Coastal Engineering 116 (2016) 258–274

Contents lists available at ScienceDirect

Coastal Engineering

journal homepage: www.elsevier.com/locate/coastaleng

An experimental study on near-orthogonal wave–current interaction

over smooth and uniform fixed roughness beds
Kian Yew Lim a,⁎, Ole Secher Madsen b
a
Department of Civil and Environmental Engineering, National University of Singapore, 2 Engineering Drive 2, EW1 03-01A, 117577, Singapore
b
Department of Civil and Environmental Engineering, Ralph M. Parsons Laboratory, 48-317, MIT, Cambridge, MA 02139, USA

a r t i c l e i n f o a b s t r a c t

Article history: An experimental study of periodic waves interacting with near-orthogonal turbulent currents is presented in this
Received 31 December 2014 paper. Mean velocity profiles for current-alone, wave-alone and combined wave–current flows are measured
Received in revised form 3 May 2016 with Acoustic Doppler Velocimeters to resolve the changes in mean flow kinematics due to wave–current inter-
Accepted 11 May 2016
action. In this study, the near-bottom wave orbital velocity is approximately 1.5 times greater than the depth-
Available online 5 August 2016
averaged current velocity, and the log-profile method is used to determine the bottom roughness from the mea-
Keywords:
sured velocity profiles. The primary focus is on 90° wave–current interaction, while selected findings for 60° and
Wave–current interaction 120° wave–current orientations are also presented. In the smooth bed experiment, the interaction is shown to be
Boundary layer flow linear due to the relatively low Reynolds-number flows generated in the facility — a relatively common scenario
Bottom roughness in small-scale laboratory setups that has not received much attention in previous studies. The smooth turbulent
Grant–Madsen model flow equation, with modification of the input shear velocity, is found to accurately predict the mean flow rough-
ness for linear interaction. The bottom roughness is subsequently increased with the introduction of a layer of
uniform 12.5 mm marbles to achieve a more realistic rough turbulent flow regime. The results agree qualitatively
with previous experimental findings that showed a reduction in the near-bottom mean velocity due to a wave-
enhanced (apparent) roughness. However, the Grant–Madsen (GM) model is found to over-estimate the appar-
ent roughness when the angle of wave–current interaction is large, implying a lack of directional sensitivity of
this model under “strong-wave, weak-current” conditions. A tentative explanation of this shortcoming is given
in the paper. However, it is not possible to extend the GM model analysis to the present near-orthogonal
wave–current flows, as results suggest that both 60° and 120° wave–current cases are sufficiently contaminated
by the wave-induced mass transport component in the current direction to invalidate the use of the log-profile
method to resolve the bottom roughness. In addition to the modification of current profiles by waves, the present
study also shows a drastic transformation of wave-induced mass transport profiles by the external turbulent cur-
rents. The veering of mean flow over depth is consistent with the superposition of a wave-induced return flow on
the external current, while additional veering in the near-bottom region may be attributed to turbulence asym-
metry induced by the current component in the direction of the near-bottom wave orbital velocity which is
shown to vary locally by ±10° due to parasitic waves emanating from the current inlet and outlet.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction
It is commonly recognized that coastal circulation and sediment
transport are significantly influenced by combined waves and current
flows. In coastal waters, the direction of wave propagation is predomi-
nantly near-orthogonal to currents, and both flows exhibit considerable
differences in time scales. The period, T, of coastal currents is of the
order of several hours, and the boundary layer thickness, which is
pffiffiffi
roughly proportional to T, is of the order of meters. Wind-wave pe-
riods are usually less than 20 s and suggest that the wave boundary
layer thickness is only of the order of centimeters. Hence, wave–current
⁎ Corresponding author.
E-mail addresses: dprlky@nus.edu.sg (K.Y. Lim), osm@mit.edu (O.S. Madsen).
interaction is restricted to a thin layer of water column close to the sea-
bed, further complicated by the presence of bedforms in many cases.
Since this interaction is a highly nonlinear process, the combined
wave–current flow is considerably different from a direct summation
of current-alone and wave-alone kinematics, implying that meaningful
studies require both currents and waves to be present simultaneously.
Immense progress has been made in the last few decades on the
simulation of wave–current flows, starting from the simple models de-
veloped in the 1970s and 1980s [Lundgren (1972), Smith (1977), Grant
& Madsen (1979), Fredsøe (1984), Christofferson & Jonsson (1985),
Coffey & Nielsen (1986), Myrhaug & Slattelid (1989)], to more recent
numerical models that adopt state-of-the-art turbulence closure
methods, e.g. Holmedal et al. (2013) for collinear wave–current flows,
Davies et al. (1988), Olabarrieta et al. (2010) and Afzal et al. (2015)

http://dx.doi.org/10.1016/j.coastaleng.2016.05.005
0378-3839/© 2016 Elsevier B.V. All rights reserved.
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
for arbitrary angles of wave–current flows. Detailed review of this sub-
ject can be found in Grant & Madsen (1986) and Soulsby et al. (1993). A
primary strategy of early-day theoretical studies was to adopt simple
turbulence closure models when solving the Navier–Stokes equations
for both steady and oscillatory components of the combined flow. The
available closure models include the eddy viscosity model, the turbulent
kinetic energy closure method and the momentum-deficit integral
method (Soulsby et al., 1993). The eddy viscosity model is one of the
simplest to use, and wave–current models based on this scheme are
generally integrated into large-scale ocean circulation models to en-
hance computational efficiency of numerical simulation.
Validation of theoretical models has been performed mostly with
collinear wave–current experiments (e.g. Kemp & Simons, 1982, 1983;
Klopman, 1994; Mathisen & Madsen, 1996a, b; Fredsøe et al., 1999),
though a fair amount of experimental work involving near-orthogonal
wave–current interaction has also been conducted (e.g. Havinga,
1992; Simons et al., 1992; Arnskov et al., 1993; van Rijn & Havinga,
1995; Musumeci et al., 2006; Madsen et al., 2008; Fernando et al.,
2011). The general conclusion is that when waves are present with cur-
rents, the near-bottom mean velocity increases (decreases) over
smooth (rough) beds. Besides, collinear wave–current studies have
also demonstrated that for currents following (opposing) the waves,
the near-surface mean velocity decreases (increases) with respect to
current-alone conditions (Kemp and Simons, 1982, 1983; Klopman,
1994). When waves interact with currents in orthogonal or near-
orthogonal orientations, the mean flows would experience changes
not only in terms of magnitude but also in the flow directions due to
the superposition of wave-induced return flow on the nominal current,
as well as steady streaming induced by momentum transfer (Longuet-
Higgins, 1953 — henceforth LH53) and turbulence asymmetry
(Trowbridge & Madsen, 1984) within the wave boundary layer. Numer-
ical simulations by Davies et al. (1988 — henceforth DSK88), accounting
only for turbulence asymmetry, and Afzal et al. (2015 — hereafter
AHM15) who also included mean momentum transfer, have revealed
details of this veering of the mean flow for non-collinear wave–current
flows. However, to the authors' knowledge, these details have not been
supported by experimental evidence.
The comparison between theoretical models and experimental re-
sults require a precise knowledge of the bed resistance over which the
flows interact, and this is commonly expressed as the equivalent
Nikuradse sand grain roughness (henceforth bottom roughness, kn).
For the case of steady rough turbulent flows over a uniform flat sand
bed, kn is characterized by the diameter of the sand grains, with numer-
ous studies suggesting that kn is of the order 1 to 10 d50 (Nielsen, 1992).
Several notable studies on oscillatory flows over movable beds (Carstens
et al., 1969; Lofquist, 1986) conclude that kn is approximately four times
the ripple height, η, when bedforms are present. However, for oblique
wave–current flows in a basin, the movable bedforms tend to exhibit
greater spatial variability (Madsen, et al., 2008), which results in ambigu-
ity of the value of kn. This problem can be partly overcome by using fixed
2D bedforms, but strong directional-dependency of kn is usually ob-
served for this type of bed configuration due to drastic changes in bed re-
sistance when the angle between mean flow and bedform axis varies
(Barrantes & Madsen, 2000). The above shortcomings may be resolved
by generating wave–current flows over a uniform, fixed 3D bottom con-
figuration, where kn is spatially homogenous as well as directional-
independent. Unfortunately, experimental measurements involving
near-orthogonal waves and currents over this type of bed configuration
are relatively scarce.
The aforementioned gap motivates the present study, which is
aimed at obtaining high-quality experimental data for near-
orthogonal wave–current flows over smooth and uniform fixed rough-
ness beds. The focus is on waves interacting with currents at 90°, and a
large number of near-bed measurement data were collected to facilitate
the determination of bottom roughness with the log-profile analysis.
Selected findings on 60° and 120° wave–current flows are also
259
discussed. In addition to the modification of current flows by waves,
the present study is also aimed at understanding the changes in wave-
induced mass transport flows due to wave–current interaction, a phe-
nomenon that has not received much attention in past studies but
could have significant implications for cross-shore sediment transport
since there is a strong dependency of bedload transport on near-bed
mean currents.
The organization of this paper is as follows: Section 2 discusses the
experimental setup and preliminary tests used to establish the mea-
surement regions and uncertainties. The smooth bed experiments for
90° wave–current interaction are detailed in Section 3. Section 4 pre-
sents the 90° flow experiments conducted over a uniform fixed rough-
ness bed and compares them with predictions afforded by the GM
model. Potential mechanisms influencing the veering of near-bottom
mean flow are also discussed. Selected results on 60° and 120° wave–
current interaction are presented in Section 5, with discussions of the
contamination of log-profile analysis by near-orthogonal waves. The
final section presents conclusions of the study and recommendations
for future work.
2. Experiments and methodology
2.1. Experimental setup
Experiments were performed in the Hydraulic Engineering Labora-
tory of the Department of Civil & Environmental Engineering, National
University of Singapore (NUS). Current-alone, wave-alone and com-
bined wave–current flows were generated in a 33 m (L) × 10 m
(W) × 0.9 m (D) wave–current basin. A photo of the facility is shown
in Fig. 1 while the overall layout is presented in Fig. 2. Water depth, h
was maintained at 0.4 m for all experiments. A section of the basin
(19 m × 1.5 m) parallel to the direction of the wave propagation was
converted into a reservoir to generate steady current flows in the test
area. The flows exited the basin over an adjustable weir at the outlet
and were re-circulated to the reservoir by two 75HP centrifugal
pumps. Current inlets of width 2.5 m were constructed along the wall
of the reservoir and aligned such that currents would intersect with
waves at 60°, 90° and 120° when both flows were present. For notation
purpose, the current channel that connects Inlet 2 with the outlet is
known as the 90° current channel, while those that connect Inlets 1
and 3 with the outlet are denoted as the 60° and 120° current channel,
respectively (refer to Fig. 2). Honeycomb filters consisting of 50 cm-
long, 5 cm-diameter PVC pipes were installed at the inlets to ensure
Fig. 1. Bird's eye view of the wave–current basin in the Hydraulic Engineering Laboratory,
NUS.
260 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274

Fig. 2. Layout of wave–current basin. Shaded areas in the middle of the basin denote the near-uniform current flow regions (to be discussed in Section 2.3.2).

uniformity and directionality of the inflows. During the experiment, the
flow discharge was constantly monitored with an electromagnetic flow
meter. Two types of bed conditions were examined in the present study,
i.e. a smooth concrete bed and a uniform roughness bed with ceramic
marbles of d50 = 12.5 mm covering the concrete bottom.
Waves were generated in the basin by a multi-element, piston-type
wave-maker, each paddle controlled by an electric-servo motor. A 1:8
passive absorbing beach was installed at the end opposite from the pad-
dles for wave absorption, with reflection coefficient estimated to be less
than 8% for the waves generated in this study. The wave paddles were
located at least 2 m away from the current channels, ensuring that the
influence of evanescent waves, which are significant within ~ 3 times
the water depth from the wave-maker, can be safely neglected in the
measurement regions (Madsen, 1970). The wave and current condi-
tions generated in the 90° current channel are summarized in Table 1,
and the coordinate system used is: positive-x in the direction of the
wave propagation, positive-y in the direction perpendicular to the x-
axis in a horizontal plane towards the outlet, and positive-z vertical up-
ward from the theoretical bottom. The theoretical bottom for smooth
bed is assumed to be the same as the concrete floor level, and for the
uniform roughness case it is taken as 0.3 diameter below the top surface
of marbles (see Section 2.4 for further description). Velocity components
along the x-, y- and z-axes are denoted by u, v and w, respectively, while
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
the resultant horizontal mean velocity is denoted by uR (¼ u2 þ v2 Þ.
From an aerial view of the basin, the angle of mean flow, θR, increases
Table 1
Experimental conditions in the 90° current channel.
Currents 
Measured depth-averaged velocity, U
(cm/s)
CS: smooth bed 11.0 ± 0.1
CU: uniform roughness bed 10.6 ± 0.2
Waves T(s) ubm (cm/s) Rew = ubmυAbm
WS: smooth bed 1.4 16.9 ± 0.6 7,800
WU1: uniform roughness bed 1.4 16.6 ± 0.5 7,700
WU2: uniform roughness bed 1.6 17.7 ± 0.6 9,900
Wave–current
WCS: CS + WS
WCU1: CU + WU1
WCU2: CU + WU2
T = wave period; ubm = measured near-bottom orbital velocity amplitude
υ = kinematic viscosity = 0.008cm2/s at temperature of 30 °C.
Abm = near-bottom particle excursion amplitude = ubmT/2π.
anticlockwise from the positive x-direction, which is the direction of
wave propagation (θR = 0°).
The presence of current along the direction of wave propagation (for
oblique wave-current cases) may change the wave properties shown in
Table 1, but this was found to be negligible. For instance, with Ta =1.4 s
and ωr ¼ ωa −ku, where ωr is the relative radian frequency and ωa is the
absolute radian frequency (without current), the relative wavelength,
Lr, is approximately 2.5 m, which is merely 4% different from the abso-
lute wavelength, La (when waves are alone).
2.2. Velocity and free surface elevation measurement
Velocity measurements in the basin were performed with Acoustic
Doppler Velocimeters (Nortek's Vectrino Plus model). The sampling
volume was cylindrical in shape with a diameter of 0.6 cm and length
of 0.7 cm. It was approximately 5 cm below the probe tip to minimize
interference of the flow field by the physical body of the probe. The sam-
pling frequency was fixed at 200 Hz in all tests. Fine clay powder was
used as seeding particles to enhance signal reflection from the flow
and hence the quality of measurements. Data with an acceleration of
more than 1 g (981 cm/s2) were classified as spikes and filtered from
the record. Fourier analysis was then conducted on the remaining data
to obtain time-averaged and harmonic velocities.
With the sampling volume located 5 cm below the probe tip, the up-
permost 5 cm of the water column was out of the measuring range.
When waves were present, the highest measurable point was 5 cm
below the wave trough, since the probes were required to be sub-
merged at all time. Separate instruments were normally used for near-
surface flow measurements. For instance, Fredsøe et al. (1999) used a
Rec ¼ Uh

bi-directional micro-propeller to measure the flow between wave crests
υ
and troughs, and a Laser Doppler Anemometer (LDA) to cover the re-
55,000 maining part of the water column. Since the present study is concerned
53,000 mainly with the near-bottom mean flow kinematics, in particular the
region where the logarithmic velocity profile is expected to prevail, no
Abm
kn further attempt was made to measure the near-surface flow.
– The free surface elevation was measured by several capacitance-
1.24 type wave gauges at a frequency of 50 Hz. Up to three wave gauges
1.50
were set-up on tripods close to the wave-maker to monitor the station-
arity of incident waves, while an additional wave gauge (alongside
three units of ADVs) was mounted on the transverse carriage (see Fig.
1) and moved along the x- and y-axis of the wave–current basin for de-
tailed measurement within the current channels. The wave gauge was
fixed at a specific elevation, whereas the ADVs were connected to ad-
justable point gauges to allow changes of probe elevation to capture ve-
locity profiles over depth.
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
2.3. Preliminary tests
2.3.1. Sampling duration
To establish a sampling duration for velocity measurements, an
hour-long velocity record was collected for a steady current flow condi-
tion in the basin. The record was then subdivided into smaller time se-
ries to determine a duration which gave a mean velocity with an
acceptable level of uncertainty, i.e. similar to the approach adopted by
Mathisen & Madsen (1996a) and Chanson (2008). The standard devia-
tions, σ, of the sub-divided time series are tabulated in Table 2. When
the sampling duration is 3 min, σ is approximately 0.3 cm/s, which is
less than 3% of the time-averaged current velocity. By shortening the
sampling duration to 1 min, σ remains less than 5% of the time-
averaged velocity for the hour-long measurement record.
For consistency, a fixed sampling duration was used in current-
alone, wave-alone and wave–current experiments. The duration should
therefore cover a sufficient number of wave cycles to allow accurate de-
termination of wave properties. In this study, waves with periods of
1.4 s and 1.6 s were generated. Based on the suggestion of Sleath
(1987) that no less than 50 wave cycles should be used for phase aver-
aging, a 3-min sampling duration, corresponding to more than 100
wave cycles, was adopted for all measurement cases in the present
study.
2.3.2. Near-uniform flow regions
The inlet conditions were adjusted such that the near-uniform flow
region in each current channel was established close to the center of the
wave–current basin. These regions should be sufficiently far from the
inlet to allow the development of current boundary layers, yet main-
taining an adequate distance from the outlet to avoid the influence of
the weir. The boundary layer thickness, δc, is predicted from the formu-
las and graphs given by Schlichting (1968), an approximation of which
was obtained by Madsen (2009) as follows:
 0:08
υ 0:92
δc ¼ 0:09 l for smooth bed
U∞
0:16 0:84
δc ¼ 0:13ðkn Þ l for rough bed
where U∞ is the free-stream velocity (approximately equal to the depth-
averaged velocity) and l is the distance from the inlet along the current
channel axis (l = y for the 90° case). Based on Eqs. (1) and (2), the
boundary layer thickness is expected to exceed 10 cm at locations
more than 3.5 m away from the inlet. Since this should give a satisfacto-
ry number of measurement points for log-profile fitting, the upstream
and downstream limits of the measurement region along the y-axis
were fixed at 3.5 m and 5 m, respectively, from the inlet. In addition,
preliminary measurements were conducted along the spanwise direc-
tion of the current channel (x-axis) to establish the up-wave and
down-wave limits of the near-uniform flow region. The length of the
measurement domain along x-axis was progressively narrowed until
an acceptable standard deviation of the order 0.5 cm/s was achieved.
Using this approach, a width of 0.5 m was obtained for the measure-
ment region in the 90° current channel, i.e. 0.25 m on either side of
Table 2
Standard deviations for datasets subdivided from an hour-long current velocity
measurement.
Sampling duration (minutes) Standard deviation (cm/s)
10 0.24
5 0.25
3 0.33
1 0.47
261
the current channel's centerline. A similar approach was used to identify
the near-uniform flow regions in the 60° and 120° channels. The distri-
bution of measurement stations within the near-uniform flow regions
in the three current channels are shown in Fig. 3. Up to 25 points
were measured along the water column at each station to obtain well-
defined velocity profiles and ensure that sufficient data points were
available for log-profile fitting. The upper limit for log-profile fitting
was set at 1/3 of the current boundary layer thickness predicted with
Eqs. (1) and (2), while the lower limit was taken as twice the diameter
of the roughness element, or 7 mm from the concrete floor for smooth
bed experiments. Adopting these spatial limits for log-profile fitting typ-
ically resulted in 6–9 data points for smooth bed cases and 11–13 data
points for rough bed cases. For wave–current flow, the lower limit
would be dictated either by the above criterion or the wave boundary
layer thickness, whichever was larger.
An example of the spatially-averaged velocity profile over a uniform
roughness bed is presented in Fig. 4a. For clarity, data points are
grouped into vertical bins of size 1 cm when presented in spatially-
averaged velocity plots (i.e. points within ±0.5 cm from a mean z posi-
tion would be averaged and shown as one data point). Since the near-
uniform flow regions were first established based on current-alone
flows, the uniformity of these regions for combined wave–current sce-
narios was also checked in subsequent tests. It was found that the
near-uniform flow regions were generally valid for both current-alone
and wave–current flows. For instance, the standard deviations of
wave–current mean flow over uniform roughness bed in the 90° current
channel (Fig. 4b), while obviously greater than their current-alone
counterparts, are still of the order 0.5 cm/s and comparable to the tem-
poral uncertainty of the measurement.
However, in the experiment for 90° wave–current flow over smooth
bed, three up-wave locations in the measurement region (along
x = −0.25 m in Fig. 3a) were found to have notably smaller mean ve-
locities than the other six measurement stations. For this particular
case, the measurement region was “truncated” by omitting the stations
located in the non-uniform section of the region. While the reason for
ð1Þ this deviation is unknown, it is unlikely to have significant implications
since the remaining number of measurement stations in the “truncated”
region is sufficiently large to establish the wave–current flow
ð2Þ
characteristics.
2.4. Log-profile analysis and bottom roughness
Based on Prandtl's mixing length hypothesis and assumption of a
constant shear stress layer close to the bed, the near-bed mean velocity
distribution can be described by the logarithmic law:
 
uc z
uR ¼ ln ð3Þ
κ z0
where u⁎c is the current shear velocity, κ is the von Karman's constant
(= 0.4), z = z′ + Δ z where z′ = elevation from a fixed datum and
Δz = distance between the theoretical bottom and the datum. The log-
arithmic velocity profile is in principle valid only over a small fraction of
the bottom boundary layer, with upper limits ranging from 0.1δc
(George, 2007) to 0.3δc (Fredsøe et al., 1999). For rough bed scenarios,
a major difficulty in applying Eq. (3) is the ambiguous location of the
theoretical bottom where the no-slip boundary condition applies.
Grass (1971) proposed an approach to resolve this issue, which involves
the plotting of semi-log profiles with different assumptions of Δz and
identifying the value of Δz that yields the best-fit of Eq. (3) to the exper-
imental data. Jackson (1981) concluded that by taking the theoretical
bottom as 0.3D below the top surface of the roughness elements, with
D representing the characteristic length (height or diameter) of the
roughness elements, satisfactory results can be obtained for many com-
mon types of rough beds. This level was also found to be a good repre-
sentation of the theoretical bottom by Yuan et al. (2012), who
262 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274

Fig. 3. Measurement stations within the near-uniform flow regions for (a) 90° current channel (b) 60° current channel (c) 120° current channel. x′ denotes the local channel axis, with
origin along the centreline of the current channel.

performed their rough bed experiment with a single layer of uniformly-
arranged marbles identical to those used in the present study. Based on
these earlier works, an elevation of 0.3D below the top surface of the
roughness elements was adopted as the theoretical bottom for the pres-
ent uniform roughness experiment. For the smooth bed experiment, the
theoretical bottom is intuitively the same as the concrete floor level of
the basin.
Once Δz was established, Eq. (3) was used to determine the bottom
roughness, kn(=30z0). Conventionally, the log-profile fitting of experi-
mental data with Eq. (3) is done with uR as the dependent variable and
ln(z) as the independent variable. An undesirable outcome of this
approach is the relatively large uncertainty for z0 when the roughness
is small and there are few data points available for regression analysis,
notably for smooth bed cases where the boundary layer growth is
slow. An alternative method of analysis, with ln(z) as the dependent
variable, was performed. Comparison shows that the uncertainty for
z0 obtained with the latter approach is considerably smaller than the
former, with error factors reduced by 2 to 4 times for most locations.
Since z0 is the parameter of interest in this study, the use of ln(z) as
the dependent variable was chosen for smooth bed analysis. The differ-
ence between the two methods is, however, insignificant for the case of
uniform roughness bed, since the bottom roughness is significantly

Fig. 4. Mean velocity, uR, and one-standard-deviation error bar for (a) current-alone (b) wave–current mean flow over a uniform roughness bed in the 90° current channel. Similar results
were obtained for 60° and 120° current channels. Dashed lines indicate the regions analyzed with log-profile method.
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
larger and the boundary layer development is accelerated by the pres-
ence of marbles, resulting in a larger number of data points within the
boundary layer and a better quality of log-profile fitting.
Two filtering schemes were used in this study to eliminate outliers.
First, if the shear velocity, u⁎c, determined from log-profile analysis at
a measurement station had a 95% Confidence Interval (C.I.) exceeding
50% of the mean value, the station would be dropped from further anal-
ysis. The remaining stations that satisfied the first test were then
checked with the Median Absolute Deviation method (Pham-Gia &
Hung, 2001), which helped to eliminate spatial outliers before the
spatially-averaged results were computed. In most cases, the number
of locations rejected by this criterion was at most two (out of the nine
measurement stations available).
For measurement stations that satisfied both filtering schemes de-
scribed above, the bottom roughness at these stations were converted
into natural log prior to averaging, i.e.:
1 X
lnkn ¼ ln kn ¼ M ð4Þ
n notes one standard deviation of the data. Note that the uncertainty is
X 2
ln kn − ln kn
σM ¼2
: ð5Þ
n−1
When the results were converted back to linear scale, the spatially-
averaged kn was obtained together with a standard deviation factor,
eσM (hereafter referred to as the error factor):
8 
< eM  eσ M −upper limit of CI
kn ¼ eM ð6Þ
: M σM
e =e −lower limit of CI
The advantage of this approach is that the lower confidence limit of
kn remains positive at all time, which is desirable as negative values of kn
are physically unrealistic.
3. Orthogonal (90°) wave–current flow over smooth bed
3.1. Current-alone
 is approximately 11 cm/s
The depth-averaged current velocity, U,
and measurements were made in the near-uniform flow regions
shown in Fig. 3. An example of the velocity profile, plotted in semi-log,
is presented in Fig. 5 for current-alone in the 90° current channel.
From analysis, the spatially-averaged kn is 0.059 cm with an error factor
of 2.0, while the current shear velocity, u⁎c = 0.52 cm/s ± 9.1%. The
smooth turbulent bottom roughness, kn , smooth, can be predicted with
Fig. 5. An example of current velocity profile over a smooth bed in the 90° current channel.
Points located between the two dashed lines are used for log-profile fitting.
263
the following expression (Schlichting, 1968):
υ
kn;smooth ¼ 30z0 ¼ 3:3 ð7Þ
uc
which gives kn,smooth = 0.051 cm, i.e. only 15% smaller than the exper-
imental kn and well within the experimental confidence interval. Hence,
it can be concluded that the current-alone flow experiences a smooth
turbulent roughness when generated over the concrete bed of the basin.
3.2. Wave-alone
Wave-alone measurements were performed at measurement sta-
tions within each of the near-uniform flow region shown in Fig. 3. In
order to maintain the same sidewall boundary conditions as wave–cur-
rent experiments, the current inlet and outlet were left open during
wave-alone experiments. The wave period, T, is 1.4 s and spatially-
averaged ubm is 16.7 ± 1.2 cm/s (or ±7.2%), where the uncertainty de-
greater than that shown in Table 1, as it is computed based on measure-
ment stations across all the three channels illustrated in Fig. 3. Rew =
ubmAbm/υ is therefore less than 9,000 and the wave-alone boundary
layer is expected to be laminar based on the regime delineation pro-
posed by Jonsson (1966) and Davies & Villaret (1997).
Fig. 6 shows the spatially-averaged wave-induced mass transport
flow profile. The conduction solution proposed by LH53 for the Eulerian
wave-induced mass transport flow may be expressed as follows:
a2 ωk 
uE ¼ 2
3 þ kh sinhð2khÞ 3μ h 2 −4μ h þ 1 ð8Þ
4 sinh kh
 
sinhð2khÞ 3  2
þ3 þ μ h −1
2kh 2
where a is the wave amplitude, ω is the wave radian frequency, k is the
wave number, μh = 1 − z/h and the near-bottom velocity (μh = 1) is
3ubm 2 =4c (c = wave celerity = ω/k). Since the LH53 solution
was computed based on measured ubm, the results are subject to uncer-
tainty in establishing the values of ubm in the basin, e.g. due to the con-
tamination by wave reflection. This uncertainty can be quantified in
terms of the standard deviation of measured near-bottom orbital veloc-
ity at all stations, which is approximately 7.2% of ubm. The uncertainty of
 E, which is scaled by ubm 2, should therefore be approximately twice the
u
error of ubm, i.e. 14.4%. The theoretical result is plotted in Fig. 6 with the
experimental data.
It should be noted that the flow generated in the present experiment
is turbulent, since the Stokes' prediction of a wave-induced return
Fig. 6. Wave-induced mass transport flow over a smooth bed. Error bars for experimental
data represent one standard deviation. Error bars for the theoretical solution are
associated with ±14.4% of u. Cross: Experimental data; Full Line: LH53 model.
264 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
 ES , in a closed basin (Nielsen, 1992):
current, u
gH 2
 ES ¼−
u ð9Þ
8ch
where H is the wave height (~10 cm), c is the wave celerity (=1.7 m/s)
and h is the water depth (=0.4 m), suggests a simple return current of
1.8 cm/s for our experimental conditions, with a corresponding Reh ¼
 ES h=υ of almost 9,000. On the other hand, the conduction solution pro-
u superimposed at an angle of 90° for case WCS, defined in Table 1. The
posed by LH53 is, strictly speaking, applicable only to laminar waves with
a/δw bb 1 (where δw is the wave boundary layer thickness). Nonetheless,
the following reasons help justify the use of LH53 model for comparison
with the present experimental results. Since the distance between the
measurement region and the passive absorbing beach was more than
10 m, the mass transport current boundary layer predicted by Eq. (1)
with U∞ = u  ES = 1.8 cm/s is expected to be more than 40 cm in the mea-
surement regions. Besides, as measurements were made at least 30 min
after start of wavemaker, our setup and procedures suggest that the vor-
ticity generated at the bed had diffused over the entire depth at the time
and locations of measurement. Thus, treating the turbulent mass trans-
port current as “pseudo-laminar” supports our adoption of the LH53 con-
duction solution for comparison with our data.
As observed in Fig. 6, the measured wave-induced mass transport flow
exhibits a near-linear variation over depth and a streaming velocity near
the smooth bed of about 0.7 cm/s, which is 40% less than the theoretical
result of LH53. As a consequence of LH53's surface boundary condition
for the mean flow, which amounts to imposing a mean shear stress in
the direction of wave propagation, the LH53 model predicts a reduction
of the offshore-directed mean flow as the free surface is approached, i.e.
for z N ~0.6h in our experiment, a trend not exhibited in our measure-
ments. However, our measurements extended only to ~0.2h below the
free surface, so it cannot be precluded that the mean flow in the offshore
direction is reduced in the immediate vicinity of the free surface when the
mean flow is turbulent. Similar observations were made by the authors
(Lim, 2013) in a flume experiment (h = 0.5 m) for wave periods ranging
from 1.0 s to 2.6 s (0.57 b kh b 2.10), and also in previous studies, e.g.
Nadaoka & Kondoh (1982), Klopman (1994) and Scandura & Foti
(2011). These observations of a continuous linear decrease of the mean
flow up to ~ 0.2h below the free surface may, at least partially, be ex-
plained by the observations for collinear wave–current interaction
(Klopman, 1994), where the magnitude of the near-surface current is
shown to increase in the presence of opposing waves. For the case of
waves interacting with a turbulent wave-induced return current, the
flow orientation is exactly the same as waves opposing a current, but
the near-surface characteristic described above is not accounted for by
Fig. 7. (a) Mean velocity, uR (b) Angle of mean flow, θR, for current-alone and combined wave–current flows over a smooth bed. Full circle: current-alone; Hollow square: wave–current.
the laminar LH53 model. The present experimental results suggest that
the “Klopman effect”, e.g. Huang & Mei (2003), must be included in a the-
oretical model of wave-induced mass transport currents for such a model
to be valid over the entire water depth.
3.3. Wave–current interaction
The current and wave flows described in Sections 3.1 and 3.2 were
measured ubm during wave–current interaction was 16.4 cm/s, which
is well within the uncertainty of ubm obtained for the wave-alone case.
Fig. 7a depicts the current-alone and wave–current mean velocity pro-
files, and clearly demonstrates that the introduction of waves on an or-
thogonal smooth turbulent current leads to an increase in the near-
bottom mean velocity. Correspondingly, the near-surface flow de-
creases in order to maintain a constant discharge along the current
channel. The increase in near-bottom mean flow agrees qualitatively
with collinear wave–current experiments over smooth beds, e.g.
Kemp & Simons (1982, 1983). Musumeci et al. (2006) conducted an or-
thogonal wave–current experiment over a fixed bed with relatively
small roughness (d50 = 0.24 mm) and also observed a similar increase
in the near-bottom mean flow upon superposition of waves.
In addition to the changes in mean velocity, the superposition of
waves also modifies the direction of mean flow, θR, over depth. Initially
aligned at 90° for current-alone, θR increases by almost 10° after the in-
troduction of waves (Fig. 7b). For a depth-averaged nominal current ve-
locity of 11 cm/s, θR is expected to veer by about 9° from its initial 90°
orientation based on Stokes theory (Eq. (9)), which is in excellent
agreement with our experimental observation. The mean velocities in
the wave direction, u, for current-alone, wave-alone and combined
wave–current flows are presented in Fig. 8. When both current and
waves are present simultaneously, the vertical distribution of mass
transport flow changes drastically from a near-linearly increasing veloc-
ity profile (for wave-alone) to a near-constant velocity distribution (for
wave–current flow). This observation agrees qualitatively with the re-
sults obtained by Musumeci et al. (2006) over fixed roughness elements
of diameter 0.24 mm and 30 mm, and their study with Rew ~ 1,000–
12,000 showed that changes in u are more significant for small rough-
ness configuration (Abm/kn ~ 20–90) than large roughness configuration
(Abm/kn ~ 0.2–0.7). It appears from Fig. 8 that the u velocity distribution
for wave–current flow can be simply modeled by a uniform return cur-
 ES given by Eq. (9), instead of by the more complex LH53 solution,
rent, u
except in the near-bottom region where a boundary layer is developing.
While this may be a result of the enhanced mixing within the potential
flow region for the waves by the external turbulent current, a more
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
Fig. 8. Mean velocity profiles in the wave direction, u, for current-alone, wave–alone and
combined wave–current flows over a smooth bed. Dashed line denotes the Stokes
return current (− 1.8 cm/s). Full circle: current-alone; cross: wave-alone; hollow
square: wave–current.
likely explanation for the drastic change in the u-velocity distribution
over depth from near-linear for wave-alone to near-constant for com-
bined waves and current is obtained by considering the mass transport
current established in the down-wave region from the nominal current
channel. Since only waves are present here, the mass transport velocity
profile would be near-linear. However, when this pure wave-induced
mean flow encounters the external current of ~11 cm/s upon entering
the nominal current channel, the established near-linear mass transport
current will be advected in the direction of the combined mean flow
which is approximately at an angle of θR = 100°. Thus, even the water
column entering the nominal current channel at y = 0 would not
reach a location further than ytan(10°) into the nominal current chan-
nel. With the location of measurement stations at y ~ 4.3 m this distance
would be approximately 0.75 m and the characteristic, i.e. the near-
linear mass transport current, carried by this water column would
never reach the measurement locations, which are at least 1 m into
the nominal current channel. This simple conceptual model suggests
that water columns passing through the measurement region originate
from the current inlet and should exhibit the features of a developing
boundary layer originating from y = 0. This conclusion is supported
by the mean velocity profiles shown in Figs. 7a and 8 which both sug-
gest a boundary layer thickness of about 10 cm, in excellent agreement
with the value obtained from Eq. (1).
3.4. Analysis of smooth bed experiment
Since the wave Reynolds number, Rew, over a smooth bed is approx-
imately 7,800 in the present study, the wave boundary layer is laminar
when waves are alone. The laminar wave boundary layer thickness, δw,
is given by Sleath (1987) as follows:
rffiffiffiffiffiffi
2υ
δw ¼ 3
ω
which gives δw = 1.8 mm for a 1.4 s wave period. Meanwhile, for
smooth turbulent current flow, the viscous sub-layer of the current is
approximated by:
υ it appears that the prediction obtained with Eq. (2) for rough beds is less
δvc ¼ 11:6
uc
which gives δvc = 1.9 mm for the experimentally determined value of
u⁎c = 0.52 cm/s. As δw ≈ δvc for the present experimental conditions,
the laminar wave boundary layer is embedded within the viscous sub-
layer of the current flow. This suggests the interaction between the
two flows is purely linear, which is further supported by similar mean
shear velocities obtained for current-alone (u⁎c = 0.52 cm/s ± 9.1%)
265
and combined wave–current flow (u⁎c = 0.49 cm/s ± 7.9%). Without
nonlinear enhancement of shear stress, the maximum combined
wave–current shear velocity, u⁎m, can be computed directly using the
results from current-alone and wave-alone:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2 2
um 2 ¼ ðuwm 2 þ uc 2 j cosϕcw jÞ þ ðuc 2 sinϕcw Þ ð12Þ
where u⁎wm is the maximum wave-alone shear velocity and ϕcw is the
angle of wave–current flow, approximately 100° in this case. The max-
imum bottom shear stress for laminar waves can be obtained with the
following expression (Schlichting, 1968):
pffiffiffiffiffiffiffi 1=2
uwm ¼ υωubm ð13Þ
which gives u⁎wm =1.77 cm/s for the present experimental conditions.
To estimate the smooth turbulent roughness for a mean flow in the
presence of waves, which is denoted by kn,cw, the maximum combined
wave–current shear velocity, u⁎m, should be adopted as the input shear
velocity in Eq. (7) instead of current shear velocity, u⁎c:
υ
kn;cw ¼ 3:3 : ð14Þ
um
The bottom roughness is expected to decrease by a factor u⁎m/u⁎c of
3.4 from current-alone to wave–current condition, based on Eqs.
(7) and (14). The experimental bottom roughness was found to be
0.059 cm (error factor of 2.0) for current-alone, and for wave–current
flows a value of 0.015 cm (error factor of 2.2) was obtained. This trans-
lates into a reduction factor of 3.9 and agrees closely with the theoretical
factor. The relatively large error factor is not surprising given the fact
that the value of kn , cw is small in this case and any minor changes in
slope of the log-profile fit will result in huge variations on the log-axis.
The overall result suggests that the use of Eq. (14), with u⁎m replacing
u⁎c, is fairly accurate in predicting the bottom roughness when laminar
waves are present simultaneously with an orthogonal current over a
smooth bed, provided δw b δvc. Similar conclusion was also obtained by
Yuan & Madsen (2014) on the validity of Eq. (7) for collinear wave–cur-
rent flows, though their oscillating water tunnel experiment (with
Rew N 400,000 and wave boundary layer certainly in smooth turbulent
regime) suggests that the average wave–current shear velocity should
be used as the scaling shear velocity in Eq. (7) for their flow conditions.
The above analysis should be a step forward in providing a means to
resolve wave–current interaction over smooth beds. However, this type
of flow condition is relatively rare in the field. In order to achieve a more
realistic case of rough turbulent wave–current interaction, the bottom
roughness was increased by covering the smooth bed with a layer of ce-
ramic marbles.
4. Orthogonal (90°) wave–current flow over uniform roughness bed
4.1. Current-alone
ð10Þ
For uniform roughness experiments, the bed was covered with a
uniform, single layer ceramic marbles with d50 = 12.5 mm throughout
the entire basin (Fig. 9). An example of the velocity profile plotted in
semi-log scale is shown in Fig. 10. While Eq. (1) is shown to be accurate
in estimating the boundary layer thickness for flows over smooth beds,
ð11Þ
satisfactory. The boundary layer thickness, obtained from Eq. (2), is ap-
proximately 24 cm, while Fig. 10 shows a thickness of less than 15 cm.
Hence, Eq. (2) over-predicts the boundary layer thickness for uniform
rough beds by more than 60%. Despite this inaccuracy, the results
from log-profile analysis are unlikely to be compromised since the anal-
ysis utilizes measurement points within 1/3 of the boundary layer thick-
ness predicted with Eq. (2). From Fig. 10, it can be seen that this region is
well-described by the log-law.
266 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
Fig. 9. Uniform, single layer ceramic marbles, held in place by 2 m × 2 m grids formed
using 4 mm-diameter aluminum rods (inset: a close-up view of the marbles).
From the analysis, current-alone kn was determined to be 3.0 cm
with an error factor of 1.5, and u⁎c = 0.92 cm/s ± 9.1%. The flow is
fully rough turbulent since Re⁎ = knu⁎c/υ = 345, and the value of kn is ap-
proximately 2.4d50. A comparison was made with a 2D flume experi-
ment using the same marbles as roughness elements. The water depth
in that experiment was 0.4 m. Two discharges were generated and the
theoretical depth-averaged velocities were 10 cm/s and 16.7 cm/s, re-
spectively. Velocity profiles were measured at a distance of 7.5 m from
the inlet, and the values of kn obtained were 3.0 cm ( U  = 10 cm/s)
and 2.7 cm (U  = 16.7 cm/s), with error factors of 1.2 for both cases.
These results are in good agreement with the current-alone roughness
obtained in the wave–current basin. Furthermore, the current-alone kn
also agrees with the wave-alone kn obtained using a wave flume located
in the same laboratory (Lim et al. 2012). In the wave experiment, waves
of 1.6 s period and 10 cm height were generated over smooth and
marble-covered beds, and kn was derived based on the energy dissipa-
tion method. The analysis shows that wave-alone kn over the marble-
covered bed is 2.7 cm (error factor of 1.3), supporting the conjecture
that a single roughness length scale can be used for both currents and
waves over uniform 3D bed roughness.
4.2. Wave-alone
Similar to the current-alone flow, the wave boundary layer is in the
fully rough turbulent regime with Rew ~ 10,000 and Abm/kn ~ 1.5
(Jonsson, 1966; Davies & Villaret, 1997). Results for case WU2 is pre-
sented in Fig. 11, as measurement data (over depth) for case WU1 are
not available. With the increase in bottom roughness, the near-bed
streaming flow is smaller than the streaming velocity observed over a
Fig. 10. An example of current velocity profile over a uniform roughness bed in the 90°
current channel. Points located between the two dashed lines are used for log-profile
fitting.
Fig. 11. Wave-induced mass transport velocity over a uniform roughness bed (WU2).
Error bars of the experimental data represent one standard deviation. Error bars for the
theoretical solution are associated with ±11.2% of u. Cross: experimental data; full line:
LH53 model.
smooth bed, i.e. about 30% of the value predicted by the LH53 solution.
The reduction in streaming velocity with increasing bottom roughness
is in qualitative agreement with the theoretical analysis of Trowbridge
and Madsen (1984), who predicted this trend based on a time-varying
eddy viscosity model. Davies & Villaret (1999) also predicted similar
trend for flows over very rough and rippled bottom (Abm/kn b 1.5). Not-
withstanding the difference in the near-bed streaming velocity, the var-
iation of mean flow with elevation is similar for both smooth and rough
bed experiments, i.e. a near-linearly increasing velocity in the offshore
direction as the water surface is approached. This trend was also ob-
served in a separate flume experiment with the same marble bed for
kh = 0.57–2.10 (Lim, 2013). As discussed in Section 3.2, it appears nec-
essary to incorporate the near-surface flow characteristic observed by
Klopman (1994) in modeling the wave-induced mass transport flows
to achieve a better agreement with experimental observations.
4.3. Wave–current interaction
In contrast to the smooth bed experiment, a reduction in near-
bottom mean velocity was observed over the uniform roughness bed
upon superposition of waves (Fig. 12a). The decrease, which is attribut-
ed to the enhanced turbulence within the wave boundary layer, is in
qualitative agreement with existing nonlinear wave–current boundary
layer models (e.g. the Grant-Madsen (GM) model), and has been ob-
served in numerous collinear (e.g. Kemp & Simons, 1982; Mathisen &
Madsen, 1996b; Fredsøe et al., 1999) and orthogonal (e.g. Musumeci
et al., 2006; Madsen et al., 2008; Fernando et al., 2011) wave–current
experiments. The veering of θR is slightly larger for the rough bed exper-
iment - up to 14° from the current-alone direction (Fig. 12b). As in the
smooth bed experiment, such veering can be attributed to Stokes' re-
turn current. The veering in the near-bottom region (z b 10 cm) is asso-
ciated with turbulence asymmetry and mean momentum transfer
(DSK88 and AHM15) and will be discussed further in Section 4.4.
The mean velocity profile in the wave direction, u, was found to be
similar in trend to the smooth bed experiment, where the near-
linearly increasing profile for wave-alone is transformed into a near-
homogenous velocity distribution upon the addition of an external cur-
rent (Fig. 13). As discussed in Section 3.3, the mass transport flow ob-
served in the measurement region is expected to have developed from
the current inlet, since water columns that enter from outside the cur-
rent channel would propagate at a resultant flow direction, θR ~ 105°
and would miss the near-uniform flow region. Again, this is supported
by the observed boundary layer thickness for the mass transport flow,
which is approximately the same as the external current flow which
originates from the current inlet.
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
Fig. 12. Mean velocity, uR, and angle of mean flow, θR, for current-alone and 90° wave–current flows (WCU1) over a uniform roughness bed. Full circle: current-alone; hollow square:
wave–current.
The observed reduction in the near-bottom mean velocity can be ex-
plained by the presence of two layers of log profiles in combined wave–
current flows, one within the wave boundary layer and another outside
the wave boundary layer but within the current boundary layer. Since
the wave boundary layer is in the range of 1–2 cm in this study and
smaller than 2d50, the use of 2d50 as the lower limit in log-profile anal-
ysis is valid. The log-profile fitting to the data, which are outside the
wave boundary layer, is expected to yield the apparent roughness, kna.
On the other hand, the physical bottom roughness, kn , cw, is required
for comparison with current-alone kn as a means of model validation.
To obtain kn,cw, the following procedure was adopted. First, kna and u⁎c
were obtained from log-profile fitting of the measured velocity data.
An iteration was then performed to obtain kn , cw which, when used as
input to the GM model, yielded a kna value that matched the experimen-
tal kna. The details of the computational procedures are elaborated in
Appendix A. In this study, kna was found to be 7.1 cm (error factor of
1.3) and 5.5 cm (error factor of 1.8) for WCU1 (T = 1.4 s) and WCU2
(T = 1.6 s), respectively. The analysis shows that kna predicted by the
GM model matches the experimental kna when kn , cw = 1.7 cm for
WCU1 (error factor of 1.4) and kn , cw = 1.0 cm for WCU2 (error factor
of 2.3). Both results are considerably smaller than the current-alone
and wave-alone bottom roughness over the same bed configuration,
which are in the range of 2.7–3.0 cm. The amount of under-prediction
Fig. 13. Comparison of current-alone, wave-alone and 90° wave–current (WCU1) mean
flows along the wave direction, u, over a uniform roughness bed. Full circle: current-
alone; cross: wave-alone; hollow square: wave–current. Dashed line denotes the Stokes'
return current.
267
of kn ,cw by the GM model is approximately 45% for WCU1 and 67% for
WCU2.
At least part of the observed mismatch between kn , cw and current-
alone kn may be attributed to the insensitivity of the GM model to ϕcw,
in particular when the waves are strongly dominating the currents. In
the GM model, the formulation for the combined wave–current shear
stress is expressed as follows (refer to Appendix A for details):

τm ¼ C μ τwm ¼ ð1 þ μ Þτwm for ϕcw ¼ 0 ð15Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

τm ¼ C μ τwm ¼ ð1 þ μ 2 Þτwm for ϕcw ¼ 90 ð16Þ
where μ is the ratio of the current shear stress to the maximum wave
shear stress (τc/τwm). For strong wave conditions where μ approaches
zero, τm is virtually the same as τwm, indicating that the result is insen-
sitive to the angle of wave–current interaction. Additional evidence
supporting the above argument of a “directional-insensitive” GM
model comes from the numerical simulation of DSK88, where ubm =
100 cm/s, kn , cw = kn = 15 cm, Rew = 1.6 × 106, Abm/kn = 8.5 and
u⁎c = 5.8 cm/s. For these specifications and ϕcw = 0, the GM model pre-
dicts an apparent roughness kna = 107 cm which is in good agreement
with the value obtained from DSK88's Fig. 12 (~115 cm). Since the esti-
mation of DSK88's kna was done graphically, an uncertainty of ±5 cm is
anticipated. When ϕcw = π/2 is chosen, the GM model predicts kna =
92 cm which is only slightly smaller than the value for ϕcw = 0 and con-
siderably larger than kna obtained from DSK88's Fig. 12 (~63 cm). These
results support the conclusion that the GM model yields acceptable re-
sults for collinear flows, but over-estimates the wave effect on currents
for orthogonal flows.
To improve the original GM model's directional sensitivity, consider
the case of orthogonal wave–current interaction for which the bottom
shear stress, τy, in the current direction is given by:
τy =ρ ¼ −v0 w0 ð17Þ
where v′ and w′ denote the turbulent velocity fluctuations in the y (cur-
rent) and z directions, respectively. Since the waves are in the x-
direction their contribution to turbulent fluctuations in the spanwise di-
rection (y), v0 w , is expected to be smaller than in the streamwise direc-
tion (x), u0 w , i.e.
v0 w ¼ γ u0 w ð18Þ
268 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
where the reduction factor, γ, may be roughly estimated from the tur-
bulence intensity measurements by Klebanoff (1955) as presented in
Schlichting (1968, Fig. 18.5), to be:
!0:5
v0 w 2
γ ¼ 0:76 ð19Þ
u0 w 2
Since u0 w is scaled by u⁎wm = (fw/2)0.5 ubm, it follows from Eq. (18) that
orthogonal wave–current interaction may be considered approximately
equivalent to the collinear interaction of a current and an “imaginary”
wave specified by its near-bottom velocity, vbm = γubm.
When the DSK88 orthogonal wave–current case is analyzed in the
manner suggested above, the GM model predicts an apparent rough-
ness of 64 cm, in close agreement with DSK88's value of 63 cm. Applying
this directional modification of the original GM model to the present ex-
perimental results, a value of kn , cw = 2.5 cm was obtained for case
WCU1, i.e. cutting the difference between kn , cw and kn for current
alone more than in half (18% versus 45%). However, the agreement ob-
tained for case WCU2 is still unsatisfactory, although it reduces the dif-
ference between kn,cw and kn from 67% to 49%.
While the modification of the original GM model has improved its
directional sensitivity, it should be recognized that the suggested mod-
ification is limited to orthogonal wave–current flows. Thus, additional
theoretical developments are called for to generalize the directional
modification of the GM model to arbitrary angles between waves and
currents. In this respect, it is interesting to note that the reduction factor
γ = 0.76, here obtained by considering turbulence anisotropy which is
not included in neither DSK88's nor AHM15's numerical models, is near-
ly equivalent to assuming that the eddy viscosity affecting the orthogo-
nal current is scaled by the average (u⁎ave) rather than the maximum
wave shear velocity (u⁎wm), which would result in a reduction factor
of (2/π)0.5 = 0.80. Thus, to preserve the GM model's ability to predict
collinear wave–current interaction, it may be necessary to consider tur-
bulence anisotropy in the development of an improved theoretical
model for wave–current interaction at arbitrary angles.
4.4. Veering of near-bottom mean flow
As shown in the previous sections, the observed mean flows in
wave–current cases exhibit a tendency to veer towards the wavemaker
(increase of θR) by 10°–15° over depth, which is consistent with the su-
perposition of a wave-induced return current on the nominal current
flow. However, the veering of near-bottom mean flow appears different
from that in the upper column, which may be attributed to other factors
such as turbulence asymmetry and momentum transfer resulting from
the presence of currents along the wave axis (DSK88 and AHM15). For
co-directional waves and currents, the wave velocity in the half-cycle
following the current is enhanced compared to the half-cycle opposing
the current. This induces an asymmetry in turbulence intensity and gen-
erates a mean stress that drives a current opposite the direction of wave
advance. For waves and currents interacting at oblique angles, this leads
to a veering of near-bottom mean flow towards the axis perpendicular
to the waves (e.g. Fig. 13 of DSK88). The veering trend gradually dissi-
pates with distance from the bed but remains detectable along the
water column even if one is unable to measure within the wave bound-
ary layer.
The above mechanism should clearly manifest for the 60° and 120°
flow cases of the present study (see Section 5). For the 90° case, the
mean flow is generally shifted by 10°-15° (θR ~ 100°–105°) due to a
wave-induced return current, and the near-bottom flow should exhibit
a tendency to veer towards 90° as the bed is approached. However, an
opposite trend is observed in Fig. 12b. With further analysis of the mea-
surement data, it was found that disturbances generated at the inlet and
outlet openings (referred to as “parasitic waves”) may have contributed
to the observed veering behavior.
4.4.1. Parasitic waves
When waves propagate across the current inlet, they induce a differ-
ence in water level (pressure) between the basin and reservoir sides of
the honeycomb filters. The pressure difference fluctuates with wave pe-
riod and hence generates an oscillatory flow with the same frequency as
incident waves through the honeycomb filters. Similar oscillatory flow
is also expected to be generated at the current outlet. Despite the pres-
ence of the honeycomb filters, the nature of these wave disturbances
would be akin to the diffracted waves transmitted into the basin (the
shadow zone) by a train of waves propagating along the outside of the
basin walls encountering an open gap of length equal to the width
B = 2.5 m of the inlet or outlet openings. These diffracted waves
would emanate from the openings as approximately radial waves
with amplitudes that vary along the crests. For a gap width of B =
2.5 m, which happens to coincide with the length, L, of the 1.4 s
waves, the spatial wave amplitude variation, i.e. the diffraction coeffi-
cient pattern for B/L = 1, can be found in Johnson (1953). It shows
that the wave amplitude decreases with distance from the opening
and is larger in the down-wave than in the up-wave direction. Given,
with reference to Fig. 3, the rather large distance from the current
inlet (4.3 m) and outlet (3.5 m) the radial parasitic waves would be
nearly plane progressive waves of constant amplitude as they travel
across the small area (0.5 m by 0.8 m) of the measurement region, i.e.
the parasitic waves would show up as a partially standing wave in the
current channel direction that would introduce a non-zero wave veloc-
ity in the y-direction, ve = vm cos (ωt), where vm would be zero in the ab-
sence of parasitic waves. Further evidence of the existence of the type of
parasitic waves identified in the present experimental set-up and their
approximate description by the diffraction analogy can be found in
Fig. 16a of Musumeci et al. (2006), which shows parasitic wave velocity
amplitudes to decrease from vm ~ 30% to ~ 10% of um when increasing
the distance from the outlet from 75 cm to 200 cm.
For notation purpose, the parasitic waves generated at the inlet and
outlet are denoted by vi and vo, respectively. A subscript “bm” is used to
denote the near-bottom velocity amplitudes of the parasitic waves, i.e.
vi , bm and vo , bm. As anticipated from the diffraction analogy described
above, analysis of the pattern of experimentally obtained values of vbm
within the measurement region does indeed reflect the pattern associ-
ated with a partially standing wave, and resolving this pattern into its
wave components gives a slightly larger vo , bm ~ 1.85 cm/s than
vi ,bm ~ 1.12 cm/s as anticipated from the diffraction analogy. Thus, at a
velocity antinode the parasitic wave velocity amplitude is estimated to
be vbm = vi ,bm + vo,bm ~ 3.0 cm/s, i.e. roughly 20% of the velocity ampli-
tude of the “incident” waves propagating in the x-direction, suggesting
that the parasitic waves are a non-negligible source of disturbance in
the wave basin.
Fig. 14. Changes in the effective wave direction from x to x'', due to parasitic waves being
(a) in-phase and (b) out-of-phase with the incident wave motion. Correspondingly, the
axis perpendicular to the wave motion changes from y to y' '.
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
Fig. 15. Angle of mean flow, θR, for 90° wave–current flow over a uniform roughness bed,
with dashed lines indicating the variability across the measurement region.
When the parasitic waves interact with incident waves, the resultant
wave motion would deviate from the x-direction. For instance, when in-
cident and parasitic waves are in-phase, the effective wave direction
would rotate anti-clockwise from the x-axis (Fig. 14a). On the other
hand, when incident and parasitic waves are out-of-phase, the effective
wave direction would rotate clockwise from the x-axis (Fig. 14b). Corre-
spondingly, the axis normal to the wave motion would veer by the same
amount from the original y-axis. Based on the measured strength of ubm
and vbm, the veering is expected to be as large as 10°. Assuming a perfect
90° mean flow (without the presence of wave-induced mass transport
flow), the near-bottom mean flow direction may veer towards 100° at
locations where incident and parasitic waves are in-phase, and towards
80° when incident and parasitic waves are out-of-phase.
Given the continuous change in phase difference between incident
and parasitic waves throughout the current channel, the effective
wave direction (and the axis perpendicular to it) is expected to be
location-dependent. Hence, it is not surprising to observe a considerable
variation of flow veering across the measurement region (Fig. 15). How-
ever, at some local measurement stations, the veering of flow appears to
correlate with the effective wave direction. As shown in Fig. 16a, when
both incident and parasitic waves are nearly in-phase (effective wave
direction N 0°), the near-bottom mean flow appears to veer towards
110°. In contrast, when incident and parasitic waves are nearly out-of-
phase (effective wave direction b 0°), there seems to be a tendency for
Fig. 16. Local angle of mean flow over a uniform roughness bed at (a) location where incident and parasitic waves are nearly in-phase (b) location where incident and parasitic waves are
nearly out-of-phase.
269
the near-bottom mean flow to veer towards 90° (Fig. 16b). This is in
qualitative agreement with the conjecture made above, that the near-
bottom veering of mean flow may be locally influenced by the potential
±10° variation of effective wave direction arising from the presence of
parasitic waves.
It should be noted that the conditions leading to the generation of
parasitic waves are inherent in experiments where currents are intro-
duced and removed through sidewall openings. Since these distur-
bances tend to dissipate with distance from the sidewalls, their effects
are likely to be negligible only if the basin width is significantly greater
than the parasitic wave length, a condition clearly not satisfied in the
present setup. Hence, future studies should take into account these dis-
turbances in experimental design and data analysis.
5. Wave–current interaction at 60° and 120° over uniform roughness
bed
5.1. Mean velocity profiles
When waves are superimposed on 60° and 120° current flows, the
mean velocity profiles, as in the 90° case, exhibit a logarithmic trend
in the near-bottom region (Fig. 17). The reduction of near-bottom
mean velocity after introduction of waves is also consistent with the
GM model. However, the mean velocity decreases substantially for the
60° current near the water surface, while it remains virtually unchanged
for the 120° current. The overall discharge that flowed through the
near-uniform flow region was smaller compared to the current-alone
experiment (despite no variation in flow discharge), which was likely
caused by an increase in the lateral spreading of flow when waves
were present.
Current-alone θR in the 60° and 120° channels are shown in Fig. 18. It
can be seen that the angles of mean flow in the near-uniform flow re-
gions are close to the desired flow directions, underlining the capability
of the present setup to ensure the directionality of the current inflow.
With the superposition of waves, θR increases by 10° to 20° which, as
discussed in Sections 3.3 and 4.3, can be attributed to Stokes' return
flow. The current component in the direction of wave propagation, u,
is presented in Fig. 19. For the 60° case, the strength of u decreases in
the near-surface region, while the contrary is true for the 120° case.
This shows that when a near-orthogonal current flow is resolved into
the direction of wave propagation, the near-surface flow characteristics
are similar to co-directional wave–current flows (Klopman, 1994), i.e.
reduction (increase) of the near-surface mean flow when the current
270 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274

Fig. 17. Mean velocity, uR, profiles for current-alone and combined wave–current flows at interaction angles of (a) 60° (b) 120° over a uniform roughness bed. Wave period, T = 1.4 s. Full
circle: current-alone; hollow square: wave–current.

is following (opposing) waves, though the flows in the present study,
unlike Klopman's study, are not fully developed.
Parasitic waves are also expected to be present within the measure-
ment regions for the 60° and 120° experiments. The origin and nature of
these parasitic waves are similar to those observed and discussed for the
90° case, i.e. over the limited extent of the measurement regions the
parasitic waves will be approximately plane waves propagating in the
direction of the nominal current channel, one coming from the inlet
and the other from the outlet, leading to a partially standing wave pat-
tern. Wave velocity measurements indeed verify this conjecture and
show that the parasitic wave coming from the inlet is larger (smaller)
than the one coming from the outlet for the 60° (120°) case, as expected
from the diffraction analogy presented in Section 4.4. The parasitic wave
velocities in the y-direction cause the effective direction of the local
wave motion to vary ±10°, just as was the case for the 90° experiments.
However, in contrast to the 90° case, for which the ±10° variation in ef-
fective wave direction potentially could reverse the veering direction,
the effect of parasitic waves is not as dramatic for the 60° and 120° ex-
periments since this local variation in effective wave direction is smaller
than the 30° angle between the current and the direction of the normal
to the incident waves. Indeed, both 60° and 120° flows show a tendency
to veer towards the y-axis (90°) as the bottom is approached (Fig. 18),
though the trend is somewhat weaker for the 120° case. A second factor
that could potentially modify the effective wave direction is the current-
induced wave refraction, but this was found to be insignificant in the
present study (i.e. change of less than 2°).
5.2. Contamination of log-profile analysis
The spatially-averaged kna, u⁎c and kn , cw for the 60° and 120° cases
are shown in Table 3. The values of kn,cw were computed in accordance
with the procedures described in Section 4.3 and Appendix A. For com-
pleteness, the results obtained for 90° wave–current flows are also pre-
sented. When the angle between waves and current deviates from 90°,
kna should intuitively be larger since the wave-induced turbulence is in-
creasingly aligned with the current flow direction. However, the exper-
imental outcome is different from this anticipation, with the 60° wave–
current interaction yielding the smallest kna (~3.5 cm) among the three
angle cases. When the bottom roughness is resolved using the GM
model, kn ,cw is only 0.5 cm for the 60° case, which is more than a factor
of two smaller than the value obtained for the 90° case. On the other
hand, kn , cw for the 120° case (1.4–1.7 cm) is larger than the 90° case,
though only marginally so for 1.4 s waves.
The above observation may be attributed to the contamination of
log-profile analysis by wave-induced mass transport flow. A conceptual
explanation of the contamination of the 60° current profile by the wave-

Fig. 18. Angles of mean flow, θR, for current-alone and combined wave–current flows at interaction angles of (a) 60° (b) 120° over a uniform roughness bed. Wave period, T = 1.4 s. Full
circle: current-alone; hollow square: wave–current.
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274 271

Fig. 19. Velocity component in the wave direction, u, for current-alone and combined wave–current flows at interaction angles of (a) 60° (b) 120° over a uniform roughness bed. T = 1.4 s.
Full circle: current-alone; hollow square: wave–current.

induced mass transport flow is presented in Fig. 20. First, a perfect pure For wave-alone flow, the near-bottom streaming is increasingly
current log-velocity profile with kn = 1.25 cm and u⁎c = 0.8 cm/s is offshore-directed as the bottom roughness increases, and this agrees
plotted in Fig. 20a. The component of the wave-induced mass transport qualitatively with Trowbridge & Madsen (1984). The mass transport
(here taken as the projection of the mass transport current obtained ex- flow also exhibits a near-linearly increasing velocity profile with eleva-
perimentally for orthogonal wave–current flows, i.e., similar to that tion for both smooth and rough bed experiments, and the agreement
shown in Fig. 13) in the direction of the external current is shown in with Longuet-Higgins (1953) solution deteriorates as the water surface
Fig. 20b. Combining these two current flows produces the velocity pro- is approached. Since experiments involving collinear wave–current
file depicted in Fig. 20a (denoted by hollow squares), showing a de- flows have shown that the near-surface mean flow is increased by op-
crease in the roughness inferred from the log-profile analysis relative posing waves (Klopman, 1994), incorporating this mechanism in a tur-
to the roughness obtained from the “non-contaminated” log profile. bulent wave-induced mass transport flow model appears needed in
Similar considerations of mean profile contamination by wave- order to improve our ability to predict this type of mean flow. The pres-
induced mass transport would lead to an overestimate of the bottom ent measurement may serve to validate such future models.
roughness for the 120° case, which also agrees with our findings With the addition of an orthogonal external turbulent current, the
(Table 3). This conjecture is supported by the results obtained by van wave-induced mass transport profile is transformed into a near-
Rijn & Havinga (1995), where the 60° case appears to have a larger uniform flow distribution. The resulting mass transport profile appears
near-bottom mean velocity and smaller apparent roughness compared to be more accurately described by Stokes' theory than the Longuet-
to the 120° case. Consequently, a new approach capable of isolating Higgins solution, except in the near-bed region where the boundary
the wave-induced mass transport from the external current profile, or layer effect prevails. Since the direction of mean flow is at 100° to the
a new boundary layer theory incorporating the effect of wave-induced x-axis, the mass transport profile observed in the near-uniform flow re-
mass transport on the wave–current mean flow, will need to be devel- gion is unlikely to originate from outside the channel, but instead is de-
oped before log-profile analysis of mean velocity profiles to resolve veloping from the inlet with a uniform flow distribution over depth.
the bottom roughness for near-orthogonal wave–current flows can be This is supported by the similar boundary layer thickness observed for
used with confidence. the mass transport flow profile and the external current flow that is
generated at the inlet, i.e. the mass transport current we observe has
6. Conclusions not reached full development over depth.
The superposition of waves on an orthogonal smooth turbulent cur-
A series of experiments on near-orthogonal combined wave–current rent increases the near-bottom mean velocity and reduces the bottom
flows have been performed at the Hydraulic Engineering Laboratory of roughness. The decrease in bottom roughness can be estimated with
NUS. Current-alone, wave-alone and combined wave–current flows the smooth turbulent roughness equation for steady flows, if the current
were generated over smooth and uniform marble-covered rough beds. shear velocity, u⁎c, is replaced by the maximum combined wave–current
Near-uniform flow regions were established in the current channels shear velocity, u⁎m, when the wave boundary layer is laminar and of a
for multiple measurements to be performed and to minimize uncertain- thickness, δw, smaller than the thickness of the viscous sublayer of the
ty associated with local bed irregularities. Velocity data were collected
over a duration of 3 min at each location to allow more than 100 wave
cycles to be measured during wave-alone and wave–current Table 3
kna, kn,cw and u⁎c for wave–current interaction over marble-covered bed
experiments.
The current-alone experiments show that the direction of mean Angle between waves and T kna Error u⁎c Error kn ,cw Error
flows are approximately the same as the desired flow directions created currents, ϕcw (deg) (s) (cm) (Factor) (cm/s) (±%) (cm) (Factor)
by the inlet and outlet alignments, i.e. at 60°, 90° and 120° to the direc-
1.4 3.61 1.42 0.91 10.83 0.48 1.74
tion (x) of the incident waves generated in the basin. The current expe- 60
1.6 3.51 1.80 0.99 11.44 0.47 2.23
riences a smooth turbulent roughness when it is generated over the 1.4 7.05 1.27 1.19 4.86 1.65 1.37
90
concrete bottom of the wave–current basin, while the bottom rough- 1.6 5.48 1.85 1.06 12.89 1.00 2.28
ness induced by a single layer of uniformly-arranged marbles is equiva- 1.4 7.06 1.56 1.10 5.54 1.68 1.67
120
1.6 8.09 1.40 1.08 17.17 1.43 1.78
lent to 2.4d50.
272 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
Fig. 20. (a) An example of a ‘perfect’ log-profile for an external current at 60° with the direction of the wave propagation (full circles) and the log-profile contaminated by mass transport
(hollow squares) (b) wave-induced mass transport profile projected onto the direction of the external current.
current, δvc, i.e. δw b δvc. For rough beds, the addition of waves to an or-
thogonal current leads to a reduction in the near-bottom mean velocity,
as predicted by the Grant–Madsen (GM) model. However, both the ex-
perimental results and the numerical simulation by Davies et al. (1988)
indicate that the GM model has a tendency to over-predict the wave
influence on near-orthogonal currents, in particular when the waves
are dominating the currents. A plausible explanation is that the wave
turbulence influencing the current flow is associated with the
streamwise turbulence (u') when waves are collinear with the cur-
rent, and spanwise turbulence (v') when waves are orthogonal to
pffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffi
the current. Since v0 2 ̅ is considerably smaller than u0 2 ̅, the use
of the maximum wave–current shear velocity for all angles of
wave–current interaction inevitably over-accounts for the wave ef-
fect on near-orthogonal currents, hence resulting in directional in-
sensitivity of the GM model. A proposed approach to correct this
weakness has been discussed and shown to improve the prediction
of the bottom roughness, but further studies are required to
strengthen the theoretical basis of this approach. In addition, other
wave–current models employing more advanced turbulence closure
methods may also be validated and improved based on the results
and insights obtained in the present study.
As shown by Davies et al. (1988) and Afzal et al. (2015), the mean
current flow for combined non-collinear waves and current should
veer towards the direction normal to the direction of the near-bottom
wave orbital velocity as the bottom is approached from above. However,
this wave direction, expected to coincide with the x-axis, was locally af-
fected by the presence of parasitic waves emanating from the current
inlet and outlet. Within the region of measurement these parasitic
waves were shown to form a partially standing wave in the direction
of the nominal current channel and therefore introduce unexpected
wave velocities in the y-direction. Resolving the parasitic wave charac-
teristics in our experiments, the resulting effective wave direction was
found to potentially vary by ± 10° around the x-axis. In the 60° and
120° flow experiments, the nominal angle between current and the nor-
mal to the incident waves is ±30° and the parasitic wave effect merely
changes the magnitude of the veering angle but not its direction, and
our observed veering is in qualitative agreement with Davies et al.'s
(1988) prediction. However, for the 90° case, the direction of near-
bottom mean velocity veering was found to be significantly affected
by the presence of parasitic waves and to depend on the local effective
wave direction. This finding highlights the potential importance of rec-
ognizing the existence of parasitic waves in experimental set-up
employing gravity fed current generation and consider their presence
in experimental design and data analysis.
Attempts to resolve the bottom roughness for the 60° and 120°
wave-current cases have also been made. However, the contamination
of the wave–current mean flow by the wave-induced mass transport
current may lead to errors in estimating the bottom roughness using
the log-profile analysis. Hence, it is not possible to obtain accurate
values of bottom roughness from the present 60° and 120° studies
until a new approach capable of isolating the wave-induced mass trans-
port from the external current, or a new boundary layer theory incorpo-
rating the effect of wave-induced mass transport on the wave–current
mean flow, is developed.
Acknowledgment
The authors gratefully acknowledge the financial support of the Na-
tional Research Foundation of Singapore (NRF) through the Singapore-
MIT Alliance for Research and Technology's (SMART) Center for Envi-
ronmental Sensing and Modeling (CENSAM) program. The authors
also thank Prof. Cheong Hin Fatt (NUS) for his helpful discussions and
especially his assistance with lab infrastructure issues, and Prof. Alan
Davies and two other reviewers for their insightful comments and sug-
gestions on the paper.
Appendix A
A two-layer time-invariant, linearly-varying eddy viscosity distribu-
tion was adopted by Grant & Madsen (GM) in their wave–current
boundary layer model (Grant & Madsen, 1979, 1986). Within the
wave boundary layer, the eddy viscosity is scaled by the maximum com-
bined wave–current shear velocity, u⁎m, whereas outside of it, the eddy
viscosity is solely characterized by the current shear velocity, u⁎c:
υcw ¼ κ ðum Þz f or z ≤δcw ðA:1Þ
υc ¼ κ ðuc Þz f or zNδcw ðA:2Þ
where δcw = combined wave–current boundary layer thickness, i.e. the
hypothetical elevation at which the two logarithmic velocity profiles
(one within and one outside the wave boundary layer) intersect. Com-
putational procedures of the GM model are shown as follows:
(1) Calculate the ratio of the current shear stress to the maximum
wave shear stress, μ:
τc
μ¼ ðA:3Þ
τwm
 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274 273

where τc is the, assumed known, current shear stress and τwm is the (6) Predict apparent roughness, kna.
maximum wave shear stress. Since τwm is unknown, the initial value
of μ is taken as zero (assuming τwm NN τc).
The velocity profiles in the GM model are expressed as follows:
(2) Obtain the enhancement factor, Cμ:
 
uc uc z
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u¼ ln f or z ≤δcw ðA:12Þ
um κ z0
τ m ¼ τwm 2 þ 2τc τwm j cosϕcw j þ τ c 2 ðA:4Þ
¼ C μ τwm  
uc z
u¼ ln f or zNδcw ðA:13Þ
κ z0a
where τm is the vector summation of the wave and current shear stress
aligned at an angle of ϕcw, and: where kna = 30zoa is the apparent roughness, and equating the above
two formulas at z = δcw yields:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Cμ ¼ 1 þ 2μ j cosϕcw j þ μ 2 ðA:5Þ  uc =um
z0
zoa ¼ δcw : ðA:14Þ
δcw
(3) Compute the combined wave–current friction factor, fcw
(Madsen, 1994):
References
"  #
C μ Abm −0:078 Afzal, M.S., Holmedal, L.E., Myrhaug, D., 2015. Three-dimensional streaming in the seabed
f cw ¼ C μ exp 7:02 −8:82 ðA:6Þ boundary layer beneath propagating waves with an angle of attack on the current.
kn J. Geophys. Res. Ocean, 120, 4370–4391.
Arnskov, M.M., Fredsøe, J., Sumer, B.M., 1993. Bed shear stress measurements over a
smooth bed in three-dimensional wave–current motion. Coast. Eng. 20, 277–316.
C μ Abm Barrantes, A.I., Madsen, O.S., 2000. Near-bottom flow and flow resistance for currents
for 0:02b kn
b102 obliquely incident to two-dimensional roughness elements. J. Geophys. Res. 105
(C11), 26,253–26,264.
"  # Carstens, M.R., Neilson, F.M., Altinbilek, H.D., 1969. Bed forms generated in the laboratory
C μ Abm −0:109 under an oscillatory flow: analytical and experimental study. US Army Corps of Engi-
f cw ¼ C μ exp 5:61 −7:30 ðA:7Þ neers, Technical Memorandum no. 28.
kn Chanson, H., 2008. Acoustic doppler velocimetry (ADV) in the field and in laboratory:
practical experiences. International Meeting on Measurements and Hydraulics of
Sewers IMMHS '08, Bouguenais, France, pp. 49–66.
C μ Abm Christofferson, J.B., Jonsson, I.G., 1985. Bed friction and dissipation in a combined current
for 102 b kn
b104 and wave motion. Ocean Eng. 12 (5), 387–423.
Coffey, F.C., Nielsen, P., 1986. The influence of waves on current profiles. Proc. 20th Inter-
national Conference on Coastal Engineering, pp. 82–96.
where Abm is the near-bottom particle excursion amplitude and kn = kn,cw
Davies, A.G., Villaret, C., 1997. Oscillatory flow over rippled beds: Boundary layer structure
is the equivalent Nikuradse sand grain roughness of the bottom. and wave-induced Eulerian drift. Chapter 6 in gravity waves in water of finite depth.
In: Hunt, J.N. (Ed.), Advances in Fluid Mechanics. Computational Mechanics Publica-
(4) Obtain the maximum wave shear stress, τwm: tions, pp. 215–254.
Davies, A.G., Villaret, C., 1999. Eulerian drift induced by progressive waves above rippled
and very rough beds. J. Geophys. Res. 104 (C1), 1465–1488.
1 Davies, A.G., Soulsby, R.L., King, H.L., 1988. A numerical model of the combined wave and
τwm ¼ f ρu 2 ðA:8Þ
2 cw bm current bottom boundary layer. J. Geophys. Res. 93 (C1), 491–508.
Fernando, P.C., Guo, J., Lin, P., 2011. Wave–current interaction at an angle 1: experiment.
J. Hydraul. Res. 49 (4), 424–436.
where ubm is the near-bottom wave orbital velocity amplitude and ρ is Fredsøe, J., 1984. Turbulent boundary layer in wave–current motion. J. Hydraul. Eng. ASCE
110, 1103–1120.
the fluid density. τwm is substituted back into (A.3) to obtain a new value
Fredsøe, J., Andersen, K.H., Sumer, B.M., 1999. Wave plus current over a ripple covered
of μ, and Eqs. (A.3) to (A.8) are iterated until the convergence of μ. bed. Coast. Eng. 38, 177–221.
Grant, W.D., Madsen, O.S., 1979. Combined wave and current interaction with a rough
(5) Compute the transition level for the two-layer eddy viscosity, bottom. J. Geophys. Res. 84 (C4), 1797–1808.
δcw: Grant, W.D., Madsen, O.S., 1986. The continental shelf bottom boundary layer. Annu. Rev.
Fluid Mech. 18, 265–305.
 P Grass, A.J., 1971. Structural features of turbulent flow over smooth and rough boundaries.
κum Cμ 1 J. Fluid Mech. 50 (2), 233–255.
δcw ¼ A ðA:9Þ
ω μ 16:3 George, W.K., 2007. Is there a universal log law for turbulent wall-bounded flows? Phil.
Trans. R. Soc. A 365, 789–806.
Havinga, F.J., 1992. Sediment concentration and sediment transport in case of irregular
where, non- breaking waves with a current. Part E. Technical Report. Delft University of
Technology, Netherlands.
"   # Holmedal, L.E., Johari, J., Myrhaug, D., 2013. The seabed boundary layer beneath waves
C μ Abm −0:071 opposing and following a current. Cont. Shelf Res. 65, 27–44.
A ¼ exp 2:96 −1:45 ðA:10Þ
kn Huang, Z., Mei, C.C., 2003. Effects of surface waves on a turbulent current over a smooth or
rough seabed. J. Fluid Mech. 497, 253–287.
pffiffiffiffiffiffi Jackson, P.S., 1981. On the displacement height in the logarithmic velocity profile. J. Fluid
Cμ Mech. 111, 15–25.
P¼ pffiffiffiffiffiffi pffiffiffi ðA:11Þ Johnson, J.W., 1953. Engineering aspects of diffraction and refraction. Trans. Am. Soc. Civ.
2 Cμ − μ
Eng. 118 (2556), 617–652.
Jonsson, I.G. 1966. Wave boundary layers and friction factors. Proc. 10th International
It should be noted that Eqs. (A.9) through (A.11) differ from the orig- Coastal Engineering Conference, ASCE, 1, 127–148.
Kemp, P.H., Simons, R.R., 1982. The interaction between waves and a turbulent current:
inal GM model formulation proposed by Grant & Madsen (1979, 1986).
waves propagating with the current. J. Fluid Mech. 116, 227–250.
The derivation of Eq (A.10) was given by Madsen & Salles (1998), while Kemp, P.H., Simons, R.R., 1983. The interaction between waves and a turbulent current:
the additional factor of (Cμ/μ)P/16.3 in Eq. (A.11), which is typically close waves propagating against the current. J. Fluid Mech. 130, 73–89.
to 1/3, is added to obtain an improved transitional level between the Klebanoff, P.S., 1955. Characteristics of turbulence in a boundary layer with zero pressure
gradient. NACA Rep. 1247.
two logarithmic velocity profiles derived from the eddy viscosity formu- Klopman, G., 1994. Vertical structure of the flow due to waves and currents. Delft Hydrau-
lation given in Eqs. (A.1) and (A.2), respectively. lics Progress Report H 840.30, Part II.
274 K.Y. Lim, O.S. Madsen / Coastal Engineering 116 (2016) 258–274
Lim, K.Y., 2013. An experimental study on mean flow characteristics for near-orthogonal
combined waves and currents. Ph.D. Thesis. Department of Civil & Environmental En-
gineering, National University of Singapore.
Lim, K.Y., Madsen, O.S., Cheong, H.F., 2012. Current characteristics in the presence of near-
orthogonal waves. Proc. 33rd International Conference on Coastal Engineering, San-
tander, Spain, Paper No.: Currents.41.
Lofquist, K.E.B., 1986. Drag on naturally rippled beds under oscillatory flows. Technical
Report No. MP-86-13, U.S. Army Corps of Engineers. Coastal Engineering Research
Center.
Longuet-Higgins, M.S., 1953. Mass transport in water waves. Philos. Trans. R. Soc. London,
Ser. A 245, 535–581.
Lundgren, H., 1972. Turbulent currents in the presence of waves. Proc. 13th International
Coastal Engineering Conference. ASCE, pp. 623–634.
Madsen, O.S., 1970. Waves generated by a piston-type wave maker. Proc. 12th Interna-
tional Conference on Coastal Engineering, ASCE 1, Washington, DC, pp. 589–607.
Madsen, O.S., 1994. Spectral wave–current bottom boundary layer flows. Proc. 24th Inter-
national Conference on Coastal Engineering. ASCE, pp. 384–398.
Madsen, O.S., 2009. Lecture notes. Ralph M. Parsons Laboratory, Massachusetts Institute of
Technology, Cambridge, M.A.
Madsen, O.S., Kularatne, S., Cheong, H.F., 2008. Experiments on bottom roughness experi-
enced by currents perpendicular to waves. Proc. 31st International Conference on
Coastal Engineering, ASCE. 1, pp. 845–853.
Madsen, O.S., Salles, P., 1998. Eddy viscosity models for wave boundary layers. Proc. 26th
International Conference on Coastal Engineering, ASCE 3, pp. 2615–2627.
Mathisen, P.P., Madsen, O.S., 1996a. Waves and currents over a fixed rippled bed: 1. Bot-
tom roughness experienced by waves in the presence and absence of currents.
J. Geophys. Res. 101 (C7), 16533–16542.
Mathisen, P.P., Madsen, O.S., 1996b. Waves and currents over a fixed rippled bed: 2. Bot-
tom and apparent roughness experienced by currents in the presence of waves.
J. Geophys. Res. 101 (C7), 16543–16550.
Musumeci, R.E., Cavallaro, L., Foti, E., Scandura, P., Blondeaux, P., 2006. Waves plus cur-
rents crossing at a right angle: experimental investigation. J. Geophys. Res. 111,
C07019. http://dx.doi.org/10.1029/2005JC002933.
Myrhaug, D., Slattelid, O.H., 1989. Combined wave and current boundary layer model for
fixed rough seabeds. Ocean Eng. 16 (2), 119–142.
Nadaoka, K., Kondoh, T., 1982. Laboratory measurements of velocity field structure in the
surf-zone by LDV. Coast. Eng. Jpn 25, 125–145.
Nielsen, P., 1992. Coastal Bottom Boundary Layers and Sediment Transport. World
Scientific, Singapore.
Olabarrieta, M., Medina, R., Castanedo, S., 2010. Effect of wave–current interaction on the
current profile. Coast. Eng. 57, 643–655.
Pham-Gia, T., Hung, T.L., 2001. The mean and median absolute deviations. Mathematical
and Computer Modeling 34 (7–8), 921–936.
Scandura, P., Foti, E., 2011. Measurements of wave-induced steady currents outside the
surf zone. J. Hydraul. Res. 49 (sup1), 64–71.
Schlichting, H., 1968. Boundary Layer Theory. sixth ed. McGraw-Hill, New York.
Sleath, J.F.A., 1987. Turbulent oscillatory flow over rough beds. J. Fluid Mech. 182,
369–409.
Simons, R.R., Grass, T.J., Mansour-Tehrani, M., 1992. Bottom shear stresses in the bound-
ary layer under waves and currents crossing at right angles. Proc. 23rd Int. Conf. on
Coastal Eng. 45, 1, pp. 604–617.
Smith, J.D., 1977. Modeling of sediment transport on continental shelves. The Sea 6. Wiley
Interscience, New York, pp. 539–577.
Soulsby, R.L., Hamm, L., Klopman, G., Myrhaug, D., Simons, R.R., Thomas, G.P., 1993.
Wave–current interaction within and outside the bottom boundary layer. Coast.
Eng. 21, 41–69.
Trowbridge, J.H., Madsen, O.S., 1984. Turbulent wave boundary layers: 2. Second-order
theory and mass transport. J. Geophys. Res. 89 (C5), 7999–8007.
Van Rijn, L.C., Havinga, F.J., 1995. Transport of fine sands by currents and waves II. J. Wa-
terway, Port, Coastal and Ocean Eng. 121 (2). ASCE, pp. 123–133
Yuan, J., Madsen, O.S., Chan, E.S., 2012. Experimental study of turbulent oscillatory bound-
ary layers in a new oscillatory water tunnel. Proc. 33rd International Conference on
Coastal Engineering, ASCE. Paper No.: Waves.24.
Yuan, J., Madsen, O.S., 2014. Experimental study of turbulent oscillatory boundary layers
in an oscillating water tunnel. Coast. Eng. 89, 63–84.