You are on page 1of 29

Environ Fluid Mech

DOI 10.1007/s10652-013-9312-5
ORIGINAL ARTICLE

Turbulent flow characteristics and drag over 2-D


forward-facing dune shaped structures with two
different stoss-side slopes
B. S. Mazumder K. Sarkar

Received: 27 February 2013 / Accepted: 3 September 2013


Springer Science+Business Media Dordrecht 2013

Abstract The present paper explores the characteristics of turbulent flow and drag over
two artificial 2-D forward-facing waveform structures with two different stoss side slopes of
50 and 90 , respectively. Both structures possessed a common slanted lee side slope of 6 .
Flume experiments were conducted at the Fluvial Mechanics Laboratory of Indian Statistical
Institute, Kolkata. The velocity data were analyzed to identify the spatial changes in turbulent
flow addressing the flow separation region with recirculating eddy, the Reynolds stresses, the
turbulent events associated with burst-sweep cycles and the drag over two upstream-facing
bedforms for Reynolds number Reh = 1.44 105 . The divergence at the stoss side slope
between the two structures revealed significant changes in the mean flow and turbulence.
Comparison showed that during the flood-tide condition there was no flow separation region
on the gentle lee side of the structure with smaller slope at the stoss side, while for the other
structure with vertical stoss side slope a thick flow separation region with recirculating eddy
was observed at the gentle lee side just downstream of the crest. The recirculating eddy
induced on the lee-side had a strong influence on the resistance that the structure exerts to
the flow due to loss of energy through turbulence. In contrast, a great amount of reduction
in drag was observed in the case of smaller stoss side sloped structure as there was no flow
separation. The quadrant analysis was also used to highlight the turbulent event evolution
along the bed form structures under flood-tide conditions.
Keywords Turbulence Forward-facing structure ADV Flow separation
Bursting events Drag reduction

B. S. Mazumder (B) K. Sarkar


Fluvial Mechanics Laboratory, Physics and Applied Mathematics Unit,
Indian Statistical Institute, Calcutta 700 108, India
e-mail: bijoy@isical.ac.in
K. Sarkar
e-mail: soundofphysiks@gmail.com

123

Environ Fluid Mech

1 Introduction
When the flow past an immovable object or obstacle like a step, shell, pebble or wood
fragment, a complex flow field around the obstacle is usually created by the interaction of
the local flow field with the obstacle. The object/obstacle resists the flow field in a stream
channel and develops a surge at the upstream side of the object accompanied by a series of
vortices at the downstream. Turbulent flow over steps and geometrically regular or irregular
rough surface structures plays a significant role in terms of separating and reattaching flow
phenomena to engineering fields. Separation of flow appears under a variety of flow conditions
such as, flow over the artificial dunes/ripples, the backward and forward facing steps (FFSs),
etc. Several investigations are available for various flow geometries. In particular, some of
the studies have been concentrated on the flow field over moving/rigid bed form structures
in alluvial channels or riverine environments [8,9,32,36,43]. These studies were concerned
with the occurrence of flow separation, reattachment points and turbulence characteristics
due to the ebb flow over such geometries and their impact of sediment transport over such
bed forms. These experiments, simulating the ebb flow over dune type bed forms, face gentle
stoss side slope and steeper lee side slope. Flows over two/three-dimensional backward facing
steps with sharp edges, rounded faces, etc. have received much attention in many engineering
applications such as, hydraulic and aerodynamic devices, combustors, and mixing equipment
[4,6,12,15,27,46]. Recently, Singh et al. [50] have studied the turbulent flow over a backward
facing step of different step angles to determine the separation points and reattachment
lengths.
Investigations had been carried out under controlled conditions in laboratory flumes to
understand the flow separation and reattachment phenomena over FFS immersed in a boundary layer flow. Stuer et al. [52] investigated the separation bubble at the upstream side of a
FFS under laminar flow condition using hydrogen bubble technique. They observed that there
was an open three-dimensional separation bubble characterized by span-wise quasi periodic
unsteadiness. The flow characteristics over a FFS and through the sudden contraction of a
pipe were examined by Ando and Shakouchi [3] visualizing the mean and fluctuating components using laser Doppler anemometer (LDA). Experiments were carried out to examine
the flow over the FFS with a smooth and rough upper surfaces for various Reynolds numbers [11,26,45,47,49]. Experiments showed the existence of two recirculation zones in the
FFS flow: one was just upstream of the sharp edge and the other was just downstream of
the sharp edge. The upstream separation was developed from the adverse pressure gradient
due to the blockage of the flow at the step face. The other separation of flow was developed
at the sharp edge and was characterized by the shedding of vortices which were convected
downstream [22]. The size of separation bubbles and the position of reattachment points
were dependent on Reynolds number and the roughness of the top surface of the structure.
Also the reattachment length X r was observed to be influenced by various flow parameters,
such as the turbulence level, the ratio of boundary layer thickness () to the step height (h  ),
i.e. / h  . Addad et al. [1] used a commercial Computational Fluid Dynamics technique for
large eddy simulation to study the flow over a forwardbackward facing step which showed
a good agreement with the previous experimental findings with LDA data. Sherry et al. [49]
studied the flow over the FFS for several values of the ratio / h  . They categorized the
study into two groups: for / h  > 1, the reattachment length X r was strongly dependent on
the / h  and for / h  < 1, the X r was weakly affected by the ratio, and usually situated
around 46h  . Subsequently, Saleel et al. [48], on investigating the flow over the identical
structure, verified that immersed boundary method was an active way to determine the flow
field around a forwardbackward facing step. Recently, Ren and Wu [47] made a comparative

123

Environ Fluid Mech

study on the turbulent boundary layers (TBLs) over smooth and rough FFSs using particle
image velocimetry. Examining the mean flows, the Reynolds stresses, the turbulent bursting
events and the average span-wise vorticity, they found that the separated flow at downstream
of the step was weakened due to roughness of the top surface of the step, though the flow
structure was independent of the surface roughness ahead of the step.
However, investigations into turbulent flow characteristics and bedform dynamics under
flood and tidal flow conditions have been quite rare until the relatively recently and it is an
area of study that is of interest to many researchers. Very recently field studies and numerical
models using Delft3D have been performed to estimate the variation of flow separation zone
over large bedforms during tidal environments [2830]. In fact, they took the field measurements on seabed bathymetry over a transect line crossing three ebb-oriented primary
bedform structures (see Figs. 2, 6 of [30]) during full tidal cycle and investigated numerically the flow separation zone due to the bed elevations in the natural environments. They
reported that during the ebb-tide a permanent separation zone occurred on the steep lee sides
(1423 ) of the ebb-oriented bedforms, whereas during the flood-tide no flow separation zone
developed over the gentle slope of flood lee side (35 ) except over the steepest (15 ) of the
lee side, where a small separation zone was observed [16,29,30]. Here the terms, stoss side
and lee side of the structures, are used conventionally according to the flow direction. They
also reported that the shape of the flow separation zone was not influenced by the changes in
flow velocity or water level rather it was influenced by the bed morphology. It may be noted
that usually ebb-oriented geometry of large-scale bedforms with asymmetric structures is
retained through both ebb and flood portions of the tidal cycle [16,58]. In the Fraser Estuary,
the heights of large dunes change due to tidal cycle, whereas their lengths remain stable
[23], which is comparable to the findings of Ernstsen et al. [16] for both the compound dune
and the superimposed dunes with the exception of the crest dune. Kostaschuk and Best [23]
found the dune height to increase with increasing flow velocity due to trough scouring caused
by increased turbulence. The quantitative knowledge on such geometrical structures, their
dynamics, sediment transport and their interaction with turbulent flow is very important for
coastal environments, river planning and restoration, and channel evolution process. In fact,
the flood-tide condition complicates the hydrodynamics of turbulence and sediment transport
phenomena around the bedforms due to variation in flow separation and flow resistances.
In spite of all these studies mentioned above, no experimental studies were performed
to simulate such environmental condition (reverse flow) to examine the mean flow, turbulence characteristics and drag over bedforms with steeper stoss-side slopes, despite the fact
that such study in the laboratory has the potential to be useful to the researchers who study
the bedforms in the natural systems, especially those that experience reverse flow conditions [14,34,56]. Therefore, a substantial investigation is required to understand the basic
hydrodynamic phenomena experimentally in a flume over the bedform structures oriented
against the flow. The present study is to identify the spatial changes in turbulence statistics of
flow addressing the recirculation eddies (flow separation regions), the Reynolds stresses, the
turbulent events associated with burst-sweep cycles and the drag coefficients over two artificial upstream-facing waveform structures. Two isolated waveforms with different stoss-side
slopes oriented against the flow (i.e., facing steeper stoss sides) are considered separately
under the identical flow conditions. More precisely, the investigation aims, experimentally in
a flume, at characterizing the turbulent flow over two artificial 2-D bedform structures separately: (1) asymmetric bedform with 50 stoss side angle, and (2) vertical (90 ) stoss side
structure, facing against the flow with a common gentle lee slope of 6 , which are akin to the
flow over the complex bedform roughness in tidal flows. Essentially, a threshold condition
of development of flow separation zone and circulation eddies on the bedform lee side in

123

Environ Fluid Mech

Fig. 1 Schematic diagram of the hydraulic channel

tidal environments needs to be determined, and hence two different stoss side angles (50
and 90 ) are considered for these tests. Hitherto it is unclear whether the occurrence of flow
separation region and circulation eddies on the bedform lee side due to reverse flow environment depends locally on the steepest part of the lee side, or it depends on the increase of stoss
side slope of the bedforms, or both. The velocity data are analyzed to highlight the turbulent
statistics and coherent structures in the flow over such bedforms with different stoss-side
angles, which are not studied earlier. Although the artificial structures with two different
stoss angles are not the correct representation in the natural flood-tide flow regimes, this
study will provide some understanding of the turbulence structures, flow separation regions
and drag on the flow over the bedforms without any added difficulty in measurement in the
tidal flow. The approximation of 2-D static artificial dune structures is justified because the
speed of the sand dunes is small compared with the mean flow [56]. Moreover, the use of
static bedforms allows a high spatial resolution of analysis and sampling very close to the
boundary which is not possible with mobile bedforms in the field.
The description of test channel, experimental method, procedure and flow conditions are
provided in Sect. 2; experimental results including mean flows, Reynolds stresses, quadrant analysis and drag reduction are presented in Sect. 3. Discussions and conclusions are
respectively provided in Sects. 4 and 5.

2 Experimentation
2.1 Test channel
Experiments were conducted in a re-circulating closed circuit laboratory flume [35,36] especially designed at the Fluvial Mechanics Laboratory (FML) of the Physics and Earth Sciences
Division, Indian Statistical Institute (ISI), Kolkata. Figure 1 shows the schematic diagram
of hydraulic channel. The experimental channel has the dimension of 10 m length 0.50 m
width 0.50 m height. The flume walls were made of Perspex windows over a distance of 8 m
providing a clear view to the flow. A centrifugal pump for flow discharge was located outside
the main body of the flume. The outlet pipe was fitted with a by-pass and a valve, so that
the flow discharge be adjusted to a desired maximum velocity. An electromagnetic discharge
meter with digital display was fitted with the outlet pipe to facilitate the continuous monitoring of flow. The inlet and outlet pipes were freely suspended from an overhead structure. The
upstream bend of the channel was divided into three sub-channels of equal dimensions, and
a honeycomb cage placed at each end of the sub-channels to ensure smooth and vortex-free

123

Environ Fluid Mech

uniform flow. For identical operating conditions, the water depth and discharge were kept
constant for all experiments.
2.2 Experimental method
In order to ensure the fully developed flow at the sampling station over the smooth rigid
surface, flume experiments were conducted at a flow discharge Q = 0.029 m3 /s at a constant flow depth h (= 0.30 m). The hydraulic slope of the flume was approximately negligible, which was of order 0.0003. Velocity data were collected at three different locations at
the flume central line along downstream using a SonTek 0.05 m down-looking 3-D Microacoustic Doppler velocimeter (ADV) for 5 min (300 s) at a sampling rate of 40 Hz. About 25
vertical sampling positions were used for each longitudinal locations. Data were recorded for
each location starting from the lowest height 0.004 m above the bottom to the highest position
0.22 m. Total 12,000 data points were collected from each vertical position. No velocity data
could be measured near the free surface in the present study. The sampling volume of ADV
was located 0.05 m below the transmitter probe and the precise distance depended on the individual probe geometry. The sampling volume was approximately cylindrical oriented along
8 m3 .
the transmitter beam axis. The size of the sampling volume of 16 MHz ADV was 910

The flow Reynolds number Reh (= Um h/) and the Froude number Fr (= Um / gh) are
respectively 1.44 105 and 0.28, where Um = 0.48 m/s is the maximum velocity observed
at the height z = 0.22 m above the flat surface, is the kinematic viscosity of water and g is
acceleration due to gravity.
Using the phase-space threshold de-spiking method described by Goring and Nikora [18],
the velocity data were analyzed for all three locations to remove noises; and found almost no
change amongst the results of those velocity data, which indicated the steady and uniform
flow at the measuring station. The cut-off level for good-quality velocity data was 93 %, i.e. at
least 93 % of the raw data was remained unaffected after using the de-spiking method. Such
excluded signals were replaced by the data using a cubic polynomial interpolation method.
In turbulent flow, the instantaneous velocity components (u, v, w) in the Cartesian coordinate system (x, y, z) are given by,
u = U + u  ; v = V + v ; w = W + w ,

(1)
u,

v ,

w

are the
where U, V, W are the time-averaged velocities in (x, y, z)-directions;
fluctuating components of u, v, w. Time-averaged stream-wise (U) and wall-normal (W)
velocity components are defined as,
U=
W =

n
1
ui ,
n

(2)

1
n

(3)

i=1
n


wi .

i=1

The root-mean-square (rms) velocity components Ur ms , Wr ms are defined as,



 n
1 
Ur ms = 
(u i U )2 ,
n
i=1

 n
1 
Wr ms = 
(wi W )2 ,
n

(4)

(5)

i=1

123

Environ Fluid Mech

Fig. 2 Normalized profiles of a stream-wise and wall-normal mean velocity (U/u and W/u ), b streamwise velocity (U/u ) in log-scale, c turbulence intensities (Iu and Iw ) and d turbulence kinetic energy ()

in three different stream-wise location over plane bed (three different symbols correspond to three different
locations)

where n is the total number of velocity observations. The normalized turbulence intensity
components are thus defined as,
Ur ms
,
u
Wr ms
Iw =
,
u
Iu =

(6)
(7)

where u is the shear velocity. The time averaged local Reynolds shear stress is determined
by uw = u  w  .
The profiles of normalized mean velocity components ( uU , uW ), log-law for U velocity,
turbulence intensities in stream-wise and vertical directions (Iu , Iw ) and turbulence kinetic
energy (TKE) over the smooth surface are computed and plotted together in Fig. 2ad. Here
the flow parameters are normalized by the friction velocity u . The stream-wise mean
 velocity

profile

U
u

over the smooth bottom surface exhibits a standard log-law

U
u

ln

z
zo

up to

flow depth about 0.20 m (Fig. 2b) with an equivalent bed roughness z o = 0.0001 m, where
u is friction velocity determined from the log-law and = 0.4 is the von Karman constant.
Here zero values of uW through out the depth indicates no secondary current in the flume over
the smooth surface. The following semi-analytical relationships for normalized turbulence
intensities (Iu,w ) and normalized TKE ()
are proposed as ,

z
,
(8)
Iu,w = Du,w exp Cu,w
h


= Dk exp 2Ck hz ,
(9)
where Du,w , Dk , Cu,w and Ck are the dimensionless coefficients determined from the fitted
equations of Iu , Iw , and ;
and given by 2.560, 0.754, 4.471, 1.03, 0.22 and 0.845, respectively. The results of mean velocity and turbulence intensities on the smooth bottom surface
are in good agreement with [4042].
Thereafter, two artificial waveform structures with different stoss-side slopes made of
smooth Perspex were fabricated for experiments. Both the structures were of a common
wavelength = 0.30 m with a common gentle slope of 6 at the lee-side (down-stream

123

Environ Fluid Mech

slope due to tidal flow), and the structures spanned the width of the flume. The first type of
structure, named as SFFS (slanting forward facing structure), had a stoss-side angle 50 (upstream slope due to tidal flow); and the second one, named as VFFS (vertical forward facing
structure), had a stoss-side angle 90 , which showed a mild change in the crest heights
h  = 0.03 m for the SFFS case, and h  = 0.0335 m for the VFFS case. These resulted
the steepness ratio, h  / = 0.1 for SFFS case and h  / = 0.112 for VFFS case, which
are consistent with the steepness values of real dunes in river flows [5,17,20] and in tidal
environments [56]. Each bedform structure was placed opposite to the flow direction (i.e.
in a tidal flow environment) at the same measuring location of the flume bed at about 6 m
downstream from the channel inlet. Experiments were conducted separately for each structure
under the identical flow condition to make a comparative study between the two structures
of different stoss angles (50 and 90 ) on the mean flows, turbulence characteristics, flow
separation and drag coefficients over the common gentle lee angle of 6 at the downstream
face. The bedform structures of two different stoss angles of 50 and 90 with a common
gentle lee angle of 6 were considered in the present study, which were closely analogous with
the bedform structures measured in the seabed bathymetry during the tidal environments (see
Fig. 6 of [30]). The observed stoss angles of all three primary bedforms were approximately
at the order of 77 and the lee sides were almost gentle slopped (see Fig. 6 of [30]). Therefore,
in the present study, the use of steep stoss angles (50 and 90 ) with a common gentle lee
slope was justified because the slope angles of the structures were of similar order which fall
into the range of those observed in natural environments. The present investigation will focus
here: how the flow structure will be affected if the artificial bedform structures are oriented
against the flow, which are akin to the tidal flow over the ebb-oriented bedforms mentioned
as SFFS and VFFS?
Twelve different measuring locations A, B, C, D,, K and L from upstream to downstream for both SFFS and VFFS cases had been selected (Fig. 3). The respective dimensionless
distances of the locations were x/ = 0.80, 0.40, 0.23, 0.13, 0.06, 0, 0.10, 0.40,

Fig. 3 Two forward facing dune shaped structures (VFFS and SFFS) along with the measuring stations A, B,
C,, L, M and N. Flow direction is from left to right. Each structure is having a common wavelength = 30
cm and width = 50 cm

123

Environ Fluid Mech


Table 1 Experimental values of
flow parameters

Plane bed
Reynolds number, Reh = Um h/
Maximum flow velocity, Um (m/s)

1.44 105
0.48

Mean flow depth, h (m)

0.30

Froude number, Fr

0.28

Friction velocity (from log law), u (m/s)

0.01911

Equivalent bed roughness, z o (m)

0.0001

SFFS
Crest height, h  (m)

0.03

Base length, (m)

0.30

Width (m)

0.50

Steepness, h  /

0.1

Lee side angle ( )

Stoss side angle ( )

50

VFFS
Crest height, h  (m)

0.0335

Base length, (m)

0.30

Width (m)

0.50

Steepness, h  /

0.112

Lee side angle ( )

Stoss side angle ( )

90

0.66, 1.00, 1.23 and 2.56. In addition, for the case of VFFS, two more locations M and N
were selected for measurements, and their dimensionless distances were x/ = 0.2 and 4.0,
respectively. The location F of the VFFS case corresponds to the location G of the SFFS,
which were the crest points of the corresponding structures. The Reynolds numbers based on
the crest heights Reh  (= Um h  /) for VFFS and SFFS cases are 1.6 104 and 1.44 104 ,
respectively. A summery of experimental conditions is provided in Table 1. The velocity
measurements were taken using ADV at all longitudinal locations (A, B,,L, M, N) separately under identical flow conditions. In a similar way, the ADV data were processed to
remove the spikes using a phase space threshold de-spiking method described by Goring and
Nikora [18].

3 Experimental results
3.1 Mean velocity components
The vertical profiles of normalized stream-wise mean velocity (U/u ) over both the structures
(VFFS and SFFS) from upstream to downstream at different selected locations (A, B,,L) are
plotted against z/ h in Fig. 4. It is observed from the figure that for the SFFS case, the normalized velocity (U/u ) profiles at all locations along the downstream show almost the identical
pattern like standard log-law [28] except the location G at the crest point, where a reverse
pattern is observed with a maximum normalized velocity (Um /u 30) at the crest level.
Moreover, there is an indication of decrease in mean stream-wise velocity near the bottom
just upstream of the crest position. For the VFFS case, the normalized velocity (U/u ) looks

123

Environ Fluid Mech

Fig. 4 U/u [x/] at the locations A, B, C,, to L for both SFFS and VFFS cases, where the symbols filled
circles and triangles indicate data points of SFFS and VFFS, respectively

like a log-law from locations A to E before the crest point; and then the velocity profiles seem
to be strongly affected by the presence of forward facing structure. Interestingly, different
flow patterns are observed for each case. From the figure, it is observed that stream-wise mean
velocities show reversed flow at E and on the lee side at G, H, and I. Significant flow separation
and the recirculating eddy within the separation zone occur just downstream of the crest F on
the gentle lee side; and then the mean velocities are seen to redevelop for the VFFS case, while
for the case of SFFS, the velocity patterns possess a small indication of flow reduction beyond
the crest position G (no flow separation; [16]). The flow separation region with recirculating
eddy induced on the lee-side has a strong influence on the resistance that bedform exerts on
the flow due to energy loss through turbulence [54]. Observations from the VFFS case reveal
that a sharp growth of mean velocity initiates from the crest position F up to the position K that
reaches to a maximum velocity at certain level z/ h near bottom and then decreases slowly.
The peak of maximum velocity grows towards the downstream along x/. An indication of
strong shear layer within the recirculation region, more akin to the TBL, is observed. The shear
layer reflects the changes in the velocity gradient between the two overlaying flow regions:
flow reversal region near the boundary and high velocity away from the boundary. The mean
velocity profiles gradually become fully developed at further downstream of the recirculation
region and collapse in the outer flow region, which indicates that the mean flow redevelops
from the perturbation and follows the log-law at a distance of 3.0(x/) from the toe point J.

123

Environ Fluid Mech

Fig. 5 Shear velocity, u [x/] derived from log-law at each of the measuring location for SFFS (with filled
circles symbols) and VFFS (with triangles symbols) case. Black dashed vertical line denotes the crest position
G for SFFS and grey dashed vertical line denotes the crest position F for VFFS

A comparative study between the two cases indicates that at the location A the streamwise velocity (U/u ) profiles show a log-law and overlap each other. As it proceeds towards
downstream, at each location there is an intersecting point of two different velocity profiles.
Below the points of intersections, mean velocity profiles collapse each other up to the location
D at x/ = 0.13. Further downstream, the mean velocity for the VFFS case seems to
be separated and dramatically reduces with flow reversal at the locations G and H and a
reattachment point at I (x/ = 0.66). Above the intersecting points, the mean velocities
for the VFFS case seem to increase and differences are becoming much smaller further
downstream at L (x/ = 2.56). The distance of the reattachment point from the crest point
F is approximately 6h  , which is at the location I (x/ = 0.66). It is interesting to note that
the mean velocity profile at the crest location F of the VFFS case follows a log-law with a
maximum velocity (Um,V F F S /u = 29) at the level z/ h = 0.2, and it is concave in shape
at the crest position G for the SFFS case with Um,S F F S /u = 30 at the level z/ h = 0.1. It is
worthwhile to note that the mean stream-wise velocity is higher for the case of VFFS over
the entire measuring zone than that for the SFFS case.
For the SFFS case, the velocity profiles are tested for log-law with the universal von Karman constant (= 0.40) at all locations except the location G at the crest point (computed
log-law profiles are not shown in the figure), which is similar to results obtained by Lefebvre
et al. [28] for flood tide over the ebb oriented bedforms. The friction velocity u
computed from respective log-law ranges from 0.022 to 0.035 m/s along the flow with
dynamic roughness height ( 0.00043 m). The computed friction velocity u is plotted against x/ except the location G in Fig. 5 (filled circle symbols), which shows
the increase of u from A to F up to the head point, and then shows the decrease
of u from H to L over the lee-side of the structure. It is observed that the value
of u at the toe point J is lower than the other values, where the log-law fitted better with regression coefficient R 2 = 0.99 than the other locations. The upward slope
of u before the head point and downward slope over the lee-side represent respectively the deceleration and acceleration of flow near the boundary. On the other hand,
for the VFFS case, the stream-wise velocities follow log-law from A to E before the
crest point, and the corresponding values of friction velocity u are shown in Fig. 5
(triangle symbols). The mean velocity is strongly perturbed by the structure and hence no
log-law is observed until the long distance. Friction velocities for VFFS case show greater
values than that of SFFS case along the flow up to the crest location. The extreme upstream
and extreme downstream locations show approximately same value of u for both cases,
implying that there is no influence of the structures.

123

Environ Fluid Mech

Fig. 6 W/u [x/] at the locations A, B, C,to L for both SFFS and VFFS cases, where the symbols filled
circles and triangles indicate data points of SFFS and VFFS, respectively

The profiles of normalized vertical mean velocity (W/u ) for both the cases are shown
against z/ h in Fig. 6 for different stream-wise locations. It is observed that the (W/u ) at the
location A for both VFFS and SFFS cases is almost zero, which means that there is no effect
of waveforms. Mean velocity profiles for both the structures (SFFS and VFFS) increase with
vertical height near the boundary approximately up to z/ h 0.2 from the location B to the
location G, and then decrease to zero away from the boundary. It is interesting to note that
the vertical velocity profile at the crest position F for the VFFS case is qualitatively similar
to that for the SFFS case at the position G with about 3.5 times higher in magnitude. Just
after the crest on the lee side, the negative vertical velocity (directed towards the bed) occurs
with a maximum value near the crest level at about z/ h = 0.12 for the VFFS case, which
is associated to the reattachment of the flow to the bed. In addition, the zero-normal mean
velocity for the SFFS case occurs in between K and L, whereas for the VFFS case it is at
point L.
To illustrate the visualizations, streamline plots of velocity vector (U, W) for the both
SFFS and VFFS cases are shown in Fig. 7a, b. Figure 7c is the magnified version of the
separation bubble of the VFFS case. This showed a clear recirculation bubble for the case
of VFFS located just downstream of the crest on the upper lee side, which was comparable
with the observations of [11,47,49], who executed the work with FFS with flat surface. The
extrapolation technique was used for plotting of streamlines. A separation point was obtained

123

Environ Fluid Mech

Fig. 7 Streamline plot of velocity vectors for a SFFS case, b VFFS case, and c magnified version of the VFFS
case

in the x-direction at the upstream of the structure where the stream-wise mean velocity was
nearly equal to zero; and a reattachment point at the downstream of the recirculation bubble
on the lower lee side was attained, where stream-wise mean velocity became nearly equal to
zero. From the results of the Figs. 4, 6 and 7c, it was evident that the flow separated at nearly
x/ = 0.05 upstream of the case of VFFS case. However, for the case of SFFS, no flow
separation region was observed. According to the observed data, no such reattachment point
was observed at the vertical face of the VFFS case, i.e. where the vertical velocity component
became nearly equal to zero. Figure 7c shows that the length of the recirculation bubble at
the downstream of the structure, X r is approximately equal to 0.19 m with a bubble center at
x/ = 0.26, z/ h = 0.112 (exactly at the crest height). The disappearance of recirculation
bubble at downstream of the crest for the SFFS case was probably associated with the bed
topography of stoss side, which generated local favorable pressure gradient to prevent any
flow separation. This finding was comparable with Ernstsen et al. [16] and Lefebvre et al.
[30].
3.2 Turbulence intensities
The plots of normalized stream-wise turbulence intensity (Iu ) are shown along downstream
in Fig. 8 for both SFFS and VFFS cases. It is observed that the pattern of turbulence intensity
(Iu ) is almost identical through out the depth z/ h except near the boundary from the locations
A to E for both cases. In fact, as it proceeds towards downstream leading from the location

123

Environ Fluid Mech

Fig. 8 Plots of normalized stream-wise intensity for: a SFFS case and b VFFS case. Flow direction is from
left to right

A, there are successive increments of intensity near the boundary up to location E, then from
the crest F of VFFS a drastic growth of intensity with a maximum value Iu = 9.95 at F is
observed. It may be pointed out that beyond the crest F of VFFS within the recirculation
region on the lee side, Iu increases sharply and reaches a maximum value at each location
at a level of z/ h 0.15 and finally decreases and coincides with the intensity profiles of
SFFS case. The position of the maximum intensity for the case of VFFS moves toward the
free surface further downstream, which is opposite to the result achieved by Mazumder et
al. [36] who conducted experiments with isolated scalene waveform structure; and similar to
the observation of [13]. The magnitude of stream-wise intensity profiles for the case of SFFS
is much smaller than that of the VFFS case near the boundary leading from the location E
up to the location L. The intensity Iu attains its maximum value over the crests for both the
cases. It is noted that the zones of high turbulence intensity are characterized by low mean
velocity. The presence of a peak in the vicinity of crest level indicates the presence of a shear
layer, which is caused by a separation of the flow.
The normalized vertical turbulence intensity (Iw ) is shown along the flow in Fig. 9 for
both SFFS and VFFS cases. The profiles of Iw along the stream-wise locations show the
similar trend as Iu with smaller in magnitude. The intensity Iw is seen to be more prominent
at the downstream of the crest for the VFFS case and this effect persists for a longer distance
than that in the case of Iu for both the cases. For the VFFS case, a slight increase in Iw is

123

Environ Fluid Mech

Fig. 9 Plots of normalized wall-normal intensity for: a SFFS case and b VFFS case. Flow direction is from
left to right

observed while approaching the forward face, which is about 5 % of the Um . This change in
Iw is due to the vertical forward face and it is more prominent compared to Iu . It is observed
that Iw for the case of VFFS is higher than that of the SFFS along the stream-wise locations,
and after the crest G of VFFS case within the recirculation region, Iw increases sharply and
reaches a maximum value at each location at a level of z/ h 0.2 and finally decreases
gradually towards the main flow.
3.3 Reynolds shear stress
The dimensionless Reynolds shear stress (uwdim = u  w  /u 2 ) is shown against z/h for all
stream-wise locations in Fig. 10 for the both SFFS and VFFS cases for comparison. It is noted
from the figures that the magnitude of turbulence shear stress (uwdim ) is almost identical
through out the depth z/ h from the locations A to E for the both VFFS and SFFS cases, while
at the crest F of VFFS case shear stress uwdim initiates to increase due to change in velocity.
At the location G of the immediate downstream of the crest F of VFFS case and at the crest
G of the SFFS, a small region of negative Reynolds shear stress is observed, indicating the
outward flux of momentum. Further downstream, especially in the case of VFFS, a large
region of high positive Reynolds shear stress is seen within the region z/ h 0.5 with a
clear peak near the boundary. Interesting to note that the maximum shear stress falls at a

123

Environ Fluid Mech

Fig. 10 uwdim [x/] at the locations A, B, C,to L for both SFFS and VFFS cases, where the symbols filled
circles and triangles indicate data points of SFFS and VFFS cases, respectively

distance within 0.50 x/ 0.66, where the reattachment point exists. Furthermore, the
position of the maximum uwdim moves toward the free surface and decreases gradually as
it proceeds towards the downstream, which has the evident in the open channel flow over
dunes [13]. Reynolds shear stress profiles collapse in the upper part of the flow within the
range 0.3 z/ h 0.6. Close to the bottom, the Reynolds shear stress profiles are greatly
perturbed and the development of a shear layer is confirmed by the presence of peaks.
3.4 Quadrant analysis of Reynolds shear stress
The observations reveal that the TBL is directly associated with large-scale coherent structures, occurring irregularly. Coherent structures with large flux events have been proposed to
explain the bursting phenomena responsible for resistance to motion, transport processes,
turbulence production, and hence mixing. These coherent structures are quasi-periodic and
occupy the total boundary layer depth. The turbulence over the bedform structures against the
flow is examined through quadrant analysis to estimate the major turbulent events characterizing the coherent structures. The analysis clearly indicates the impact of the chosen structures
on the dominant Reynolds shear stress. The quadrant threshold technique for direct estimation of observed data of conditional statistics of Reynolds shear stress is presented briefly.

123

Environ Fluid Mech

The longitudinal and vertical velocity components U and W are aligned along x and z directions, respectively and Reynolds shear stress = u  w  is the inward flux of stream-wise
momentum.
The quadrant analysis is originally devised to sort out the contributions to from each
quadrant of instantaneous values on the u  w  plane. The quadrants are usually referred
by the following names: (a) outward interactions (i = 1; u  > 0, w  > 0), Quadrant-1
(Q1), (b) ejections (i = 2; u  < 0, w  > 0), Quadrant-2 (Q2), (c) inward interactions
(i = 3; u  < 0, w  < 0), Quadrant-3 (Q3) and (d) sweeps (i = 4; u  > 0, w  < 0),
Quadrant-4 (Q4). This method of quadrant decomposition is similar to that used in some
of the previous researches like [31,36,39,47,57]. The existing quadrant analysis is used to
analyze how the ebb oriented bedform structures under flood-tide conditions influence the
turbulent events. The analysis essentially highlighted the turbulent event evolution along the
bedform structures. Each velocity pair from ADV data may be investigated either through
examining the entire signal data or filtering those data above a threshold value (i.e., excluding
a Hyperbolic region of size H) defined as,
H = |u  w  |/ [(Ur ms ) (Wr ms )] .

(10)

At any point in a turbulent flow, the contribution of the total Reynolds stress from quadrant
i, excluding the region H, is defined as,


u w i,H

1
= lim
n n

u  (t)w  (t)Ii,H [u  , w  ]dt,

(11)

where n is the total number of measurements and the angle brackets denote a conditional
average and the indicator function Ii,H is defined as,

1, if (u  , w  ) is in ith quadrant and




if |u  w  | H [(Ur ms ) (Wr ms )] ,
(12)
Ii,H (u , w ) =

0, otherwise.
Here, H is the threshold parameter in the Reynolds stress signals by which one can
extract the values of u  w  from the whole set of signal data, which are greater than H times
[(Ur ms ) (Wr ms )] value. In the Eq. 12, the expression |u  w  | H [(Ur ms )(Wr ms )] is used as
a filter which cancels out all those data whose strength is less than H times [(Ur ms ) (Wr ms )].
The stress fraction [31,36] by ith quadrant is defined as,
Si,H =

u  w  i,H
,
[(Ur ms )(Wr ms )]

(13)

which gives the Reynolds shear stress fraction associated with each of the turbulent events. By
definition, S1,H , S3,H < 0 and S2,H , S4,H > 0, and for H = 0, S1,0 + S2,0 + S3,0 + S4,0 = 1.
The space fraction [47,57] gives the fraction of space , i.e., fraction of total observations
(n) contributing to the instantaneous Reynolds shear stress by each of the quadrant events for
a given H, is attained by using

Ii,H
Ni,H (u  , w  ) =
,
(14)
n
where Ii,H is the indicator function (Eq. 12) and n is the total number of observations.
For comparison, contributions of stress fraction Si,H were computed for the both SFFS
and VFFS for H = 0 and 4 for all the four quadrant events.

123

Environ Fluid Mech

Fig. 11 Contour plots of stress fraction |Si,H | for each quadrant for H = 0: a SFFS case and b VFFS case.
Flow direction is from left to right

Figure 11a, b shows the contribution of all four quadrant events to the Reynolds shear
stress i.e., stress fraction |Si,H | for i = 14 for the threshold parameter H = 0 over the
both SFFS and VFFS cases. These figures show a trend that the contributions of ejections
and sweeps to the shear stress are much higher than that of the outward and inward interactions, though there are some conflicts regarding the statement at some locations which
might be the point of interests. In the vicinity of the crest position of the SFFS case and at
a distance just downstream of the crest for the VFFS case, it is interesting to note that the
contributions of ejections and sweeps are negligible, whereas outward and inward interactions dominate equally at the referred locations for both of the cases. Contour plots show
that contributions from ejections and sweeps to the shear stress are almost same for the
case of SFFS at the downstream of the crest, where no recirculation region is formed. On
the other hand, for the case of VFFS, sweeps have larger effects to the shear stress in the

123

Environ Fluid Mech

Fig. 12 Contour plots of stress fraction |Si,H | for each quadrant for H = 4: a SFFS case and b VFFS case.
Flow direction is from left to right

recirculation region. Further downstream near the bed, sweeps overshadow the ejections
for the VFFS case. In addition, larger contributions of ejections to the Reynolds shear
stress are distributed at a slightly away from the bedform structure which is similar to
the observation of Ren and Wu [47]. At the interface of ejections and sweeps over the
trough region, kolkboils phenomenon is observed [40,41]. A kolk phenomenon is associated with an ejection when rapidly rushing fluid passes the forward face (stoss side) of
the structures [33,43]. Existence of upward tilting stream-wise vortex motion in the flow
is known as kolk; while boils are the structures like circular or oval shaped patch lift
up from the kolk on the water surface, which dissipate or merge with the surroundings
[40,41].
Figure 12a, b shows contributions to Reynolds shear stress from all the four quadrants for
H = 4 for the both SFFS and VFFS cases. It is obvious that other than the most strong events,
all will filter out. Only inward interactions contribute notably in the vicinity of crest region of

123

Environ Fluid Mech

Fig. 13 Contour plots of the ratio |S2,0 |/|S4,0 | : a SFFS case and b VFFS case. Flow direction is from left to
right

SFFS case. Over the crest for VFFS, intense ejections largely contribute to the shear stress.
As we go further downstream, on the surface of the structure in the recirculation bubble,
sweeps contribute most as in the case of H = 0. Away from the surface at the downstream
of the crest of the VFFS case, intense ejections largely dominate over sweeps similar to the
observation of [40,41]. In the case of SFFS, contributions from ejections and sweeps vanish
as compared to the case of H = 0.
Figure 13a, b depicts the ratio of ejections to sweeps (|S2,0 |/|S4,0 |) that contributes to
the Reynolds shear stress. This figure verifies the previous statements about the positions
where the sweeps have greater contributions to the Reynolds shear stress and where the
ejections overshadow it. If the ratio |S2,0 |/|S4,0 | < 1 at any position, the sweeps are more
prominent than the ejections at that position; and if |S2,0 |/|S4,0 | > 1 at any point, the ejections
overshadow the contributions of the sweeps to the shear stress at that point. Just in the vicinity
of the crest and in the recirculation region sweeps exceed ejections by approximately a factor
of 1.4 for the VFFS case. Typically contributions of ejections exceed that of sweeps in further
vertical distance and downstream by a factor ranging between 1.13 to 2.5. However, for the
SFFS case in the vicinity of the crest position, contributions of sweeps exceed that of ejections
by a factor of 2.4 approximately.
The space fractions Ni,H (u  , w  ) occupied by all the four quadrants for H = 0 are shown
in Fig. 14a, b for both SFFS and VFFS. Clearly, Q1 and Q3 events (interactions) have low
probability to occur in near bed region and the events generally increase above the bed
z/ h > 0.25. In contrast, the percentage of space occupied by ejections Q2 and sweeps Q4
is high near the bed and decreases slowly toward the water surface. In the vicinity of the
crest for the SFFS case, outward and inward interactions occupy around 30 % of the space,
while for the case of VFFS at the same location, sweeps overshadow the other three. Near
the surface of the structure in the recirculation region for the VFFS case, ejections are seen
to occupy higher space about 40 % than sweeps which is about 25 %. Sweeps occupy the
larger space just at the upper regions of recirculation bubble and go on increasing further
downstream.

123

Environ Fluid Mech

Fig. 14 Contour plots of space fraction Ni,H to each quadrant for H = 0: a SFFS case and b VFFS case. Flow
direction is from left to right

3.5 Flow resistance and drag reduction


There were several approaches to deduce the flow resistance, i.e. total boundary shear stress
and hence the overall drag. Flow resistance due to the form drag exerted by bedforms is
important to quantify its influence on water surface fluctuation. In general, vertical profiles
of stream-wise velocity measured over the waveform structures were used to calculate the
boundary shear stresses. The profiles may be either locally derived or by means of spatial
averaging. Smith and McLean [51], McLean et al. [37] preferred the latter procedure. There
were several ways to construct a spatial average: (1) averaging over constant elevations
above the waveform structure, (2) averaging along streamlines of constant velocities, or
(3) averaging the velocity data at constant values of z above the surface along the streamwise direction. The last procedure was applied for our present study.

123

Environ Fluid Mech

Fig. 15 Spatially averaged mean stream-wise velocity profiles (U ) for a SFFS case and b VFFS case, where
the symbols filled circles and triangles indicate data points of SFFS and VFFS cases, respectively

Spatial averages of mean stream-wise velocity along the flume centerline, U were computed for both the SFFS and VFFS cases using the third procedure of spatial averaging as
stated earlier and plotted in Fig. 15a, b. As recommended in [37], velocity average excluded
those data in the separation zone. Spatially averaged profile near the bed was characterized
by shear velocity associated with the skin friction. The spatially averaged velocity profiles
for both the cases follow log-law with coefficient of regression, R 2 0.98 and are given by,
U = 7.713 log(z/ h) + 54.46 (for VFFS),

(15)

U = 5.493 log(z/ h) + 52.98 (for SFFS).

(16)

and

The corresponding shear velocities, u T associated with the fitted log-law and hence the
total boundary shear stresses, T (= u 2T ) are given by 0.0309 m/s and 0.9518 N/m2 for the
VFFS case and 0.0220 m/s and 0.4828 N/m2 for SFFS case, respectively. It is notable that
mean flow resistance over the VFFS case is much larger than that of the SFFS case.
Again, it can be argued that, rather than comparing T , drag coefficient can be compared
according to [55], but as the flow discharge was constant for all the experiments, i.e. steady
flow condition was maintained, the flow resistance can be thought of only dependant on the
waveform shape. Drag coefficient C D is defined here as, C D = (u /U )2 , where u is the
shear velocity calculated by extrapolating the Reynolds stress profiles at each stream-wise
locations and U is the depth averaged mean stream-wise velocity. Stream-wise values of drag
coefficients for both SFFS and VFFS cases are plotted in Fig. 16. Therefore, it is seen that in
the flow reversal region on the waveform there is an increase in drag and it is much higher for
the case of VFFS than that of SFFS. So, by reducing the stoss-side angle, we can minimize the
overall flow resistance. Thus the reduction in drag occurs. It can be concluded that waveform
shape with different stoss-side slopes against the flow (forward facing to the flow) plays an
important role for controlling the mean flow resistance and hence the movement of sediments
in a fluvial channel.

123

Environ Fluid Mech

Fig. 16 Drag coefficient over the two structures SFFS and VFFS along the stream-wise direction, where the
symbols filled circles and triangles indicate data points of SFFS and VFFS, respectively. Flow direction is
from left to right

4 Discussion
The flow structure generated over forward-facing dunes/obstructions has many important
implications to the bedform dynamics, bed shear stress and flow resistance due to form drag
in the tidal flow situation. The above considerations motivated the need to investigate the
role of forward-facing dune-shaped structures to the flow field. Study of flow associated
with these structures in the tidal environment is inherently difficult since it is usually hard to
measure the near-bed flow and hard to quantify the nature of turbulence over these structures.
The current paper presents the experimental data for the mean flow, the turbulence and the
resistance due to drag associated with these bedform structures, which could be thought of
as a bluff body or object in unidirectional flow or a relic bedform subjected to a tidal bore
oriented in the opposite direction (i.e. in reverse flow condition). The novelty of this study is
the execution of the flume experiments to characterize the turbulent flow over two forwardfacing bed form structures, which are akin to the tidal flow environments, and to make the
necessary comparative discussions with the field and numerical investigations of dynamics
of dunes in tidal environments [16,29,30]. Moreover, the test data are used to analyze the
turbulent statistics of flow over such bedforms with different stoss-side angles, which were
not studied earlier. Thus the results were evaluated in terms of turbulence characteristics
and the coherent structures in the perturbed flow region. This study showed that there were
significant changes in the turbulence, especially in the near-bed area and in the vicinity of
the crest position due to different geometries, i.e. difference in stoss-side angle.
Shear layers generated along flow in the SFFS case had a much smaller velocity than the
velocity from the characteristics of shear layer generated by the flow separation in the case
of VFFS. Turbulence production associated with the VFFS case was dominated by eddies
generated along the shear layer, which produced high horizontal and vertical velocities and
large Reynolds stresses in this region.
For the SFFS case, the velocity profiles followed log-law [28] with the universal von
Karman constant (= 0.40) at all locations except for the location G at the crest point. On
the other hand, for the VFFS case, the log-law was observed from A to E before the crest
point; and then log-velocity was strongly perturbed by the structure; and hence no log-law
was observed until far away from the crest. Interesting to note that the mean velocity profile
at the crest F of the VFFS case followed a log-law, while the shape of the velocity profile

123

Environ Fluid Mech

with vertical hight was concave at the crest G for SFFS case. The vertical velocity profile at
the crest F for the VFFS case was qualitatively similar to that for the SFFS case at the crest
position G with about 3.5 times higher in magnitude.
The bedform structure with forward-facing vertical stoss side (VFFS case) acquired a thick
flow separation region with a prominent recirculating eddy on the gentle lee side, whereas
there was no indication of flow separation or flow reduction on the lee side of the forwardfacing bed form structure (SFFS case). The flow separation induced wakes that grew and
transported downstream occupying the outer flow towards the free surface. The length of the
recirculation bubble at the downstream of the structure was about 0.19 m with a bubble center
exactly at the crest height for the VFFS case (Fig. 7c). The disappearance of recirculation
bubble at downstream of the crest for the SFFS case was probably due to the topography
of slanted stoss face, which prevented flow separation (Fig. 7a). Similarly, in the numerical
study by Lefebvre et al. [29], for bedform nos. 2 and 3, there was no flow separation zone
during the reverse flow; whereas, permanent flow separation zones were achieved during the
ebb flow condition. The bedform structures considered in the present study are analogous to
the natural dune structures measured in the seabed bathymetry during the tidal environments.
It is interesting to determine the threshold condition of the development of flow separation
zone and recirculation bubbles in lee side of the bedform structures; and the dependence of
stoss and lee angles to turbulence parameters, drag coefficients during the tidal conditions.
Therefore, the implication of this study is to understand the turbulence phenomena, what
occur in the natural dunes. Mazumder et al. [36] reported experimentally that during the ebbtide flow a permanent separation zone occurred on a sharply sloping lee side (50 ) for the
ebb-oriented bedforms, whereas in the present study during the flood-tide no flow separation
zone developed over the gentle slope of flood lee side (6 ), that agrees well with the ebb
flow condition in the work of Kostaschuk and Villard [25], Best and Kostaschuk [10] and
Paarlberg et al. [44], which confirmed the occurrence of permanent flow separation on the lee
side slope of greater than 10 . Importantly, for ebb flow condition, flow separates at the crest
point [29,36] but in the present study, there was a flow separation for only VFFS case which
occurred at about x/ = 0.05 upstream of the structure. For VFFS case, the flow patterns
were classified as the boundary layer development region before the crest point; the flow
recirculation and the reattachment on the lee-side, and the boundary layer redevelopment
region along the flow. No definite recirculation bubble was observed at the upstream side
of the structure for either cases, which is contrary to the previous result using a FFS and a
forwardbackward facing step [11,47].
A high turbulence intensity was characterized by low flow in the flow reversal region,
while the outer flow region was characterized by high velocity and lower turbulence intensity.
Vertical normalized intensity seemed to be more prominent at the downstream of the crest
than the stream-wise intensity and the effect persisted for a longer distance. The intensity
Iu attained its maximum value over the crest for both the cases. In contrast, the vertical
intensity Iw attained its maximum over the lee slope at about x/ = 0.66 distance from the
crest for the VFFS case and interestingly for the SFFS case the maximum of it was observed
in the vicinity of the crest position. Whereas, for the ebb flow condition over such SFFS,
Mazumder et al. [36] concluded that both the Iu and Iw attained their maximum at about
x/ = 0.70 distance from the crest. Again, comparison with Mazumder et al. [36] showed
the disappearance of definite peak in Iu . As in the case of reverse flow, the maximum Iu
was always obtained at lower-most point at 0.004 m above the bed. The maximum value of
stream-wise intensity Iu at the crest F of the VFFS case was about two times greater than that
of Iw at I, which was about 4.5 for the VFFS. For ebb flow condition over the same structure
as our SFFS, according to the study of [36] maximum Iu is about three times greater than Iw ,

123

Environ Fluid Mech

whereas for our case in reverse flow condition, Iu was measured to be five times greater than
Iw . For the SFFS case, the maximum of Iu is nearly equal to 25 % of the Um (= Um /u ).
Again, at downstream of the crest region, Iu greatly reduces to 12 % compared to that of
the VFFS case, and that value is lower than the value obtained by Ren and Wu [47]. The
maximum of Iu for VFFS case attained about 39 % of the normalized maximum velocity
Um (= Um /u ) which agreed reasonably well with the earlier studies by Agelinchaab and
Tachie [2] ( 3040 %), Ren and Wu [47] ( 30 %) and Sherry et al. [49] ( 3745 %). Just
downstream of the crest F, a high increment in Iw is observed in the layer near the boundary,
where the negative vertical velocity occurred. In this region, for the VFFS case, Iw is found
to have magnitude of about 18 % of Um . In the case of SFFS, maximum intensity is observed
above the crest region which is just about 4 % of Um . It diffuses towards the outer region as
well as the bottom boundary further downstream. Due to the presence of waveform structure,
enhancement of turbulence intensity occurred which was more prominent in the case of VFFS
due to significant flow separation, recirculation and development of a thin shear layer. The
recovery of Iu at the flat surface beyond the downstream of the toe was much faster than Iw
for both the cases which was qualitatively similar with the study of [36].
A change in the Reynolds shear stress was noticed due to the local flow conditions at
different locations relative to the structural geometry. A strong enrichment was observed
between the locations G and J for the VFFS case compared to that of the SFFS, which
revealed a greater flow perturbation in that region. A clear peak in the vicinity of the crest
height (z/ h 0.13) for the VFFS case was observed, which was similar to the turbulent flow
over dunes using numerical technique obtained by Mendoza and Shen [38]. For our VFFS
case, the maximum uwdim achieved was about 15 times greater than that for the SFFS case.
Comparison with the study by Mazumder et al. [36] showed that for ebb flow condition, the
maximum uwdim is about two times greater than that for reverse flow condition over SFFS
as in our case. The recovery of the shear stress was much faster in the SFFS case than that of
the VFFS. The significant scattered points at the recirculation region just beyond the crest for
the VFFS case were noticed, which were due to the sudden change in the depth. For SFFS
case, the profiles of Reynolds shear stress were almost identical in pattern for all locations
along the flow.
A quadrant decomposition technique was adopted to explain the bursting phenomena
responsible for most turbulent production. The contributions of ejections and sweeps have
greater influence to the Reynolds shear stress overall except at the crest locations indicating
that the momentum transfer between the flow and the channel bed is mostly carried by
these two events. The crest locations can be treated as rough edges and hence it verifies that
contributions to the Reynolds shear stress by the outward and inward interactions become
significant and dominate over the other two events at the rough surfaces as reported by Kaftori
et al. [21]. It is in contrary with the findings of [36] in ebb flow condition, whose study showed
ejections and sweeps to be always dominant over and beyond the structure and have greater
influences to the Reynolds shear stresses than the other two interactions. Large-scale vorticity
is manifested as ejection event and arises both along the shear layer and at flow reattachment.
It should be noted that for the VFFS case the relative intensity of each event (specially
ejections and sweeps) for H > 0 is significantly greater as compared to the SFFS case. This
is probably due to relatively high turbulence level. As a result, the relative importance of
contributions of the turbulent events to the shear stress for VFFS case is much higher than
that for SFFS case. Contribution of ejections is clearly the most responsible event for moving
sediment particles away from the bed, and sweeps provide an avenue for entraining particles.
In the vicinity of the crest region for the SFFS case, outward and inward interactions
occupy around 30 % of the space, while in the case of VFFS at the same location, sweeps

123

Environ Fluid Mech

overshadow the other three. Although near the surface of the structure in the recirculation
region for the VFFS case, ejections are seen to occupy higher space about 40 % than sweeps
which is about 25 %. However, the sweeps and ejections extract energy from the mean flow to
create turbulence near the bed, which are considered to be more important than the interaction
events. In this region sweeps contribute more to the Reynolds shear stress than ejections as
we have discussed earlier. That emerges the fact that in this region, duration of sweeps are
larger.
The interactions between the turbulent coherent structures at different flow layers are
associated with the kolkboils phenomena. These coherent flow structures are advected with
the mean flow, often reaching the free surface and erupting as surface boils. The origin
of kolks and boils in the fluvial systems has been vigorously attributed to the interface of
ejections and sweeps over the trough region [40,41]. Existence of upward tilting stream-wise
vortex motion in the flow is known as kolk [33,43]; while boils are the structures like circular
or oval shaped patch lift up from the kolk on the water surface, which dissipate or merges with
the surroundings. Boils indicate the presence of macro-turbulence structures generated by the
dunes at the bed [19,24]. The formation of such macro-turbulent structures conceives from
the flow separation during high flows. Overall, sweeps are seen to have greater contribution
to the shear stress near the top surface in the case of VFFS and ejections largely contribute
at a slightly higher vertical height.
Bed forms in alluvial channels exerts flow resistance according to their shapes and sizes
and hence can greatly influence the sediment transport process. It is revealed from this study
that slope of the upstream-facing stoss-side has an effect on the flow resistance due to drag
and hence the sediment transport rate similar to the findings of [7,28,53]. So, it is an important
issue and of great importance in the filed for river engineering to predict channel evolution
process, river planning and restoration in the reverse flow condition/tidal flow. The flow
resistance is calculated from spatially averaged stream-wise velocity profiles for both the
cases and it is concluded that with increase in the stoss-side slope, there is an increase in
overall flow resistance and the drag.

5 Conclusions
The purpose of the present study was to ascertain the effect of different stoss-side slopes of
forward-facing bedform structures on the mean flow, turbulence and flow resistance/drag;
and thus to simulate of flow over such bedforms in tide/flood environments. Two isolated
rigid bedform structures with a common lee-side angle were considered under identical flow
conditions. An ADV was used to collect the velocity data at different horizontal locations to
examine the time-mean velocities, Reynolds stresses and contributions of burst-sweep cycles
to the Reynolds shear stress.
Comparative study shows that there were substantial changes in the turbulence properties
and drag coefficients associated with different stoss-side slopes especially in the near-bed
area and in the vicinity of the crest positions. Although the artificial rigid isolated upstreamfacing bedform structure is not the exact representative situation in the riverine environment,
this study provides a knowledge of flow characteristics and resistance due to form drag over
the structures in reverse/tidal flow condition. Several conclusions can be drawn from the
present experimental simulations:
The longitudinal velocity profiles for the bedform with smaller stoss-side angle (SFFS
case) follow the log-law (except at the crest location), similar to the observation of

123

Environ Fluid Mech

Lefebvre et al. [28]. However, for the structure with vertical stoss face (VFFS), the velocity
profiles follow the log-law up to just upstream of the corresponding crest location.
A permanent flow separation region with a recirculating eddy is observed at just downstream of the crest of the bedform with vertical stoss-side (VFFS case), while no flow
separation zone is observed over the bedform with smaller stoss-side angle (SFFS case).
Comparison with previous studies for ebb flow condition readily showed that the size and
the location of flow reversal zone are influenced by changes in bed geometry during the
tidal/flood flow condition. The threshold condition of the development of flow separation
zone in lee side of the bedform structures; and the dependence of stoss and lee angles
to turbulence parameters, drag coefficients during the tidal conditions are worthy of this
study.
Maximum value of the longitudinal and vertical turbulence intensities for the SFFS case
and the longitudinal turbulence intensity for VFFS attained at the vicinity of the respective
crest locations. However, the vertical turbulence intensity for VFFS case attained its
maximum at some distance downstream of the crest over the lee portion.
Maximum longitudinal intensity was five times greater than the maximum vertical intensity for the structure with smaller stoss-slope (SFFS case) for the case simulating a tidal
flow condition, which is greater in comparison with ebb flow condition [36] over similar
structure.
Greatly affected Reynolds shear stress was noticed due to different structural geometries
enhancing different local flow conditions. The maximum normalized Reynolds shear
stress for the structure with smaller stoss-slope is just about half of that for the ebb flow
condition [36] over the similar structure.
A quadrant decomposition technique was adopted to discriminate the turbulent events due
to different structures. Importantly in the vicinity of the crest locations for both structures,
inward and outward interactions were mostly contributed to the Reynolds shear stress.
Other-wise sweeps and ejections were the two most dominant events.
Near the bottom region, adjacent to the lee face, the sweeps have greater contributions
to the Reynolds shear stress for the structure with vertical stoss-face (VFFS), which
contributed to enhance the avenue of entraining particles and for the other structure
(SFFS), it was the ejections which is responsible for movement of sediment particles.
A great enhancement in flow resistance due to the structure with vertical stoss-face (VFFS
case) was observed and hence a reduction in overall drag was achieved in the case of
smaller stoss side sloped structure (SFFS case) as there is no flow separation.

This study could provide a better understanding of turbulence, if there would be a bed
covered by such waveforms rather than an isolated feature. It is of future interest to improve
our understanding of turbulence and resistance due to form drag over a series of forwardfacing bedform structures in the flood-tide condition; and to study the effect of different stoss
side angles of bedform structures. Detailed investigations are important to formulate a better
modelling of three-dimensional flow structures for different Reynolds numbers associated
with a series of forward-facing dunes oriented against the flow for a wide range of parameters,
which will help the researcher to understand the mean flow, turbulence and drag in the fields
of geology, geophysics, hydraulics and river engineering.
Acknowledgments One of the authors (BSM) would like to express his sincere thanks to Professor Stephen
G. Monismith, Director of Environmental Fluid Mechanics Laboratory (EFML), Stanford University, USA for
suggesting the problem and constructive technical discussions on the work during BSMs visits to the EFML,
Stanford University, California, USA. The authors would like to acknowledge the Department of Science
and Technology (DST), New Delhi for the financial support for this research work at the Fluvial Mechanics

123

Environ Fluid Mech


Laboratory (FML), Indian Statistical Institute, Kolkata, India; and many thanks to two anonymous referees
and Associate Editor, Professor Hubert Chanson for their detailed and constructive input on the paper.

References
1. Addad YD, Laurence CT, Jacob MC (2003) Large eddy simulation of a forwardbackward facing step
for acoustic source identification. Int J Heat Fluid Flow 24:562571
2. Agelinchaab M, Tachie M (2008) PIV study of separated and reattached open channel flow over surface
mounted blocks. ASME J Fluids Eng 130(6):061206061214
3. Ando T, Shakouchi T (2004) Flow characteristics over forward facing step and through abrupt contraction
pipe and drag reduction. Res Rep Fac Eng Mie Univ 29:18
4. Armaly BF, Durst F, Pereira JCF, Schonung B (1983) Experimental and theoretical investigation of
backward-facing step flow. J Fluid Mech 127:473496
5. Ashley GM (1990) Classification of large-scale subaqueous bedforms: a new look at an old problem.
J Sediment Petrol 60:160172
6. Barkley D, Gomes M, Henderson R (2002) Three-dimensional instability in flow over a backward-facing
step. J Fluid Mech 473:167190
7. Bennett JP (1997) Resistance, sediment transport, and bedform geometry relationships in sandbed channels. In: Proceedings of the U.S. Geological Survey (USGS) sediment, workshop, 47 Feb
8. Bennett SJ, Best JL (1995) Mean flow and turbulence structure over fixed, two-dimensional dunes:
implications for sediment transport and bedform stability. Sedimentology 42:491513
9. Best J (2005) The fluid dynamics of river dunes: a review and some future research directions. J Geophys
Res 110:F04S02. doi:10.1029/2004JF000218
10. Best J, Kostaschuk RA (2002) An experimental study of turbulent flow over a low-angle dune. J Geophys
Res 107(C9):18(9)18(19). doi:10.1029/2000JC000294
11. Camussi R, Felli M, Pereira F, Aloisio G, Marco AD (2008) Statistical properties of wall pressure fluctuations over a forward-facing step. Phys Fluids 20(7):075113
12. Casarsa L, Giannattasio P (2008) Three-dimensional features of the turbulent flow through a planar sudden
expansion. Phys Fluids 20(1):015103
13. Cellino M, Graf WH (2000) Experiments on suspension flow in open channels with bed forms. J Hydraul
Res 38(4):289298
14. Dinehart RL (2002) Bedform movement recorded by sequential single-beam surveys in tidal rivers.
J Hydrol 258:2539
15. Eaton JK, Johnston JP (1981) A review of research on subsonic turbulent flow reattachment. AIAA
J 19:10931100
16. Ernstsen VB, Noormets R, Winter C, Hebbeln D, Bartholoma A, Flemming BW, Bartholdy J (2006)
Quantification of dune dynamics during a tidal cycle in an inlet channel of the Danish Wadden Sea.
Geo-Mar Lett 26(3):151163
17. Gabel SL (1993) Geometry and kinematics of dunes during steady and unsteady flows in the Calamus
River, Nebraska, USA. Sedimentology 40(2):237269
18. Goring DG, Nikora VI (2002) Despiking acoustic Doppler velocimeter data. J Hydraul Eng 128(1):
117126
19. Jackson RG (1976) Sedimentological and fluid-dynamic implications of the turbulent bursting phenomenon in geophysical flows. J Fluid Mech 77:531560
20. Julien PY, Klaassen GJ (1995) Sand-dune geometry of large river during floods. J Hydraul Eng
121(9):657663
21. Kaftori D, Hetsroni G, Banerjee S (1998) The effect of particles on wall turbulence. Int J Multiphase
Flow 24(3):359386
22. Kiya M, Sasaki K (1983) Structure of a turbulent separation bubble. J Fluid Mech 137:83113
23. Kostaschuk RA, Best JL (2005) The response of sand dunes to variations in tidal flow. J Geophys Res
110:F04S04. doi:10.1029/2004JF000176
24. Kostaschuk RA, Church MA (1993) Macroturbulence generated by dunes: Fraser River, Canada. Sediment
Geol 85:2537
25. Kostaschuk RA, Villard PV (1996) Flow and sediment transport over large subaqueous dunes: Fraser
River, Canada. Sedimentology 43:849863
26. Largeau J, Moriniere V (2007) Wall pressure fluctuatings and topology in separated flows over a forwardfacing step. Exp Fluids 42(1):2140
27. Le H, Moin P, Kim J (1997) Direct numerical simulation of turbulent flow over a backward-facing step.
J Fluid Mech 330:349374

123

Environ Fluid Mech


28. Lefebvre A, Ernstsen VB, Winter C (2011) Influence of compound bedforms on hydraulic roughness in
a tidal environment. Ocean Dyn 61(12):22012210
29. Lefebvre A, Ernstsen VB, Winter C (2013) Estimation of roughness lengths and flow separation over
compound bedforms in a Natural Tidal inlet. Cont Shelf Res 6162:98111
30. Lefebvre A, Ferret Y, Paarlberg AJ, Ernstsen VB, Winter C (2013) Variation of flow separation over large
bedforms during a tidal cycle. In: Marine and river dune dynamics MARID IV, Bruges, Belgium, 15 and
16 April, pp 169175
31. Lu SS, Willmarth WW (1973) Measurements of the structure of the Reynolds stress in a turbulent boundary
layer. J Fluid Mech 60:481512
32. Lyn DA (1993) Turbulence measurements in open channel flows over artificial bed forms. J Hydraul Eng
119(3):306326
33. Mao Y (2003) The effects of turbulent bursting on the sediment movement in suspension. Int J Sediment
Res 18(2):148157
34. Martinius AW, Gowland S (2010) Tide-influenced fluvial bedforms and tidal bore deposits (Late Jurassic Lourinha Formation, Lusitanian Basin, Western Portugal). Sedimentology. doi:10.1111/j.1365-3091.
2010.01185.x
35. Mazumder BS, Ray RN, Dalal DC (2005) Size distributions of suspended particles in open channel flow
over bed materials. Environmetrics 16(2):149165
36. Mazumder BS, Pal DK, Ghoshal K, Ojha SP (2009) Turbulence statistics of flow over isolated scalene
and isosceles triangular-shaped bedforms. J Hydraul Res 47(5):626637
37. McLean SR, Wolfe SR, Nelson JM (1999) Spatially averaged flow over a wavy boundary revisited.
J Geophys Res 104(C7):1574315753
38. Mendoza C, Shen HW (1990) Investigation of turbulent flow over dunes. J Hydraul Eng 116(4):459477
39. Nakagawa H, Nezu I (1981) Structure of spacetime correlations of bursting phenomena in an openchannel flow. J Fluid Mech 104:143
40. Nezu I, Nakagawa H (1993a) Three-dimensional structures of coherent vortices generated behind dunes
in turbulent free-surface flows. Refined flows modelling and turbulence measurements. In: Proceedings
of the 5th international symposium, Paris, pp 603612
41. Nezu I, Nakagawa H (1993b) Turbulence in open channel flows. IAHR monograph. Balkema, Rotterdam.
ISBN 90-5410-118-0
42. Nezu I, Rodi W (1986) Open-channel measurements with a laser Doppler anemometer. J Hydraul Eng
112(5):335355
43. Ojha SP, Mazumder BS (2008) Turbulence characteristics of flow region over a series of 2-D dune shaped
structures. Adv Water Resour 31:561576
44. Paarlberg AJ, Dohmen-Janssen CM, Hulscher SJMH, Termes P (2009) Modelling river dune evolution
using a parameterization of flow separation. J Geophys Res 114:F01014
45. Pearson DS, Goulart PJ, Ganapathisubramani B (2011) Investigation of turbulent separation in a forwardfacing step flow. In: 13th European turbulence conference. J Phys Conf Ser 318. doi:10.1088/1742-6596/
318/2/022031
46. Poole RJ, Escudier MP (2003) Turbulent flow of a viscoelastic shear-thinning liquid through a plane
sudden expansion of modest aspect ratio. J Non-Newton Fluid Mech 112:126
47. Ren H, Wu Y (2011) Turbulent boundary layers over smooth and rough forward-facing steps. Phys Fluids
23(4):117
48. Saleel CA, Shaija A, Jayaraj S (2011) Numerical simulation of fluid flow over a forwardbackward facing
step using immersed boundary method. Int J Eng Sci Technol 3(10):77147729
49. Sherry MJ, Jacono DL, Sheridan J (2010) An experimental investigation of the recirculation zone formed
downstream of a forward facing step. J Wind Eng Ind Aerodyn 98:888894
50. Singh AP, Paul AR, Ranjan P (2011) Investigation of reattachment length for a turbulent flow over a
backward facing step for different step angle. Int J Eng Sci Technol 3(2):8488
51. Smith JD, McLean SR (1977) Spatially averaged flow over a wavy surface. J Geophys Res 82(12):
17351746
52. Stuer H, Mass HG, Virant M, Becker J (1999) A volumetric 3D measurement tool for velocity field
diagnostic in microgravity experiments. Meas Sci Technol 10:904913
53. Toyama A, Shimizu Y, Yamaguchi S, Giri S (2007) Study of sediment transport rate over dune-covered
beds. In: 5th IAHR symposium on river, coastal and estuarine morphodynamics. University of Twente,
Enschede
54. Vanoni V, Hwang LS (1967) Relation between bed form and friction in streams. J Hydraul Div Am Soc
Civ Eng 93(HY3):121144
55. Venditti JG (2007) Turbulent flow and drag over fixed two- and three-dimensional dunes. J Geophys Res
112:F04008

123

Environ Fluid Mech


56. Winter C, Ferret Y, Lefebvre Y, Ernstsen VB (2013) Geometric properties of hydraulically-relevant tidal
bedforms. In: Marine and river dune dynamics MARID IV, Bruges, Belgium, 15 and 16 April, pp 337338
57. Wu Y, Christensen KT (2006) Reynolds-stress enhancement associated with a short fetch of roughness
in wall turbulence. AIAA J 44(12):30983106
58. Zarillo GA (1982) Stability of bedforms in a tidal environment. Mar Geol 48:337351

123

You might also like