioned the fullinpicture
short-falls of detailed
the literature, multidirectional
the goal of this work
field snapshots of measurement.
is to investigate turbulent open-channel flows under relatively unconstrained conditions.
To help furnish the above-mentioned short-falls in the literature, the goal of this work
What sets this work apart from other studies is that while other works are usually limited
is to investigate turbulent open-channel flows under relatively unconstrained conditions.
to one or two unconstrained conditions, this work aims to consider a much wider range
What sets this work apart from other studies is that while other works are usually limited
of unconstrained conditions, namely, a non-developed flow, a low Reynolds number, a
to one or two unconstrained conditions, this work aims to consider a much wider range
low AR channel, and a low to near-critical Fr. In particular, to the author’s knowledge, the
of unconstrained conditions, namely, a non-developed flow, a low Reynolds number, a
lower limits of the Reynolds number and AR conditions explored in this work are the
low AR channel, and a low to near-critical Fr. In particular, to the author’s knowledge,
lowest recorded in the literature for measurements extracted from flows that are not fully
the lower limits of the Reynolds number and AR conditions explored in this work are the
developed. Indeed, the detailed whole flow field velocity measurements provided
lowest recorded in the literature for measurements extracted from flows that are not fully
through the planar particle image velocimetry technique for such flow conditions are ap-
developed. Indeed, the detailed whole flow field velocity measurements provided through
parently non-existent in the literature. Therefore, with such unique measurements, this
the planar particle image velocimetry technique for such flow conditions are apparently
work aims toinadd
non-existent the significant contributions
literature. Therefore, withtosuch
the unique
database in the literature
measurements, thisby (1) charac-
work aims to
terizing the effects of aspect ratio, Reynolds number, and Froude number of a
add significant contributions to the database in the literature by (1) characterizing the effects narrow
open-channel
of aspect ratio, turbulent
Reynolds flow thatand
number, is not fullynumber
Froude developed; (2) providing
of a narrow ways to
open-channel predict
turbulent
essential parameters; and (3) determining how the flow features compare
flow that is not fully developed; (2) providing ways to predict essential parameters; and with fully de-
veloped open-channel flows.
(3) determining how the flow features compare with fully developed open-channel flows.

2. Experimental System and Measurement Procedure

experimentswere
The experiments wereconducted
conductedatat Bucknell
Bucknell University’s
University’s Fluid
Fluid Flow
Flow andand Experi-
Experimen-
mental Research Laboratory, USA. The experimental set-up consisted of a flume
tal Research Laboratory, USA. The experimental set-up consisted of a flume and a particle and a
particlevelocimetry
image image velocimetry
system. system. A schematic
A schematic diagramdiagram of the arrangement
of the arrangement of the exper-
of the experimental
imentalissystem
system shownisinshown
Figurein1.Figure 1.

(a)Experimental
Figure 1. (a) Experimentalset-up,
set-up,and
and(b)(b) schema
schema of of
thethe
testtest section
section andand
the the coordinate
coordinate system
system em-
employed in this
ployed in this work.
work. TheThe green
green shaded
shaded section
section in (b)
in (b) covers
covers thethe focus
focus of PIV
of PIV measurement.
measurement. All All
di-
mensions areare
dimensions in in
millimeters. TheThe
millimeters. symbols x, y,x,and
symbols z represent
y, and the directions
z represent alongalong
the directions the stream, wall-
the stream,
normal, and the
wall-normal, andspan, respectively.
the span, respectively.

The flume used in this work is an open flow transport channel having a test section of
rectangular cross-section. The test section is equipped with transparent acrylic walls, and
Fluids 2023, 8, 135 4 of 19

of approximate dimensions of length L = 2.5 m, width b = 0.08 m, and depth H = 0.25 m.

The system works in a closed recirculation, pumping water from a reservoir through a
flow conditioner to the test section, and then back to the reservoir [11]. For the purpose of
this research, certain modifications were made to the channel. The flow development of
the turbulent boundary layer was enhanced using a triad of 9 mm-diameter rods fixed at
the entry to the test section to serve as trips. In order to reduce reflection during velocity
measurements, a dark polyvinyl chloride (PVC) plate was glued onto the bottom wall of
the flume. This plate provided the smooth bed surface required for the tests. By utilizing a
gate close to the exit of the test section, various depths of flow (55 mm< h < 75 mm) were
achieved at a constant bed slope. This ensured that the aspect ratio (AR = b/h) of the channel
flow was sufficiently low, thus maintaining a narrow channel condition (i.e., AR < 5 [1])
throughout the testing process. As shown in Figure 1, the coordinate system adopted in this
work is fixed relative to the flume. The directions x, y, and z represent the directions along
the stream, wall-normal, and the span, respectively. The origin of the coordinate system is
specified such that x = 0 is fixed at the entrance into the test section of the flume, and y = 0
and z = 0 are, respectively, on top of the PVC plate and in the mid-span of the channel.
Velocity measurements were obtained using a planar particle image velocimetry (PIV)
system entailing a laser, camera, programable timing unit, and a computer. As each of
these PIV components utilized in this work is identical to that described in a previous
publication [12], details regarding specifications will not be repeated here. However, it is
important to note that for this work, the camera body was attached to a 28 mm focal length
Nikkor lens. The camera assembly was in turn fitted with an orange filter with a band-pass
wavelength of 532 nm ± 10 nm. For the PIV system, the laser and the camera were attached
to a translation stage. In this way, both could be traversed as a unit in the streamwise
direction without changing the distance between them. The flow was seeded with silver-
coated hollow glass spheres with a mean diameter of 10 µm and a specific gravity of 1.4.
Velocity measurements were accomplished by illuminating the seeded flow with pulses of
a laser sheet of approximately 1 mm thickness and following precautions similar to those
stated in the references [12]. For each PIV measurement, a total of 4000 instantaneous pairs
of images were recorded with the camera. The images were then transferred to the computer
and processed using specialized PIV software (DaVis-10.2). During data processing, the
interrogation area was initially set to a size of 128 pixels × 128 pixels. It was thereafter run
through several steps that are iterative. After undergoing an outlier-removal validation
step, each interrogation window was ultimately subdivided into 32 pixels × 32 pixels, with
a 75% overlap set between immediate interrogation areas. Consequently, the distances
in both x and y directions between immediate vectors in physical units are identical at
0.27 mm. With such a resolution, it was possible to obtain data in the y direction within
sufficiently small wall units.
The measurements obtained through PIV and reported herein are time-averaged
velocities in the streamwise and wall-normal directions (i.e., U and V, respectively), turbu-
lence intensities in the streamwise and wall-normal directions (i.e., u and v, respectively),
Reynolds normal stresses in the streamwise and wall-normal directions (i.e., u2 and v2 ,
respectively); and Reynolds shear stress −uv. The uncertainties of these parameters were
assessed in a manner similar to that outlined by Wieneke [13]. They were estimated at 95%
confidence level, and found to be ±1.8%, ±2.3%, ±2.5%, and ±3.5% of the respective peak
values of (U, V), (u, v), (u2 , v2 ), and −uv.
The test conditions are summarized in Table 1. For the present work, PIV measure-
ments were obtained at z = 0 for an array of x-y measurement planes. The streamwise
locations covered within the measurement planes ranged from 1.18 m to 1.59 m. In order to
compare the boundary layer profiles for all test conditions, the boundary layer parameters
utilized in Table 1 were obtained from data extracted from the same streamwise location
(x = 1.235 m). For the substantive tests, the principal aim was to conduct an extensive set of
measurements to assess the effects of AR, the Reynolds number based on the momentum
thickness (Reθ = Ue θ/ν, where Ue , θ, and ν are, respectively, the maximum streamwise
Fluids 2023, 8, x FOR PEER REVIEW 5 of 20

cation (x = 1.235 m). For the substantive tests, the principal aim was to conduct an exten-
Fluids 2023, 8, 135 sive set of measurements to assess the effects of AR, the Reynolds number based on 5 ofthe
19
momentum thickness (Re = Ue 𝜃/ν, where Ue, 𝜃, and ν are, respectively, the maximum
streamwise mean velocity, momentum thickness of the boundary layer, and kinematic
viscosity
mean of themomentum
velocity, flow), andthickness
the Froude number
of the boundary Um √(𝑔ℎ)
(Fr =layer, where Uviscosity
and kinematic m is the depth-
of the
averaged mean streamwise velocity, and √𝑔 is the acceleration due to gravity). Thus, the
flow), and the Froude number (Fr = Um ( gh) where Um is the depth-averaged mean
tests have been
streamwise named
velocity, to gindicate
and the values ofdue
is the acceleration each
to of these three
gravity). Thus,parameters. All in
the tests have beenall,
13 rounds
named of tests were
to indicate performed
the values of eachover the range
of these three1.1 < AR < 1.5,
parameters. in <all,𝑅𝑒13 rounds
All128 < 965, and 0.12
of tests
< Fr performed
were < 0.77. over the range 1.1 ≤ AR ≤ 1.5, 128 ≤ Reθ ≤ 965, and 0.12 ≤ Fr ≤ 0.77.

Table1.
Table 1. Summary
Summary of
of test
test conditions
conditionsand
andrelevant
relevantboundary
boundarylayer
layerparameters.
parameters.

Streamwise
Streamwise Depth-Aver- Momentum
Momentum Boundary
Depth of Depth- Maximum Boundary
Length
LengthRange in
Depth of Aspect aged Froude Maximum Thickness
Thickness Layer Shape
Test Name
Test Name Range in
flow,
flow,
Aspect Averaged Froude Velocity,
Velocity, Reynolds
Layer Shape
Measurement Ratio, AR
Ratio, AR Velocity,
Velocity, Number,
Number, Fr
Fr Reynolds Parameter
Parameter
Measurement h(m)
h (m) e (m/s)
UU e (m/s) Number,
Plane (m) Umm (m/s)
(m/s) Number, Reθ ℌ
Plane (m) Reθ
AR1.1Re447Fr0.18 1.34–1.43 0.072 1.15 0.149 0.18 0.160 447 1.49
AR1.1 Re447 Fr0.18 1.34–1.43 0.072 1.15 0.149 0.18 0.160 447 1.49
AR1.2Re527Fr0.18
AR1.2 Re527 Fr0.18
1.34–1.43
1.34–1.43
0.070
0.070
1.18
1.18
0.151
0.151
0.18
0.18
0.163
0.163
527
527
1.41
1.41
AR
AR Re
1.41.4 Re FrFr
512
512 0.23
0.23 1.34–1.43
1.34–1.43 0.060
0.060 1.38
1.38 0.176
0.176 0.23
0.23 0.189
0.189 512
512 1.44
1.44
1.18–1.27
1.18–1.27
AR
AR Re
1.51.5 Re
165 FrFr
165 0.13
0.13 0.055
0.055 1.50
1.50 0.093
0.093 0.13
0.13 0.099
0.099 165
165 1.88
1.88
1.50–1.59
1.50–1.59
AR1.5 Re281 Fr0.19 1.34–1.43 0.055 1.50 0.139 0.19 0.147 281 1.58
AR1.5Re281Fr0.19 1.34–1.43
1.34–1.43 0.055 1.50 0.139 0.19 0.147 281 1.58
AR1.3 Re452 Fr0.17 0.055 1.50 0.200 0.27 0.211 383 1.56
1.34–1.43
1.50–1.59
AR1.3
AR Re452Fr0.17 0.055 1.50 0.200 0.27 0.211 383 1.56
1.3 Re729 Fr0.42 1.34–1.43
1.50–1.59 0.065 1.27 0.132 0.16 0.167 452 1.50
AR1.3 Re703 Fr0.49 1.34–1.43 0.065 1.27 0.332 0.42 0.346 729 1.38
AR1.3Re729Fr0.42
AR1.3 Re836 Fr0.58
1.34–1.43
1.34–1.43
0.065
0.065
1.27
1.27
0.132
0.393
0.16
0.49
0.167
0.410
1.50
1.42
452
703
AR
AR1.31.3
Re Re 703Fr0.49
928 Fr0.62
1.34–1.43
1.34–1.43 0.065
0.065 1.27
1.27 0.332
0.462 0.42
0.58 0.346
0.482 1.38
1.42 729
836
AR
AR Re
1.31.3 Re FrFr
836
864 0.58
0.63 1.34–1.43
1.34–1.43 0.065
0.065 1.27
1.27 0.493
0.393 0.62
0.49 0.514
0.410 1.38
1.42 928
703
AR
AR Re
1.31.3 Re Fr0.32
928Fr
650 0.62
1.34–1.43
1.34–1.43 0.065
0.065 1.27
1.27 0.505
0.462 0.63
0.58 0.526
0.482 1.43
1.42 864
836
AR1.5 Re829 Fr0.77 1.34–1.43 0.065 1.27 0.253 0.32 0.270 1.38 650
AR1.3Re864Fr0.63 1.34–1.43 0.065 1.27 0.493 0.62 0.514 928 1.38
AR1.3Re650Fr0.32 1.34–1.43 0.065 1.27 0.505 0.63 0.526 864 1.43
AR1.5Re829Fr0.77 1.34–1.43 The
0.065 Reθ parameters
AR and1.27 0.253 tested 0.32
here are considerably
0.270 low
650[3,8]. Indeed,
1.38 their
lower limits rank among the lowest recorded in the literature for this class of turbulent
flow.The However,
AR andthe Revalues are modified
parameters tested to testare
here any discernible low
considerably effects of Indeed,
[3,8]. low AR their and
Re θ . Regarding
lower limits rank theamong
Reynolds numbers
the lowest in particular,
recorded it is significant
in the literature to class
for this pointof out that an
turbulent
assessment
flow. However, of thetheReynolds
values are numbers
modified based onany
to test the discernible
maximum effects
streamwise
of lowmeanAR and 𝑅𝑒 .
velocity
and the flow depth
Regarding the Reynolds (Re ) indicates that the present range is 21,700 < Re < 136,700.
h numbers in particular, it is significant to pointhout that an assess- This is
substantially wide and comparable with other measurements made
ment of the Reynolds numbers based on the maximum streamwise mean velocity and the within the turbulent
flow regime.
depth (𝑅𝑒 The Fr valuesthat
) indicates considered
the presentin this
rangework are also
is 21,700 𝑅𝑒 <within
< well 136,700.theThis
subcritical
is sub-
range.
stantiallyNonetheless, they cover low
wide and comparable withtoother
moderately high numbers
measurements that may
made within the give insight
turbulent to
flow
the effects
regime. The ofFr
thevalues
onset considered
of gravity on in the
thisfree
work surface of the
are also wellopen-channel flow. Finally,
within the subcritical range. it
is pointed out that despite tripping the flow, the flow in this measurement
Nonetheless, they cover low to moderately high numbers that may give insight to the section is not
necessarily
effects of the expected
onset oftogravity
be fullyondeveloped. Indeed,ofasthe
the free surface confirmed in the flow.
open-channel following section,
Finally, it is
the flow is still in the flow development regime, thus suiting an objective
pointed out that despite tripping the flow, the flow in this measurement section is not of presenting
results for aexpected
necessarily developing open
to be narrow
fully channel
developed. flow. as confirmed in the following section,
Indeed,
the flow is still in the flow development regime, thus suiting an objective of presenting
3. Results and Discussion
results for a developing open narrow channel flow.
In this section, the results are presented by evaluating the extent of the flow develop-
ment, and the
3. Results andcharacteristics
Discussion of the mean and fluctuating flow. To facilitate a comparison of
the flow features in Sections 3.2 and 3.3, all the flow profiles presented in those sections are
In this section, the results are presented by evaluating the extent of the flow develop-
extracted at x = 1.235 m.
ment, and the characteristics of the mean and fluctuating flow. To facilitate a comparison
of the
3.1. flow
Flow features in Sections 3.2 and 3.3, all the flow profiles presented in those sections
Development
are extracted at x = 1.235 m.
In order to show the nature of flow development for the present tests, a wide range
of the flow data are extracted from representative measurements taken during test con-
3.1. Flow Development
ditions AR1.5 Re165 Fr0.13 and AR1.3 Re452 Fr0.17 . According to Kirkgöz and Ardiclioglu [7],
the expected flow developing zone for the range of flow conditions covered herein should
be between 51 h and 60 h. However, the current data were taken much further upstream
(16 < x/h < 29). Thus, the following section presents unique and detailed insight into the
nature of the flow development (if any) of a narrow open-channel flow prior to the develop-
ing zone. The results of the mean velocities and turbulent statistics normalized by the local
 In order to show the nature of flow development for the present tests, a wide range
of the flow data are extracted from representative measurements taken during test condi-
tions AR1.5Re165Fr0.13 and AR1.3Re452Fr0.17. According to Kirkgöz and Ardiclioglu [7], the ex-
pected flow developing zone for the range of flow conditions covered herein should be
between 51 h and 60 h. However, the current data were taken much further upstream (16
Fluids 2023, 8, 135 < x/h < 29). Thus, the following section presents unique and detailed insight into 6the of 19
nature
of the flow development (if any) of a narrow open-channel flow prior to the developing
zone. The results of the mean velocities and turbulent statistics normalized by the local
maximummaximum velocity
velocity are plotted
are plotted in Figures
in Figures 2 and23.and 3. They
They showshow that while
that while the streamwise
the streamwise
components of the of
components mean
the velocities and turbulent
mean velocities intensities
and turbulent appear to
intensities be independent
appear of the of
to be independent
streamwise location, location,
the streamwise the wall-normal components
the wall-normal are not. Additionally,
components the wall-normal
are not. Additionally, the wall-nor-
Reynolds stress profiles
mal Reynolds stressatprofiles
increasing streamwise
at increasing locations locations
streamwise do not converge. While theWhile
do not converge.
spanwise components were not measured (due to the limitations of the PIV
the spanwise components were not measured (due to the limitations of the PIV arrange- arrangement),
it is reasonable
ment), it is to reasonable
expect that atto the locations
expect that atof the
measurement, those
locations of velocities would
measurement, thoselikely
velocities
be dependent
would likelyon the
be streamwise
dependent on location as well. Consequently,
the streamwise Figures
location as well. 2 and 3 confirm
Consequently, Figures 2
that the
andflow is not fully
3 confirm developed
that the [10].fully developed [10].
flow is not

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
1.0 1.0

U/Ue (a) U/Ue (b)

0.5 0.5

x/h = 22 x/h = 28 x/h = 25 x/h = 28

0.0 0.0
(c) (d)
0.000 0.000
V/Ue V/Ue

−-0.025 −-0.025
0.2 0.2
(e) (f)
u/Ue u/Ue
0.1 0.1

0.0 0.0

0.05
(g) 0.05
(h)
v/Ue v/Ue

0.00 0.00
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
y/h y/h

FigureFigure 2. Wall-normal
2. Wall-normal variations
variations of streamwise
of streamwise mean mean velocities
velocities (a,b), wall-normal
(a,b), wall-normal mean mean velocities
velocities
(c,d), streamwise turbulence intensities (e,f), and wall-normal turbulence intensities (g,h). TheseThese
(c,d), streamwise turbulence intensities (e,f), and wall-normal turbulence intensities (g,h).
profiles are extracted
at x/h =at22,
x/h28= for
22, test
28 for test condition
AR1.5 ReAR 1.5Re165Fr0.13 (shown in (a,c,e,g) and x/h = 25,
profiles are extracted condition 165 Fr0.13 (shown in (a,c,e,g) and x/h = 25,
28 for test condition AR1.3Re452Fr0.17 (shown in (b,d,f,h)). All velocities are normalized by the local
28 for test condition AR1.3 Re452 Fr0.17 (shown in (b,d,f,h)). All velocities are normalized by the local
maximum velocity, and the lengths are normalized by the respective depth of the flow.
maximum velocity, and the lengths are normalized by the respective depth of the flow.

Apart from this foregoing conclusion, it is important to note some particular traits.
The results indicate that at a higher Reynold’s number, the deviation of the wall-normal
mean velocities from full development is much more apparent. In contrast, the wall-
normal turbulent intensities tend toward streamwise independence as Reynolds number
increases. While subject to confirmation, it is reasonable to speculate that the lack of flow
development in the wall-normal mean and turbulence quantities may be the result of strong
three-dimensional effects. This is very likely due to the extreme narrowness of the channel.
The wall-normal intensities, on the other hand, may not be affected by such wall-effects.
In Figure 4, the levels of turbulence decay and turbulence anisotropy during the
development of the flow are presented. This is achieved using data extracted from the
mid-depth of the flow (i.e., y = h/2). The profiles of turbulence decay in the streamwise
direction (U/u)2 and that in the wall-normal directions (U/v)2 show that the latter is
significantly higher than the former. These indicate that streamwise development of
flow tends to dampen the large scales of the flow. It is important to point out that the
 Fluids 2023, 8, 135 7 of 19

Fluids 2023, 8, x FOR PEER REVIEW 7 of 20

current values of these turbulence intensity ratios are at least two times those reported by
Mahananda et al. [10]. This may be attributed to the presence of bed roughness in the latter
work, which could inhibit the decay of turbulence in a developing flow.

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

0.002 x/h = 22 x/h = 28 0.002 x/h = 25 x/h = 28

0.001
2
(a) 0.001 (b)
−uv/Ue −-uv/Ue
2

0.000 0.000

0.02 0.02
0.01 (c) 0.01
(d)
u2/Ue2 u2/Ue2
0.00 0.00

-0.01
−0.004 −-0.01
0.004
(e) (f)
v2/Ue2 0.002 v2/Ue2 0.002

0.000 0.000
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
y/h y/h

Figure
Figure3.3.Wall-normal
Wall-normal variations of of Reynolds
Reynoldsshear
shearstresses
stresses(a,b),
(a,b),streamwise
streamwise components
components of Reyn-
of Reynolds
olds stresses
stresses (c,d),(c,d), and wall-normal
and wall-normal components
components of Reynolds
of Reynolds stresses
stresses (e,f). profiles
(e,f). These These profiles are ex-at
are extracted
Fluids 2023, 8, x FOR PEER REVIEW tracted at x/h = 22, 28 for test condition AR 1.5Re165Fr0.13 (shown in (a,c,e)) and x/h = 25, 28 for test
x/h = 22, 28 for test condition AR1.5 Re165 Fr0.13 (shown in (a,c,e)) and x/h = 25, 28 for test condition 8 of 20
condition AR1.3Re452Fr0.17 (shown in (b,d,f)). All velocities are normalized by the local maximum ve-
AR1.3 Re452 Fr0.17 (shown in (b,d,f)). All velocities are normalized by the local maximum velocity, and
locity, and the lengths are normalized by the respective depth of the flow.
the lengths are normalized by the respective depth of the flow.

Apart from this foregoing conclusion, it is important to note some particular traits.
The results indicate that at a higher Reynold’s number, the deviation of the wall-normal
mean velocities from full development is much more apparent. In contrast, the wall-nor-
mal turbulent intensities tend toward streamwise independence as Reynolds number in-
creases. While subject to confirmation, it is reasonable to speculate that the lack of flow
development in the wall-normal mean and turbulence quantities may be the result of
strong three-dimensional effects. This is very likely due to the extreme narrowness of the
channel. The wall-normal intensities, on the other hand, may not be affected by such wall-
effects.
In Figure 4, the levels of turbulence decay and turbulence anisotropy during the de-
velopment of the flow are presented. This is achieved using data extracted from the mid-
depth of the flow (i.e., y = h/2). The profiles of turbulence decay in the streamwise direction
(U/u)2 and that in the wall-normal directions (U/v)2 show that the latter is significantly
higher than the former. These indicate that streamwise development of flow tends to
dampen the large scales of the flow. It is important to point out that the current values of
these turbulence intensity ratios are at least two times those reported by Mahananda et al.
[10]. This may be attributed to the presence of bed roughness in the latter work, which
could inhibit the decay of turbulence in a developing flow.

Figure
Figure 4. 4.
Streamwise
Streamwise variations of of
variations (a)(a)
ratio of of
ratio mean
mean streamwise velocities
streamwise to to
velocities streamwise
streamwiseturbulence
turbulence
intensities;
intensities; (b)(b)
ratio ofof
ratio mean
mean streamwise
streamwisevelocities
velocitiestotowall-normal
wall-normalturbulence
turbulenceintensities;
intensities;and
and(c)
(c) ra-
ratio
tioofofstreamwise
streamwiseturbulence
turbulence intensities to wall-normal turbulence intensities. The plots
intensities to wall-normal turbulence intensities. The plots are extractedare ex-
tracted from the mid-depth of the flow.
from the mid-depth of the flow.

An examination of the turbulence flow mid-depth anisotropies (u/v) in Figure 4 also

shows that the streamwise turbulence intensities are 12–22% larger than the wall-normal
components. The global and local variations relative to the streamwise direction are sig-
nificant, with the latter trending in the direction of an increasing anisotropy along the
Fluids 2023, 8, 135 8 of 19

An examination of the turbulence flow mid-depth anisotropies (u/v) in Figure 4

also shows that the streamwise turbulence intensities are 12–22% larger than the wall-
normal components. The global and local variations relative to the streamwise direction
are significant, with the latter trending in the direction of an increasing anisotropy along
the stream. These observations are important as they signify an evolving process of
energy transfer to the streamwise turbulence intensities due to the energy cascade. Again,
comparing this work with Mahananda et al. [10], it is clear that the addition of roughness
elements can lead to an enhancement of the anisotropy levels of a developing flow in a
narrow channel by over 50%.

3.2. Characteristics of the Mean Flow

In characterizing the flow, attention is first turned to the results of the mean flow
presented in Figures 5–8 and Table 2. The developing boundary layer of the narrow channel
studied in this work has certain features similar to those of other open-channel flows. This
is already demonstrated in Figure 2. Consequently, the flow may be properly stratified into
the conventional inner and outer regions, and analyzed as such. Plots of the streamwise
mean velocity in inner wall units (i.e., U+ = U/ Uτ and y+ = y Uτ /ν, for a characteristic
friction velocity Uτ ) in Figure 5, show the presence of a portion of a viscous sublayer that
appears to conform to the following well-known law of the wall:

U + = y+ (1)

This law is accepted to be valid up to y+ ≈ 5 in other turbulent boundary layer flows.

Beyond y+ ≈ 5, however, a logarithmic layer exists. Like other turbulent boundary layers,
each test condition in this work is found to have a logarithmic layer following the classical
logarithmic (log) law, namely:
1
U + = ln y+ + B (2)
κ
A von Kármán constant κ ranging from 0.40 to 0.41, and a logarithmic law constant
B from 5.0 to 6.0 both satisfy this law for the current test data. The consistency is so
remarkable that when using these constants in a Clauser plot technique, the optimal values
of Uτ may be obtained within a maximum relative deviation of 3.5%. For brevity, however,
only the friction parameters obtained from Equation (2) with constants κ = 0.41 and B = 5.0
are presented in Table 2 and analyzed hereafter. In general, the logarithmic layer is found
to fall within a wall-normal range (Ly+ ) of 30 < y+ < 300. For the tests with the lowest
Reynolds number (i.e., AR1.5 Re165 Fr0.13 and AR1.5 Re281 Fr0.19 ), the upper limit of this layer
is much less than y+ = 300, but such that y ≈ 0.2 h. The logarithmic layer tends to diminish
(<15% of the flow depth) with a decreasing Reynolds number. However, as shown in
Figure 5, this is not a monotonic trend. The current tests indicate that irrespective of AR,
the thickness of the layer for a low Reynolds number flow peaks to over 45% of the flow
depth, at around 350 < Reθ < 550, which is then followed by a decline and a levelling off
with an increasing Reynolds number.
The log-linear plots in Figure 5 clearly show that like other boundary layer flows,
Uτ is an important wall parameter with dominant effects close to the bed of the channel.
Nevertheless, it must be emphasized that Uτ is not only dependent on the local conditions.
On the contrary, there is evidence of a direct dependence of the outer layer effects of the
flow surface, such that Uτ may be predicted by the Froude number Fr. Consequently, a
friction Reynolds number Reτ defined by the friction velocity Uτ and the flow depth h,
exhibits a strong dependency on Fr. The current data plotted in Figure 6 show that the
relationships are linear, and defined by the following:

Uτ = 0.0425 Fr (3)

and
Reτ =2645Fr (4)
 found to fall within a wall-normal range (Ly+) of 30 < y+ < 300. For the tests with the lowest
Reynolds number (i.e., AR1.5Re165Fr0.13 and AR1.5Re281Fr0.19), the upper limit of this layer is
much less than y+ = 300, but such that y ≈ 0.2 h. The logarithmic layer tends to diminish
(<15% of the flow depth) with a decreasing Reynolds number. However, as shown in Fig-
Fluids 2023, 8, 135 ure 5, this is not a monotonic trend. The current tests indicate that irrespective of9 of
AR,19 the
thickness of the layer for a low Reynolds number flow peaks to over 45% of the flow depth,
at around 350 < 𝑅𝑒𝜃 < 550, which is then followed by a decline and a levelling off with an
increasing
with adjusted R2 values
Reynolds number.
of 0.997 and 0.985, respectively.

30
(a) AR1.1Re447Fr0.18 AR1.2Re527Fr0.18 AR1.4Re512Fr0.23
AR1.5Re165Fr0.13 AR1.5Re281Fr0.19 AR1.5Re383Fr0.27
25 AR1.3Re452Fr0.17 AR1.3Re729Fr0.42 AR1.3Re703Fr0.49
AR1.3Re836Fr0.58 AR1.3Re928Fr0.62 AR1.3Re864Fr0.63
AR1.3Re650Fr0.32 AR1.5Re829Fr0.77
20
Log Law Law of the Wall

U+ 15

10

0
0.01 0.1 1 10 100 1000
+
y

(b) 0.5

0.4

0.3
Ly+/h

0.2

0.1

0.0
0 200 400 600 800 1000
Req
Figure 5. (a) Logarithmic plots of streamwise velocities using inner wall coordinates, along with the
law of the wall, and the log law (Equations (1) and (2), respectively, for AR1.2 Re527 Fr0.18 ). (b) The de-
pendence of the log layer wall-normal range (Ly+ ) normalized by the flow depth h on the momentum
thickness Reynolds number (Reθ ).

Another friction parameter considered in this work is the skin friction coefficient Cf .
This is computed from the mean velocity data as two (Uτ /Ue )2 . In order to provide a
comparison of the friction parameters with those reported in other works, the skin friction
coefficient is plotted in Figure 6 along with those published in several other open-channel
flow studies. The data compared are derived from Uτ measurements in the mid-span plane.
As depicted in the figure, the variation of Cf . with Reθ trends is similar to other works,
irrespective of the differences in the aspect ratio of those works. Thus, mid-span values of
Cf are not affected by the aspect ratio of the channel. Indeed, like other turbulent boundary
 5 of 20

For the substantive tests, the principal aim was to conduct an exten-
ments to Fluids
assess the8,effects
2023, 135 of AR, the Reynolds number based on the 10 of 19
ss (Re = Ue 𝜃/ν, where Ue, 𝜃, and ν are, respectively, the maximum
elocity, momentum thickness of the boundary layer, and kinematic
w), and the Froude number (Fr = Um √(𝑔ℎ) where
layer flows, Umcoefficient
this is the depth-
tends to increase logarithmically with decreasing Reθ . However,
amwise velocity, and 𝑔 is the acceleration due to gravity). Thus, the
in the case of open-channel flows, this coefficient may be predicted up to Reθ of about
ed to indicate the values of each of these three
15,000 parameters.
through All in all,
the following equation:
ere performed over the range 1.1 < AR < 1.5, 128 < 𝑅𝑒 < 965, and 0.12
C f = −0.00755 ln[0.07521 ln ( Reθ )] (5)

est conditions and relevant boundary layerTable

parameters.
2. Summary of mean flow parameters.
Depth-Aver- Momentum Boundary
Boundary
Maximum Friction Skin Friction Cole’s Wake
ct aged Froude Thickness Layer Shape
Layer Shape
Velocity, Test Name Velocity Coefficient ymax /h Parameter
AR Velocity, Number, Fr Reynolds Parameter
Ue (m/s) Uτ (mm/s) Cf ×10−3 ∏
Um (m/s) Number, Reθ ℌ
0.149 0.18 0.160 AR Re 447Fr 1.49
1.49 8.6 5.81 0.60 −0.049
1.1 447 0.18
0.151 0.18 0.163 AR1.2 Re527
527Fr0.18 1.41
1.41 8.7 5.60 0.69 0.018
0.176 0.23 0.189 AR1.4 Re512
512Fr0.23 1.44 10.0 5.63 0.71 0.009
AR1.5 Re165 Fr0.13 1.88 6.1 7.72 0.44 −0.369
0.093 0.13 0.099 165 1.88
AR1.5 Re281 Fr0.19 1.58 8.5 6.73 0.53 −0.076
0.139 0.19 0.147 AR1.5 Re383
281Fr0.27 1.56
1.58 11.6 5.98 0.55 −0.020
AR1.3 Re452 Fr0.17 1.50 9.0 5.80 0.59 0.004
0.200 0.27 0.211 AR1.3 Re729
383Fr0.42 1.56
1.38 18.2 5.57 0.79 −0.380
AR1.3 Re703 Fr0.49 1.42 21.5 5.47 0.72 −0.388
0.132 0.16 0.167 452 1.50
AR1.3 Re836 Fr0.58 1.42 25.0 5.39 0.78 −0.476
0.332 0.42 0.346 AR Re 729Fr 1.38
1.38 26.3 5.25 0.84 −0.398
1.3 928 0.62
0.393 0.49 0.410 AR1.3 Re864
703Fr0.63 1.42
1.43 26.9 5.25 0.78 −0.460
0.462 0.58 0.482 AR1.3 Re650
836Fr0.32 1.42
1.38 14.2 5.55 0.79 −0.278
0.493 0.62 0.514 AR1.5 Re829
928Fr0.77 1.38
1.42 30.3 5.30 0.88 −0.335
0.505 0.63 0.526 864 1.43
0.253 0.32 0.270 650 2 1.38
The adjusted R for such a fit is 0.979. In 1999, Tachie et al. [14] noted subtle dif-
ferences between Cf obtained from open-channel flows and others, such as wind tunnel
e parameters tested here are considerably low [3,8]. Indeed, their
measurements, or simulations of classical turbulent boundary layer flows. With a wider
mong the lowest recorded in the literature for this class of turbulent
range of published data (such as that provided in this work), more precise confirmations
values are modified to test any discernible effects of low AR and 𝑅𝑒 .
of the differences may be made. When compared with other turbulent boundary layer
olds numbers in particular, it is significant to point
fits, such out that
as that an assess-
prescribed by Österlund [15] and Schlatter et al. [16], the Cf values from
s numbers based on the maximum streamwise mean velocity and Rethe
Equation (5) converge as θ decreases and appear to diverge beyond Reθ ≈ 1000. While
dicates that the present range is 21,700this < 𝑅𝑒 < 136,700. This is sub-
observation requires further investigation, it is reasonable to infer that such a difference
omparable with other measurements made is likelywithin the turbulent
associated with the flow
non-local (outer layer) free surface effects on the friction.
es considered in this work are also well within the subcritical range.
For the outer flow, some of the most consequential parameters are the maximum
over low to moderately high numbersstreamwisethat may give meaninsight
velocityto U the, the wall-normal location of that maximum velocity y ,
e max
of gravity on the free surface of the open-channel flow. Finally,
and the flow depth h. The occurrence it is of the maximum velocity below the flow depth
spite tripping the flow, the flow in this measurement
(i.e., dip phenomenon) section is not
is characteristic of narrow open-channel flows. This phenomenon is
to be fully developed. Indeed, as confirmed
shownin inthe following
Figure section,
7 in terms of ymax /h and a dip correction factor α = h/ymax − 1. The data
he flow development regime, thus suiting showan objective
that of presenting
both parameters are functions of the Reynolds number Reθ . This is an important
ing open narrow channel flow. finding, as previous works have only noted that the dip correction factor is mainly affected
by the spanwise location of the measurement, and hitherto the other affecting factors have
ussion not been explored. However, with the data presented herein, we can now see that for
he results are presented by evaluating ythe /h or α,
maxextent ofthe
therelationship
flow develop- with Reθ is significant, linear, and even predictable over the
narrow
cteristics of the mean and fluctuating flow. To range of Reynolds
facilitate a comparison number studied in this work through the following equation:
n Sections 3.2 and 3.3, all the flow profiles presented in those sections
.235 m. α = 1.267 − 0.0924Reθ (6)

The adjusted R2 for such a fit is 0.857. Furthermore, the trend (see Table 2, Figure 7)
t
indicates that regardless of the ultra-narrowness of the channel, the dipping phenomenon
is expected to reduce with increasing Reθ . The scatter in the data at high Reynolds number
is patent. However, this may be the result of discernible effects of medium Fr number.
 range of published data (such as that provided in this work), more precise confirmations
of the differences may be made. When compared with other turbulent boundary layer fits,
such as that prescribed by Ö sterlund [15] and Schlatter et al. [16], the Cf values from Equa-
tion (5) converge as Re𝜃 decreases and appear to diverge beyond Re𝜃 ≈ 1000. While this
Fluids 2023, 8, 135 observation requires further investigation, it is reasonable to infer that such a difference
11 of 19
is likely associated with the non-local (outer layer) free surface effects on the friction.

(a) 1800
Ret 30
1600 Ut

1400 25

1200 20

Ret 1000 Ut
15

800
10
600

5
400

200 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Fr
8.0
(b) Schultz and Swain
Tachie et al. I
7.0 Tachie et al. II
Tachie et al. III
Das et al
6.0
Österlund
Schlatter et al.
5.0 Current Work
Cf × 10-3

4.0

3.0

2.0

1.0

0.0
0 10,000 20,000 30,000

Req
Figure
Figure6.6.(a)
(a)Dependence
Dependenceofoffriction
frictionvelocity
velocityUUτ and
τ and friction
friction Reynolds
Reynolds numberReRe
number 𝜏 τon onthe
theFroude
Froude
number
numberFr. Fr.(b)
(b)Skin
Skinfriction
frictioncoefficient
coefficientCCf values
f valuesof ofcurrent
current tests
tests compared
compared with
withother
other open-channel
open-channel
flow
flowstudies
studies(Schultz
(SchultzandandSwain
Swain[17];
[17];Tachie
Tachieetetal.al.I [4],
I [4],IIII[6],
[6],IIIIII[18]
[18]and;
and;Das
Dasetetal.al.[19])
[19])and
andother
other
boundary layer studies (Ö sterlund [15] and Schlatter et al. [16]).
boundary layer studies (Österlund [15] and Schlatter et al. [16]).

For
Thethe outer flow,
deviation some
of the outerofflow
the most consequential
from the parameters
log law is often are the by
characterized maximum
the wake
streamwise
parameter mean velocity
∏. Some Ue, the wall-normal
researchers location of such
[20] have incorporated that maximum velocity
a parameter ymax, and
in a streamwise
mean velocity defect profile given by the following:
     y 2    y 3 
U Uτ y
= 1+ ln − (1 + 6 ∏ ) 1 − + (1 + 4 ∏ ) 1 − (7)
Ue κUe δ δ δ

Using Equation (7) and the Uτ already evaluated, ∏ values may be obtained for each
test condition within 0.1 ≤ y/δ ≤ 1. The results are summarized in Table 2 and plotted in
Figure 7. In that figure, the results are compared with two other measurements that are
Fluids 2023, 8, 135 12 of 19

associated with open-channel turbulent flows over smooth walls. It is noted that for both
test data taken from references [4] and [21], the data are in the range 830 < Reθ < 3000. Of
the two data sets, the results from reference [21] have the largest recorded Fr of up to unity.
The ∏ values reported herein are much lower than the 0.55 value quoted for zero pressure
gradient smooth plate flow. Additionally, there is no clear trend of dependency with Reθ or
Fr. One implication of these observations is that the free surface effects infiltrate deep into
all regions of the boundary layer [21], dampening the inner and outer layer disparities.
An examination of the wall-normal velocities and gradients of the wall-normal and
spanwise velocities in Figure 8 provides some insight into the nature of the dipping
phenomenon. The dip observed in the mean streamwise velocity distribution is often
attributed to large-scale secondary flow patterns from the corner toward the mid-span
of the channel. The observations of this work seem to point to a culmination of multi-
dimensional motion around the mid-depth region of the mid-span plane. Consequently,
a significant downward movement (V < 0) of flow toward the bed is recorded at that
location. The values increase with the Reynolds number. The wall-normal gradients in
the streamwise wall-normal velocity (∂V/∂y) and the spanwise gradients in the mean
spanwise flow (∂W/∂z, assessed from the continuity equation) are comparable. At the
highest Reynolds numbers tested in this work, the trend in the distribution of the velocity
gradients are unclear. However, at low Reynolds numbers, ∂V/∂y and ∂W/ ∂z are at least
five times that of the streamwise gradient of the mean streamwise velocity (∂U/∂x). The
significant values of ∂W/ ∂z indicate there may be a movement of flow directed from the
sidewalls toward the mid-span plane, and changes in flow directions. The dynamic changes
in ∂V/∂y and ∂W/∂z, being more prevalent at around y/h > 0.5, suggest that they may be
significant contributors to the dipping phenomenon observed below the surface of the flow.

3.3. Characteristics of Higher-Order Moments of the Turbulence Statistics

The higher-order moments of turbulence statistics are considered by focusing on
the turbulence intensities and Reynolds shear stresses. The Reynolds normal stresses
are omitted because they are qualitatively similar to the corresponding turbulent inten-
sities. The turbulence intensity and Reynolds shear stress data are normalized by the
friction velocities, plotted against the wall-normal coordinate per flow depth, and pre-
sented in Figures 9 and 10. In order to allow for the systematic study of the parametric
effects, the plots are shown to determine any trend due to comparatively large AR changes
(Figure 9a,c,e), Reθ changes (Figure 9b,d,f), Fr changes (Figure 10a,c,e), and combined AR
and Reθ modifications (Figure 10b,d,f).
Attention is first focused on the turbulence intensity profiles. These represent the level
of velocity fluctuations in the flow. It is important to point out that for the entire depth of
the flow, the normalized streamwise turbulence intensities (u+ ) are significantly larger than
the wall-normal components (v+ ). While the difference in the turbulence intensities peaks
at regions close to the wall, it reduces toward the mid-depth region. This is consistent with
the observations noted in an earlier section regarding enhanced wall-normal flow motion
around the mid-depth region. Another point to note about the turbulence intensity plots
is that while the directions of the trends of u+ and v+ are generally similar, they move in
opposite directions as the free surface is approached. The increasing values of u+ close
to the free surface indicate that the surface effects are much more dominant in the flow
direction, thus enhancing fluctuations in that direction, and attenuating those in the wall-
normal direction. Compared with data from previous studies of open-channel turbulent
flows, the turbulence intensities reported here are significantly different. This is expected
due to differences in the streamwise location of measurements. Thus, while the peak value
of u+ in Figure 9 is, for instance, ~10% greater than that reported by Tachie et al. [18], the
former is expected to decay at a location of fully developed flow.
, x FOR PEER REVIEW 13 of 20
Fluids 2023, 8, 135 13 of 19

1.0
Tachie et al.
Balachandar et al.
0.8 Current Work

0.6

0.4

∏ 0.2

0.0

−-0.2

−-0.4

0 500 1000 1500 2000 2500 3000

Req
Figure 7. (a) Dependence
Figure 7. (a) ofDependence
the locationofof thethe maximum
location streamwise
of the maximum mean velocity
streamwise ymax on y𝑅𝑒
mean velocity , Reθ ,
max𝜃 on
shown in dimensionless
shown in parameters
dimensionless(yparameters
max/h and a(ydip correction
max /h and a dip factor
correction ℎ⁄𝑦𝑚𝑎𝑥
𝛼 =factor α =− 1).
h/y (b)
max − Cole’s
1 ). (b) Cole’s
wake parameterwakefor current
parametertestfor
data compared
current with
test data other open-channel
compared flow studies
with other open-channel flow(Tachie et al. et al.
studies (Tachie
I [4] and Balachandar et Balachandar
I [4] and al. [21]). et al. [21]).
ids 2023, 8, x FOR PEER REVIEW 14 of
Fluids 2023, 8, 135 14 of 19

0.0 0.2 0.4 0.6 0.8 1.0

0.020
(a)
0.010

0.000

V/Ue −-0.010
−-0.020
AR1.1Re447Fr0.18 AR1.2Re527Fr0.18 AR1.4Re512Fr0.23
−-0.030 AR1.5Re165Fr0.13 AR1.5Re281Fr0.19 AR1.5Re383Fr0.27
AR1.3Re452Fr0.17 AR1.3Re729Fr0.42 AR1.3Re703Fr0.49
−-0.040 AR1.3Re836Fr0.58 AR1.3Re928Fr0.62 AR1.3Re864Fr0.63
AR1.3Re650Fr0.32 AR1.5Re829Fr0.77
−-0.050
0.08
(b)

0.04

(∂V/∂y) ⁄ (Ue /h)

0.00

(∂W/∂z) ⁄ (Ue /h)

−-0.04

−-0.08
∂ V / ∂y ∂ W / ∂z

0.4
(c)
0.2
(∂V/∂y) ⁄ (Ue /h)
0.0

(∂W/∂z) ⁄ (Ue /h)

−-0.2

−-0.4
∂ V / ∂y ∂ W / ∂z
−-0.6
0.0 0.2 0.4 0.6 0.8 1.0
y/h
Figure 8. (a) Dependence of the normalized mean wall-normal velocities (V/Ue ) on the wall-normal
Figure 8. (a) Dependence
location. Gradients of the ofstreamwise
the normalized mean
wall-normal wall-normal
and velocities
spanwise velocities (V/U
(∂V/∂y ande)∂W/
on ∂z)
thefor
wall-norm
location. Gradients of the streamwise wall-normal
(b) AR1.5 Re165 Fr0.13 and (c) AR1.3 Re836 Fr0.58 . and spanwise velocities (∂V/∂y and ∂W/∂z) fo
(b) AR1.5Re165Fr0.13 and (c) AR1.3Re836Fr0.58.

3.3. Characteristics of Higher-Order Moments of the Turbulence Statistics

The higher-order moments of turbulence statistics are considered by focusing on
turbulence intensities and Reynolds shear stresses. The Reynolds normal stresses
omitted because they are qualitatively similar to the corresponding turbulent intensiti
Fluids 2023, 8, x FOR PEER REVIEW 16 of 20

such a section is reported to yield a nearly constant correlation coefficient value of ~ 0.4–
Fluids 2023, 8, 135 15 of 19
0.5 [1]. This further demonstrates the significant differences in the variation of turbulence
intensities and Reynolds numbers, compared with other turbulent boundary layers.

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
3.0 3.0
AR1.1Re447Fr0.18 AR1.5Re165Fr0.13
(a) AR1.2Re527Fr0.18 (b) AR1.5Re281Fr0.19
AR1.4Re512Fr0.23 AR1.5Re383Fr0.27
2.0 2.0

u+
1.0 1.0

0.0 0.0
1.0 1.0
(c) (d)

0.8 0.8

v+ 0.6 0.6

0.4 0.4

0.2 0.2

0.8 0.8
(e) (f)
0.6 0.6
+
<-uv>
−
0.4 0.4

0.2 0.2

0.0 0.0

−-0.2 0.0 0.2 0.4 0.6 0.8 1.0 −-0.2 0.0 0.2 0.4 0.6 0.8 1.0
y/h y/h
Figure 9.
Figure 9. Variations of normalized
Variations of normalized streamwise
streamwise turbulence
turbulenceintensity
intensityuu++ (a,b),
(a,b), wall-normal turbulence
wall-normal turbulence
intensity v+ (c,d), and Reynolds shear stress <−uv +> (e,f), with the normalized wall-normal location
+ +
intensity v (c,d), and Reynolds shear stress <−uv > (e,f), with the normalized wall-normal location
y/h. This is conducted for selected test conditions to highlight effects of AR (a,c,e, using the same
y/h. This is conducted for selected test conditions to highlight effects of AR (a,c,e, using the same
legend) and Re𝜃 (b,d,f using the same legend).
legend) and Reθ (b,d,f using the same legend).

Again, with respect to the turbulence intensities, it is noteworthy that in the region
0.1 < y/h < 0.9, v+ values do not show any significant variations with AR and Reθ when
both parameters are low (i.e., AR < 1.5; Reθ < 400). However, a substantial increase in
Fr (for instance, from 0.32 to 0.64) leads to clear reductions in turbulent intensities at
0.2 < y/h < 0.4 and 0.9 < y/h < 1.0. Additionally, the combined effect of modifying Reθ and
Fr (in particular, increasing Fr by over 3 times at Reθ > 400) results in marked reductions
in streamwise turbulent intensities, outside of the experimental error or scatter. The
plots in Figure 10 show that those deviations are centered around the regions close to
the bed and the free surface. The pervasiveness of these effects raises doubts as to the
existence of a universal function for streamwise turbulence intensities in open-channel
flows. Specifically, the current data suggest that the universal functions for streamwise
and wall-normal turbulence intensities proposed by Nezu and Rodi [22] will not apply
Fluids 2023, 8, 135 16 of 19

Fluids 2023, 8, x FOR PEER REVIEW

to developing narrow open-channel low Reynolds number flows. This is a significant 17 of 20
observation, given that those functions were proposed by these researchers as the basis of a
simplified k-e turbulence model.

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
3.0 3.0
AR1.3Re650Fr0.32 AR1.3Re452Fr0.17
(a) (b)
AR1.5Re829Fr0.77 AR1.3Re729Fr0.42
AR1.3Re703Fr0.49
2.0 2.0 AR1.3Re836Fr0.58
AR1.3Re928Fr0.62
u+ AR1.3Re864Fr0.63

1.0 1.0

0.0 0.0
1.0 1.0
(c) (d)

0.8 0.8

v+ 0.6 0.6

0.4 0.4

0.2 0.2

0.8 (e) 0.8 (f)

0.6 0.6
+
<-uv
− >
0.4 0.4

0.2 0.2

0.0 0.0

−-0.2 0.0 0.2 0.4 0.6 0.8 1.0 −-0.2 0.0 0.2 0.4 0.6 0.8 1.0
y/h y/h
Figure 10. Variations
Figure 10. Variationsof ofnormalized
normalizedstreamwise
streamwiseturbulence
turbulenceintensity
intensityu+u(a,b), wall-normal
+ (a,b), wall-normal turbulence
turbu-
intensity +
lence intensity v and
v (c,d),+ (c,d),Reynolds
and Reynolds
shear shear
stress stress +
<−uv <−uv + > (e,f),
> (e,f), withwith the normalized
the normalized wall-normal
wall-normal location
location
y/h. Thisy/h. This is conducted
is conducted for selected
for selected test conditions
test conditions to highlight
to highlight effectseffects
of Fr of Fr (a,c,e
(a,c,e usingusing the
the same
same legend)
legend) and
and the the combination
combination of Reθ of
andReFr and Frusing
𝜃 (b,d,f (b,d,fthe
using
samethelegend).
same legend).

4. Summary and Conclusions

The Reynolds shear stresses denote the momentum fluxes of the unsteady turbulent
motions that work effectively
Particle image velocimetryas additional shear to
has been used stresses.
obtainUnlike two-dimensional
velocity measurements flows,
of an
the wall-normal distributions of these stresses (< − uv> + ) shown in Figures 9 and 10 are from
open-channel flow in the developing section of a flume with a narrow width. The aspect
aratio
three-dimensional flow.was
(AR) of the flow Consequently,
varied fromthey1.1 are characterized
to 1.5 by maximum
using a variant and
depth of minimum
flow h. The
Reynolds numbers based on the momentum thickness Re𝜃 and Froude number Fr were
also in the ranges of ~160 to 930 and 0.1 to 0.8, respectively.
This work demonstrates that the developing turbulent flow zone may be encoun-
tered at a much further upstream location (16 < x/h < 29) for a tripped flow. Yet, for such
Fluids 2023, 8, 135 17 of 19

points and changes in sign value. The latter characteristic is particularly consistent with
the velocity dipping phenomenon. It is important to note that the plots show that the
<−uv>+ profiles are affected by AR, Reθ , and Fr. Such effects are associated with various
sections of the flow depth. By increasing the AR from 1.1 to 1.4, the effective shear stresses
due to turbulent fluctuations increase by up to 450% over the range 0.2 < y/h < 0.9. Any
increment in the flow factors allied with a substantial change in Fr, on the other hand, tends
to decrease those stresses within a more restricted depth range (0 < y/h < 0.4). Perhaps
the most effective changes in Reynolds shear stress are obtained by varying the Reynolds
number. Increasing Reθ leads to an increment in <−uv>+ at 0 < y/h < 0.7, and a decrement
at y/h > 0.7. The location of the maximum value of <−uv>+ appears not to follow any
particular trend with respect to Reθ . However, the minimum value of <−uv>+ varies nearly
linearly from y/h = 0.65 to 0.99 as Reθ increases from 165 to 928. Overall, the complexity of
the profiles cautions a re-assessment of the simplified momentum relations of the Reynolds
equation used in the evaluation of the friction velocity.
To provide an additional comparison of the present flow with other canonical flows, the
relative effects of the turbulence intensities and the Reynolds shear stress may be assessed
through the correlation coefficient (i.e., the ratio of the Reynolds shear stress and the product
of u and v). The outcome is a range of values from −0.1 to 0.5 within 0.1 ≤ y/h ≤ 0.6. This
result is in contrast with other open-channel and boundary layer flows where such a section
is reported to yield a nearly constant correlation coefficient value of ∼0.4–0.5 [1]. This
further demonstrates the significant differences in the variation of turbulence intensities
and Reynolds numbers, compared with other turbulent boundary layers.

4. Summary and Conclusions

Particle image velocimetry has been used to obtain velocity measurements of an open-
channel flow in the developing section of a flume with a narrow width. The aspect ratio
(AR) of the flow was varied from 1.1 to 1.5 using a variant depth of flow h. The Reynolds
numbers based on the momentum thickness Reθ and Froude number Fr were also in the
ranges of ~160 to 930 and 0.1 to 0.8, respectively.
This work demonstrates that the developing turbulent flow zone may be encountered
at a much further upstream location (16 < x/h < 29) for a tripped flow. Yet, for such a
flow, only the streamwise components of the velocities are independent of the streamwise
location. For a developing open-channel turbulent flow over a smooth wall, the mean flow
consists of an inner and outer flow structure. For the inner layer, there exists a logarithmic
layer following the classical log law with a von Kármán constant ranging from 0.40 to 0.41,
and a log law constant of 5.0 to 6.0. The friction velocity obtained through the Clauser plot
technique may also be predicted through a linear equation associated with Fr. Utilizing the
current data and other data in the literature, a new logarithmic equation has been proposed
to predict the skin friction coefficient Cf with the knowledge of Reθ alone. For the outer flow,
it is shown that the wall-normal location of the maximum velocity ymax /h occurs below
the flow depth. Its associated dip correction factor α is a predictable linear function of
Reθ . The wake parameter is also found to range from −0.50 to 0.02, irrespective of the Reθ ,
Fr, or AR. With respect to the higher-order turbulence statistics, the turbulence intensities
indicate values that are consistent with an enhanced wall-normal flow motion around the
mid-depth region. The wall-normal distributions of the Reynolds shear stresses also reveal
distinctive minima, along with sign changes, due to dipping. There is a clear deviation of
the correlation coefficient (ranging from −0.1 to 0.5) from other open-channel flows.
An important conclusion derived from this work is that while some of the dynamics
of the unconstrained open-channel flow studied herein may be predicted using mean
flow assessment tools applied to conventional open-channel flows, there are important
differences. The implication of the foregoing on the evaluation of friction parameters and
turbulence modeling assessments is potentially significant and requires further probing.
Fluids 2023, 8, 135 18 of 19

Funding: This work was not funded by any external grant.

Data Availability Statement: Not applicable.
Acknowledgments: The support of this work through the C. Graydon and Mary E. Rogers Fellowship
is gratefully acknowledged.
Conflicts of Interest: The author declares no conflict of interest.

Nomenclature
English
B logarithmic law constant
b width of test channel
Cf skin friction coefficient =2 (Uτ /Ue )2
g acceleration due to gravity
h depth of flow
H depth of test channel
L length of test channel
Ly+ range of log layer in inner coordinates
Reθ Reynolds number based on maximum velocity and momentum thickness
Reh Reynolds numbers based on the maximum streamwise mean velocity and flow depth
Reτ Reynolds numbers based on the friction velocity and flow depth
U mean (time-averaged) streamwise velocity
u streamwise turbulence intensity
u+ streamwise turbulence intensity in inner coordinates = u/Uτ
u2 streamwise Reynolds normal stress
Ub depth-averaged mean streamwise velocity
Ue maximum mean streamwise velocity
Uτ friction velocity
Fluids 2023, 8, x FOR PEER REVIEW U+ time-averaged streamwise velocity in inner coordinates = U/Uτ 19 of 20
−uv Reynolds shear stress
<−uv>+ Reynolds shear stress in inner coordinates = −uv/Uτ 2
V mean (time-averaged) wall-normal velocity
vv wall-normal
wall-normal turbulence
turbulence intensity
intensity
vv++ streamwise turbulence intensity
streamwise turbulence intensity in inner in inner coordinates
coordinates = v/Uτ
= v/U τ
vv22 wall-normal Reynolds
wall-normal Reynolds normal stress normal stress
WW meanmean (time-averaged)
(time-averaged) spanwise
spanwise velocity
velocity
xx streamwise direction; streamwise
streamwise direction; streamwise distance distance
y wall-normal direction; wall-normal
wall-normal direction; wall-normal distance distance
y++ wall-normal direction in inner coordinates = y =Uyτ U
wall-normal direction in inner coordinates /ντ /ν

ymax
max location
location of of
the the
meanmean streamwise
streamwise velocity
velocity
z spanwise
spanwise direction;
direction; spanwise
spanwise distance
distance
Greek
𝛼α dipdip
correction
correction factor
factor ⁄𝑦 − 1− 1
= ℎmax
= h/y
δ boundary
boundary layer thickness
layer thickness
δ* displacement
displacement thickness
thickness
θθ momentum thickness
momentum thickness
𝜅κ vonvonKármán
Kármán constant
constant
νν kinematic viscosity
kinematic viscosity
∏∏ Cole’s wake
Cole’s wakeparameter
parameter
Other Symbols/Acronyms
ADV
ADV acoustic
acoustic Doppler
Doppler velocimetry
velocimetry
AR aspect
aspect ratio
ratio (b/h) (b/h)
Fr Froude
Froude number
number
ℌ boundary
boundary layerlayer shape
shape parameter,
parameter, = δ */θ
=δ */θ
LDA laser
laser Doppler
Doppler anemometry
anemometry
PIV particle image
particle image velocimetry
velocimetry
PVC polyvinyl
polyvinyl chloride
chloride

References
1. Nezu, I. Open-Channel Flow Turbulence and Its Research Prospect in the 21st Century. J. Hydraul. Eng. 2005, 131, 229–246.
2. Sarma, K.V.N.; Lakshminarayana, P.; Rao, N.S.L. Velocity Distribution in Smooth Rectangular Open Channels. J. Hydraul. Eng.
1983, 109, 270–289.
Fluids 2023, 8, 135 19 of 19

References
1. Nezu, I. Open-Channel Flow Turbulence and Its Research Prospect in the 21st Century. J. Hydraul. Eng. 2005, 131, 229–246.
[CrossRef]
2. Sarma, K.V.N.; Lakshminarayana, P.; Rao, N.S.L. Velocity Distribution in Smooth Rectangular Open Channels. J. Hydraul. Eng.
1983, 109, 270–289. [CrossRef]
3. Balachandar, R.; Ramachandran, S.S. Turbulent Boundary Layers in Low Reynolds Number Shallow Open Channel Flows.
J. Fluids Eng. 1999, 121, 684–689. [CrossRef]
4. Tachie, M.F.; Bergstrom, D.J.; Balachandar, R. Rough Wall Turbulent Boundary Layers in Shallow Open Channel Flow. J. Fluids
Eng. 2000, 122, 533–541. [CrossRef]
5. Bergstrom, D.J.; Tachie, M.F.; Balachandar, R. Application of power laws to low Reynolds number boundary layers on smooth
and rough surfaces. Phys. Fluids 2001, 13, 3277–3284. [CrossRef]
6. Tachie, M.F.; Balachandar, R.; Bergstrom, D.J. Low Reynolds number effects in open-channel turbulent boundary layers. Exp.
Fluids 2003, 34, 616–624. [CrossRef]
7. Kirkgöz, M.S.; Ardiçlioğlu, M. Velocity Profiles of Developing and Developed Open Channel Flow. J. Hydraul. Eng. 1997, 123,
1099–1105. [CrossRef]
8. Sarkar, S. Measurement of turbulent flow in a narrow open channel. J. Hydrol. Hydromech. 2016, 64, 273–280. [CrossRef]
9. Mahananda, M.; Hanmaiahgari, P.; Balachandar, R. Effect of aspect ratio on developing and developed narrow open channel flow
with rough bed. Can. J. Civ. Eng. 2018, 45, 780–794. [CrossRef]
10. Mahananda, M.; Hanmaiahgari, P.; Balachandar, R. On the turbulence characteristics in developed and developing rough narrow
open-channel flow. J. Hydro-Environ. Res. 2022, 40, 17–27. [CrossRef]
11. Arthur, J.K. Turbulent Flow through and over a Compact Three-Dimensional Model Porous Medium: An Experimental Study.
Fluids 2021, 6, 337. [CrossRef]
12. Arthur, J.K. A narrow-channeled backward-facing step flow with or without a pin–fin insert: Flow in the separated region. Exp.
Therm. Fluid Sci. 2023, 141, 110791. [CrossRef]
13. Wieneke, B. PIV uncertainty quantification from correlation statistics. Meas. Sci. Technol. 2015, 26, 074002. [CrossRef]
14. Tachie, M.F.; Bergstrom, D.J.; Balachandar, R.; Ramachandran, S. Skin Friction Correlation in Open Channel Boundary Layers.
J. Fluids Eng. 2001, 123, 953–956. [CrossRef]
15. Österlund, J.M. Experimental Studies of Zero Pressure-Gradient Turbulent Boundary Layer Flow. Ph.D. Thesis, Royal Institute of
Technology, Stockholm, Sweden, 1999.
16. Schlatter, P.; Li, Q.; Brethouwer, G.; Johansson, A.V.; Henningson, D.S. Structure of a turbulent boundary layer studied by DNS.
In Direct and Large-Eddy Simulation VIII; Springer: New York, NY, USA, 2011.
17. Schultz, M.P.; Swain, G.W. The Effect of Biofilms on Turbulent Boundary Layers. J. Fluids Eng. 1999, 121, 44–51. [CrossRef]
18. Tachie, M.F.; Bergstrom, D.J.; Balachandar, R. Roughness effects in low-Re θ open-channel turbulent boundary layers. Exp. Fluids
2003, 35, 338–346. [CrossRef]
19. Das, S.; Balachandar, R.; Barron, R. Generation and characterization of fully developed state in open channel flow. J. Fluid Mech.
2022, 934, A35. [CrossRef]
20. Krogstad, P.; Antonia, R.A.; Browne, L.W.B. Comparison between rough- and smooth-wall turbulent boundary layers. J. Fluid
Mech. 1992, 245, 599–617. [CrossRef]
21. Balachandar, R.; Blakely, D.; Tachie, M.F.; Putz, G. A Study on Turbulent Boundary Layers on a Smooth Flat Plate in an Open
Channel. J. Fluids Eng. 2001, 123, 394–400. [CrossRef]
22. Nezu, I.; Rodi, W. Open-channel Flow Measurements with a Laser Doppler Anemometer. J. Hydraul. Eng. 1986, 112, 335–355.
[CrossRef]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.