634

Ocean Wave effects on the daily

cycle in SST

Peter A.E.M. Janssen

Research Department

12 October 2010
Series: ECMWF Technical Memoranda

A full list of ECMWF Publications can be found on our web site under:
http://www.ecmwf.int/publications/
Contact: [email protected]

c Copyright 2010

European Centre for Medium-Range Weather Forecasts

Shinfield Park, Reading, RG2 9AX, England

Literary and scientific copyrights belong to ECMWF and are reserved in all countries. This publication is not
to be reprinted or translated in whole or in part without the written permission of the Director. Appropriate
non-commercial use will normally be granted under the condition that reference is made to ECMWF.

The information within this publication is given in good faith and considered to be true, but ECMWF accepts
no liability for error, omission and for loss or damage arising from its use.
Ocean Wave effects on the daily cycle in SST

Abstract

Ocean waves represent the interface between the ocean and the atmosphere, and, therefore, a wave model is
needed to compute not only the wave spectrum, but is also required to determine the processes at the air-sea
interface that govern the fluxes across the interface.
Here, starting from earlier results with the Turbulent Kinetic Energy Equation, a simple model is developed
that allows for the inclusion of wave dissipation effects, effects of Langmuir turbulence and buoyancy on
the simulation of the daily cycle in SST.

1 Introduction.

The work of Terray et al (1996) and Craig and Banner (1994) has highlighted the prominent role of breaking
waves and its contribution to the surface current. In the field it is customary to find considerable deviations from
the usual balance between production and dissipation of turbulent kinetic energy. These deviations are caused
by the energy flux produced by surface wave damping. When observed turbulent kinetic energy dissipation, ε ,
and depth z are scaled by parameters related to the wave field, an almost universal relation between dimension-
less dissipation and dimensionless depth is found. Here, dimensionless dissipation is given by ε HS /Φaw , with
HS the significant wave height and Φaw the energy flux from wind to waves, while the dimensionless depth is
given by z/HS .
The energy flux by surface wave damping is expected to affect the upper-ocean mixing up to a depth of the order
of the significant wave height. Transport to the deeper layers of the ocean is possible because work against the
shear in the Stokes drift generates Langmuir cells which have a penetration depth of the order of the inverse of
a typical wave number of the wave field.
In this paper I would like to develop a multi-layer model of turbulent mixing in the upper ocean that includes
effects of surface wave damping, Langmuir turbulence and stratification in addition to the usual shear produc-
tion and dissipation. The model is applied to the problem of the evolution of the diurnal cycle in SST, and it is
shown that, even for low wind speed, wave effects play an important role in determining the amplitude of the
diurnal cycle.
The programme of the paper is as follows. In §2 a brief discussion of the role of ocean waves in air-sea
interaction is given while it is shown how to obtain in a reliable way energy and momentum flux from the wave
field. §3 gives some of the details of the mixed layer model that is proposed to describe the mixing processes
in the upper ocean. The model consists of momentum equations and the heat equation. In the presence of
turbulence these equations are not closed and the level-2 12 Mellor-Yamada scheme is adopted to model the
eddy viscosity for heat and momentum. These eddy viscosities are then found to depend on the turbulent
kinetic energy (TKE) and hence the need for a TKE equation. In the present paper the TKE equation describes
the rate of change of turbulent kinetic energy due to processes such as shear production (including shear in the
Stokes drift), damping by buoyancy, vertical transport of pressure and TKE and dissipation of turbulence. It
presents an ideal context to model effects of wave dissipation and Langmuir turbulence on the mixing properties
of the upper ocean. In contrast to the Graig and Banner model effects of wave dissipation on mixing are taken
into account by following the fairly novel approach of explicitely modelling the vertical transport of pressure
in terms of the rate of change of the wave spectrum due to wave dissipation (A similar idea in the atmospheric
context was pursued by Janssen (1999)). The effect of Langmuir turbulence, following Grant and Belcher
(2009), is represented by the part in the shear production term that is connected to Stokes drift. The upper
ocean may experience extremely stable conditions, especially during the day under low wind speed conditions;
the modelling of these stable conditions therefore requires special attention. A model for buoyancy effects was
developed which for weakly stable conditions is based on results from the Kansas field campaign (assuming

Technical Memorandum No. 634 1

 Ocean Wave effects on the daily cycle in SST

that atmospheric and oceanic turbulence behaves in a similar fashion) while the modelling of extremely stable
conditions was guided by the renormalisation approach of Sukoriansky et al. (2005).
In §4 some properties of steady state solutions of the TKE equation are discussed. In particular, it is shown that
to a good approximation diffusion of turbulent kinetic energy may be neglected. This approximation is called
the local approximation because the turbulent kinetic energy then only depends on the local properties of the
turbulent flow. In the local approximation it turns out that the TKE equation reduces to an algebraic problem
and its solution indicates that the turbulent velocity (and hence the eddy viscosity) only weakly depends on the
wave energy flux and the contribution by Langmuir turbulence (according to a 1/3-power law). Nevertheless,
wave effects enhance the eddy viscosities by a factor of 2-3. Inspecting more closely the solution according to
the local approximation it is found that wave dissipation affects the mixing process very close to the surface
at a depth of the order of the significant wave height. Langmuir turbulence is found to affect mixing in the
deeper parts of the upper ocean at a depth of the order of a typical wavelength of the ocean wave field. Also
buoyancy effects are discussed in some detail. For weak stratification, the present model is shown to be in
close agreement with the results of the Kansas field campaign (Businger et al., 1971) while for extremely stable
conditions it is found that momentum transport dominates heat transport, in agreement with Sukoriansky et al.
(2005). In addition, the combined effects of waves and buoyancy are studied as well. It is found that under
stable conditions buoyancy effects, which act in particular in the deeper parts of the upper ocean, suppress
the effects of Langmuir turbulence. Finally, the TKE equation is shown to be in close agreement with the
empirically known dependence of dimensionless turbulent dissipation on depth.
In §5 results of numerical simulations with the mixed layer model are presented. First, a synthetic example with
constant momentum and heat fluxes is given, which is followed by a simulation of the sea surface temperature
(SST) at a location in the Arabian Sea. The simulated diurnal cycle in SST is found to be in close agreement
with in-situ observations. The importance of sea state effects, even for low wind speed cases, is shown as well.
Finally, §6 gives a summary of conclusions.

2 Surface layer mixing and ocean waves.

In order to be able to give a realistic representation of the mixing processes in the surface layer of the ocean,
it should be clear that a reliable estimate of energy and momentum fluxes to the ocean column is required.
A first attempt to estimate these fluxes from modelled wave spectra and knowledge about the generation and
dissipation of ocean waves was given by Komen (1987). Weber (1994) studied energy and momentum fluxes in
the context of a low-resolution coupled ocean-wave atmosphere model (WAM-ECHAM), and it was concluded
that there is no need to use a wave prediction model to determine the energy flux. A parametrization of the type
Φaw = mρa u3∗ (with u∗ the air friction velocity and m a constant) would suffice. It will be shown here that this
conclusion depends on an approximation used by Weber to estimate the energy flux.
As energy and momentum flux depend on the spectral shape, the solution of the energy balance equation is
required. It reads
∂ ∂
F + vg . F = Sin + Snl + Sdiss + Sbot , (1)
∂t ∂x
where F = F(ω , θ ) is the two-dimensional wave spectrum which gives the energy distribution of the ocean
waves over angular frequency ω and propagation direction θ . Furthermore, vg is the group velocity and on
the right hand side there are four source terms. The first one, Sin describes the generation of ocean waves
by wind and therefore represents the momentum and energy transfer from air to ocean waves. The third and
fourth term describe the dissipation of waves by processes such as white-capping, large scale breaking eddy-
induced damping and bottom friction, while the second terms denotes nonlinear transfer by resonant four-wave

2 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

interactions. The nonlinear transfer conserves total energy and momentum and is important in shaping the wave
spectrum and in the down-shift towards lower frequencies.
Let us first define the momentum and energy flux. The total wave momentum P depends on the variance
spectrum F(ω , θ ) and is defined as
Z 2π Z ∞
k
P = ρw g dω dθ F(ω , θ ), (2)
0 0 ω
which agrees with the well-known relation that wave momentum is simply wave energy divided by the phase
speed of the waves. The momentum fluxes to and from the wave field are given by the rate of change in time
of wave momentum, and one may distinguish different momentum fluxes depending on the different physical
processes. For example, making use of the energy balance equation (1) the wave-induced stress is given by
Z 2π Z ∞
k
τ aw = ρw g dω dθ Sin (ω , θ ), (3)
0 0 ω
while the dissipation stress is given by
Z 2π Z ∞
k
τ wo = ρw g dω dθ Sdiss (ω , θ ), (4)
0 0 ω
Similarly, the energy flux from wind to waves is defined by
Z 2π Z ∞
Φaw = ρw g d ω d θ Sin (ω , θ ), (5)
0 0

and the definition for the energy flux from waves to ocean, Φwo , follows immediately from the above one by
replacing Sin by Sdiss . It is important to note that while the momentum fluxes are mainly determined by the
high-frequency part of the wave spectrum, the energy flux is to a larger extent determined by the low-frequency
waves.
In an operational wave model, the prognostic frequency range is limited by practical considerations such as
restrictions on computation time, but also by the consideration that the high-frequency part of the dissipation
source function is not well-known. In the ECMWF version of the WAM model the prognostic range of the
wave spectrum is given by the condition

ω < ωc = max(2.5ωmean , 4ω pm ) (6)

where ωmean is a conveniently defined mean angular frequency and ω pm is the Pierson Moskovitch frequency.
In the diagnostic range, ω > ωc , the wave spectrum is given by Phillips’ ω −5 power law. In the diagnostic
range it is assumed that there is a balance between wind input, dissipation and nonlinear transfer. In practice
this means that all energy and momentum going into the high-frequency range of the spectrum is dissipated,
and is therefore directly transferred to the ocean column.
As a consequence, the momentum flux to the ocean, τ oc , is given by
Z 2π Z ωc
k
τ oc = τ a − ρw g dω dθ (Sin + Snl + Sdiss ) , (7)
0 0 ω
where τ a is the atmospheric stress, whose magnitude is given by τa = ρa u2∗ . Note that the ocean momentum
flux τ oc only involves the sum of the three source functions of the energy balance equation and therefore it only
involves the total rate of change of wave momentum. Any wave model that is forced by reliable atmospheric
stresses and that produces wave height results that compare well with, for example, buoy wave height data and
Altimeter wave height data, will produce reliable estimates of the ocean momentum flux τ oc .

Technical Memorandum No. 634 3

 Ocean Wave effects on the daily cycle in SST

tau_air
tau_oc
Phi_aw
3 Phi_oc
Flux

0
0 12 24 36 48
Time (hrs)

Figure 1: Evolution in time of normalized momentum flux and energy flux to the ocean for the case of a passing front after
24 hrs. The momentum flux has been normalized with ρa u2∗ , while the energy flux has been normalized with mρa u3∗ , where
m = 5.2.

Ignoring the direct energy flux from air to currents, because it is small (cf. Phillips, 1977), the energy flux to
the ocean, Φoc , is given by
Z 2π Z ωc
aw − ρw g
Φoc = Φtot d ω d θ (Sin + Snl + Sdiss ) , (8)
0 0

where Φtot
aw is the total energy flux transferred from air to ocean waves. This total energy flux is fairly well-
known, because empirically the wind input to ocean waves is well-known, even in the high-frequency part of the
spectrum (cf. Plant, 1982). Furthermore, there is now a consensus that the high-frequency part of the spectrum
obeys an ω −5 power law (Banner, 1990; Birch and Ewing, 1986; Hara and Karachintsev, 2003, to mention but
a few references). Hence, fairly reliable estimates of the energy flux Φoc may be provided by means of a wave
model provided the model has a wind input term that agrees with the observations of wave growth and provided
modelled wave heights compare well with observations.
Before results of time series for momentum and energy flux for a simple case are presented, we have to make
one remark on the numerical implementation of (4) and (5). The energy balance equation is solved by means
of an implicit integration scheme (cf. Komen et al, 1994). To be consistent with the numerical treatment of the
energy balance, the momentum and energy flux have to be treated in a similar spirit, i.e. including the implicit
factors of the integration scheme.
Let us now illustrate the sea-state dependence of the momentum and energy flux for the simple case of the
passage of a front. To that end we take a single grid-point version of the ECMWF version of the WAM model
and force the waves for the first day with a constant wind speed of 18 m/s, which is followed by a drop in wind
speed to 10 m/s and a change in wind direction by 90 deg. In Fig. 1 we have plotted time series of atmospheric
stress (τa ), the momentum flux to the ocean (τoc ), the total air-wave energy flux (Φtot
aw ) and the energy flux
into the ocean (Φoc ). The momentum fluxes have been normalized by τa , while the energy fluxes have been
normalized by mρa u3∗ , with m = 5.2 which is a convenient mean value. During the first day we deal with the
case of wind-generated gravity waves, hence windsea, and, in particular, the difference between atmospheric
stress and the momentum flux to the ocean is small, most of the time at best 2%. This is a well-known property
of windsea (JONSWAP, 1973). For windsea, the difference between total energy flux Φtot aw and the energy flux

4 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

ECMWF Monthly mean relative momentum flux (Tau/Ustar**2) for January 2003
1.020
1.015
1.010
1.005
1.000
0.995
0.990
0.985
0.97
0.97 0.980
0.98
0.99
0.975
0.970
0.965
0.9
7 0.960
0.955
0.950
0.945
0.940
0.935
0.930

Figure 2: Monthly mean of momentum flux into the ocean, normalized with the atmospheric stress. Period is January
2003.

into the ocean Φoc is somewhat larger. When the front passes at T = 24 hrs there is a sudden drop in wind, hence
in atmospheric stress. However, the waves are still steep and experience an excessive amount of dissipation in
such a way that wave energy decreases. As a consequence, considerable amounts of momentum and energy are
dumped in the ocean column, much larger than the amounts one would expect from the local wind. Therefore,
in cases of rapidly varying circumstances, the fluxes are seen to depend on the sea state. This is in particular
true for the energy flux Φoc and to a much lesser extent for the momentum flux τoc .
This different behaviour of momentum flux and energy flux is caused by a combination of two factors. By
definition momentum flux is mainly determined by the high frequency part of the spectrum while we have
assumed that in the unresolved part of the spectrum there is a balance between wind input and dissipation.
Hence, for windsea there is almost always a balance between atmospheric momentum flux and the flux into
the ocean. This holds to a lesser extent for the energy flux because this flux is partly determined by the low
frequency part of the wave spectrum as well.
The different behaviour of momentum and energy flux is also found in the monthly means on a global scale.
This is illustrated in the Figs. 2 and 3, which are taken from Janssen et al. (2004). The typical variation in the
ratio τoc /τa is then found to be of the order of 4% while the variation in the normalized energy flux, Φoc /mρa u3∗ ,
is substantially larger. The global average of the value for m turns out to be m ≃ 5.2. Note that the map for the
energy flux shows an interesting spatial pattern. In the equatorial region values of the normalized energy flux
are small, suggesting that the mixed layer is thinner than the norm. In the extra-Tropics the normalized energy
flux is considerably larger, presumably because here there is larger variability in the wind field.
We finally remark that in the work of Weber (1994) the energy flux into the ocean was approximated by the
relation Φoc ≃ hciτaw , where hci is the mean phase velocity. This generally overestimates the energy flux by at
least a factor of two and as a consequence fairly high values of m (m ≃ 14) are found. In addition, in interesting
cases such as the passage of a front, the energy flux approximated in this manner will follow the wind. For
example, in the frontal case of Fig. 1 the energy flux to the ocean would decrease dramatically at T = 24 hrs,
while, in fact, it should hardly change. Therefore, it is not surprising that with this approximation the energy

Technical Memorandum No. 634 5

 Ocean Wave effects on the daily cycle in SST

ECMWF Monthly mean relative energy flux (E/5.2Ustar**3) for January 2003
1.5
1.4
1.3
1.2
1.1
1
0.8
1 1
0.8 0.9
0.8
0.6 0.8
0.6
0.4 0.7
0.6
0.4
0.4
0.6
0.8
0.5
0.6
1
0.8 1
1 0.4
1
1
0.3
1
0.2
0.1
0
Figure 3: Monthly mean of energy flux into the ocean, normalized with mρa u3∗ where m ≃ 5.2. Period is January 2003.

flux Φoc and wind are closely related.

It is concluded that it is not a good idea to estimate the energy flux from the local stress, because significant
memory effects are present in rapidly varying circumstances. In general, when wave information is available,
it is preferred to directly use knowledge on the evolution of the sea state due to wave dissipation, c.f. Eq. (8).
Furthermore, on average 98% of the atmospheric stress is transferred locally to the ocean, while 2% of the wave
momentum is advected away and finally dissipates at the shores. However, under extreme circumstances such
as during hurricanes as much as 10% of the wave momentum may be advected away. Therefore, although on
average differences are small, it seems preferable to drive the ocean with the momentum flux from waves to
ocean, cf. Eq. (7), because the alternative choice would introduce slightly more momentum in the ocean colum,
which in long integrations may give a contribution to climate drift.

3 Mixed layer modelling.

Having found a reliable way of obtaining from the rate of change of the wave spectrum the momentum and
energy flux into the ocean, we now turn our attention to the consequences for the mean flow in the ocean. We
start from the work of Craig and Banner (1994) (and Mellor and Yamada (1982)) who introduced effects of wave
dissipation on turbulent mixing by specifying the energy flux at the surface as a surface boundary condition
to the turbulent kinetic energy (TKE) equation. Following Grant and Belcher (2009), the TKE equation is
extended by introducing the generation of Langmuir circulation through work done against the shear in the
Stokes drift. Furthermore, following Noh and Kim (1999) and Baas et al. (2008), the important effects of
buoyancy are introduced as well. We discuss the consequences for the momentum and heat equation, where
the eddy viscosity is expressed as a product of a mixing length and turbulent velocity which follows from the
solution of the turbulent kinetic energy budget.
The model is applied to the problem of the diurnal cycle in sea surface temperature (SST), which is quite a
challenge because the SST follows from a balance between absorption of solar radiation in water and turbulent

6 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

transport of heat. Assuming that the amplitude of the diurnal cycle can be measured accurately and since the
absorption profile of solar radiation is well-known, this application provides a sensitive test of our ideas of
mixing in the upper ocean. In this Section the model is presented, while in §4 the properties of the steady state
version of the momentum, heat and TKE equations are studied. This is then followed in §5 by applying the
dynamical model to a synthetic case of constant wind forcing and heat flux, while, using observed forcings, the
model is also applied to simulate the diurnal cycle in subsurface temperature for a three month period in the
Arabian Sea.

3.1 Momentum equation.

To simplify the problem, the wind/wave driven water velocity is assumed to be non-rotating and uniform with-
out any pressure gradients in the horizontal directions. For convenience increasing depth is taken in the positive
z-direction. The momentum equation then reduces to

∂u ∂τ
= − , τ = −hδ uδ wi, (9)
∂t ∂z
where τ is the stress in the water column which is usually parametrized as τ = −ν∂ u/∂ z, assuming that the
main component of the water velocity is turbulent. However, in the same spirit as done for the problem of wind
wave generation (Janssen, 1999; Janssen et al., 2004) it is suggested that in particular in the upper part of the
ocean column wave motion is an important component as well. Therefore, the fluctuating parts of the velocity
are written as a sum of wave-induced motion, denoted with a subscript w, and turbulent motion, denoted with
a prime ′ , and it is assumed that there is no correlation between wave motion and turbulence. As a result the
stress τ becomes

τ = −hδ uw δ ww i − hu′ w′ i,

and the turbulent part of the stress is modelled with a mixing length model while the wave-induced part is
given, i.e. independent of the current. The shape of the wave-induced stress is prescribed by a function whose
derivative vanishes at the surface1 , hence
τoc
− hδ uw δ ww i = × T̂ (z), 1 − T̂ (z) = (1 − e−z/z0 )2 ,
ρw
where z0 determines the gradient of the wave-induced stress and is considered to be closely related to the
significant wave height HS . In other words, it is assumed that wave dissipation affects at most a layer of
thickness of the wave height. Combining everything together and introducing the water friction velocity w∗
according to
τoc = ρw w2∗ (10)
the momentum equation becomes

∂u ∂ ∂u
 
d T̂ (z)
= νm − w2∗ . (11)
∂t ∂z ∂z dz

Here, νm is the eddy viscosity for momentum and following Craig and Banner (1994) the level-2 21 Mellor-
Yamada scheme is used (Mellor and Yamada, 1982). Hence, the eddy viscosity for momentum (and heat
denoted by νh ) is expressed as
νm,h = l(z)q(z)SM,H (12)
1 the reason for the vanishing of the first derivative will be explained shortly

Technical Memorandum No. 634 7

 Ocean Wave effects on the daily cycle in SST

where l(z) is the turbulent mixing length, e = q2 /2 is the turbulent kinetic energy (q(z) is referred to as the
turbulent velocity) and SM and SH are dimensionless parameters which may still depend on stratification. The
turbulent velocity q will be obtained from the TKE equation, while the expression for the mixing length will be
introduced during the discussion of buoyancy effects.
Eq. (11) is the basic evolution equation for the ocean current. In order to better understand the role of the wave-
induced stress profile it is of interest to study the case of a time-independent current. Then, the momentum
equation becomes
 
d du d T̂ (z)
νm − w2∗ = 0.
dz dz dz
Integrating once with respect to depth and realizing that the momentum flux to the water column is suplied
entirely by surface wave dissipation one finds
du
νm = −w2∗ 1 − T̂ (z) .


dz
and this equation can be immediately integrated for u with the result
Z z
1 − T̂ (z)
u = −w2∗ dz . (13)
H νm
The advantage of the introduction of the wave-induced profile is now immediately evident by closer inspection
of the relation (13). In the usual approach the function T̂ is absent and if one would choose an eddy viscosity
which is a linear function of depth or height a logarithmic singularity would occur upon evaluation of the
integral. This singularity can only be avoided by the introduction of a ’mysterious’ roughness length z0 . In
the present case such an ’ad-hoc’ measure is not needed. Now, the integrand has no singularity at the origin
because the function 1 − T̂ (z) vanishes sufficiently rapidly near the surface therefore cancelling the singularity
caused by the eddy-viscosity ν in the denominator. Therefore, in the following the eddy viscosity is assumed
to be given by Eq. (12), while the mixing length scale l(z) is assumed to vanish for z → 0.

Remark: One may apply a similar reasoning to the problem of air flow over wind-generated gravity waves. The wave-
induced stress is then determined by the wind-input source function, and the wind profile follows from
dU
νm = u2∗ 1 − T̂ (z) .


dz
The eddy-viscosity νm is again given by Eq. (12), i.e. νm = l(z)q(z)SM . The turbulent velocity q(z) is obtained from the
kinetic energy equation, and for simplicity I assume that this consists of a balance between production and dissipation.
This implies

dU 2
 
νm =ε
dz

where the dissipation ε = q3 /Bl. Making use of the expression for the eddy viscosity in the energy budget, the turbulent
velocity is readily found and, as a result the eddy viscosity becomes
dU
νm = l 2 (z)
dz
3/4
where the relation B1/4 SM = 1 is used. Substitution of the eddy viscosity in the momentum equation finally gives for the
wind profile
Z z
dz
1/2
U(z) = u∗ 1 − T̂ (z) ,
0 l(z)

8 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

for the boundary condition that the wind velocity vanishes at the surface. Now, taking as mixing length l(z) = κ z it is
immediately evident from the above expression that the wind velocity only remains finite provided 1 − T̂ (z) ∼ z2 for
vanishing height z. Therefore, choosing for the stress profile

1/2
1 − T̂ (z) = 1 − e−z/z0 ,
the expression for the wind profile U becomes
Z z
dz  
U(z) = u∗ /κ 1 − e−z/z0 .
0 z
The integral may be expressed in terms of the exponential integral E1 (z) (see Abramowitz and Stegun, 1964), hence,
u∗
U(z) = [log(z/z0 ) + γ + E1 (z/z0 )] ,
κ
where γ = 0.57721 is Euler’s constant. Expressions for E1 (z) for small and large z are known. The resulting form of the
wind profile for small z/z0 becomes
u∗ z
U(z) ≈ ,
κ z0
while for large z/z0 the wind profile becomes
u∗
U(z) ≈ log(z/y0 ), y0 = e−γ z0 .
κ
Remarkably, as e−γ ≈ 0.561, the outer flow experiences a smoother flow than the inner flow.
In summary, the roughness length may be explained in terms of a gradient length related to the wave-induced stress profile
T̂ (z). However, in order to obtain a finite surface velocity there are restrictions to the behaviour of the wave stress profile
near the surface, 1 − T̂ (z) ∼ z2 !

3.2 Heat equation.

The heat equation describes the evolution of the temperature T due to radiative forcing and turbulent diffusion.
Using the depth variable z, the temperature evolves according to
∂T 1 ∂R ∂ ∂T
=− + νh , (14)
∂t ρw cw ∂ z ∂ z ∂ z
where νh is the eddy viscosity for heat, given by Eq. (12), while the solar radiation profile R(z) is parametrized
following the work of Soloviev (1982), i.e.
R(z) = a1 exp(−z/z1 ) + a2 exp(−z/z2 ) + a3 exp(−z/z3 ) (15)
with
(a1 , a2 , a3 ) = (0.28, 0.27, 0.45)
while
(z1 , z2 , z3 ) = (0.013986, 0.357143, 14.28571).
The decay length scale z1 , corresponding to the absorption of light in the near UV range, is seen to be quite
small, of the order of 1 cm. Therefore, in order to capture the absorption of light in the near UV range high
resolution in z near the ocean surface is required.
The ’turbulent’ heat transport could have been modelled in a similar fashion as done for the momentum equation
by adding a explicit contribution due to the presence of growing water waves. So far this has not been done yet.

Technical Memorandum No. 634 9

 Ocean Wave effects on the daily cycle in SST

3.3 Kinetic energy equation.

The equation for the kinetic energy of the turbulent velocity fluctuations is obtained from the Navier-Stokes
equations. If effects of advection are ignored, the TKE equation describes the rate of change of turbulent
kinetic energy e due to processes such as shear production (including the shear in the Stokes drift), damping by
buoyancy, vertical transport of pressure and TKE, and turbulent dissipation ε . It reads
∂e ∂ US 1 ∂ ∂
= νm S2 + νm S − νh N 2 + (δ pδ w) + (eδ w) − ε , (16)
∂t ∂z ρw ∂ z ∂z
where e = q2 /2, with q the turbulent velocity, S = ∂ U /∂ z and N 2 = gρ0−1 ∂ ρ /∂ z, with N the Brunt-Väisälä
frequency, ρw is the water density, δ p and δ w are the pressure and vertical velocity fluctuations and the over-
bar denotes an average taken over a time scale that removes linear turbulent fluctuations
The turbulent production by Langmuir circulation is modelled following Grant and Belcher (2009) by the
second term on the right-hand side of Eq. (16) which represents works against the shear in the Stokes drift.
Here US is the magnitude of the Stokes drift for a general wave spectrum F(ω ),
Z ∞
2
US = d ω ω 3 F(ω )e−2k|z| , k = ω 2 /g.
g 0

Although in principle the depth dependence of the Stokes drift is known it still is a fairly elaborate expression
through the above integral. In the final result we will use the approximate expression

US = US (0)e−2kS z ,

where US (0) is the value of the Stokes drift at the surface and kS is an appropiately chosen wavenumber scale.
The dissipation term is taken to be proportional to the cube of the turbulent velocity divided by the mixing
length
q3
ε= , (17)
Bl
Here, B is another dimensionless constant.
It is customary (see e.g. Mellor and Yamada, 1982) to model the combined effects of the pressure term and the
vertical transport of TKE by means of a diffusion term. Thus,
1 ∂ ∂ ∂ ∂e
 
(δ pδ w) + (eδ w) = lqSq
ρw ∂ z ∂z ∂z ∂z
where Sq is a constant. As a result the TKE equation becomes

∂e ∂ ∂e ∂ US √ e3/2
 
= lqSq + νm S2 − w2∗ − νh N 2 − 2 2 ,
∂t ∂z ∂z ∂z Bl(z)
and this equation has to be supplemented by boundary conditions. Following Craig and Banner (1994) it is
often assumed that the energy flux at the surface is supplied by wave dissipation. Hence the boundary condition
at the surface becomes
∂e
− lqSq = F0 for z = 0,
∂z
while at infinite depth the gradient in TKE is assumed to vanish,
∂e
=0 for z → ∞.
∂z

10 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

The energy flux ρw F0 is related to the energy flux into the ocean by
ρw F0 = Φoc (18)
where Φoc is the energy flux by breaking and/or dissipating waves given by Eq. (8). In the absence of the
relevant information on the sea state, the energy flux is often parametrized as Φoc = mρa u3∗ . Hence writing,
F0 = α w3∗ , (19)
one then finds α = m (ρw /ρa )1/2 . With m in the range of 2 − 10, α has typical values of the order 50 − 250.
Using a wave prediction system m and α can be determined explicitely.
However, the pressure term can also be determined by explicitely modelling the energy transport caused by
wave dissipation. Janssen (1999) demonstrated how the pressure term may affect flow in the atmospheric
boundary layer by explicitely using knowledge on the growth of waves by wind. The same idea will be used
here (cf. Janssen et al. (2004)) but now applied to wave dissipation in the ocean column. The correlation
between pressure fluctuation and vertical velocity fluctuation at the surface is
Z ∞
1
Iw (0) = + δ pδ w(z = 0) = g Sdiss (k)dk (20)
ρw 0

and the main problem is how to model the depth dependence of δ pδ w. One could perhaps argue that the
depth dependence may be modelled in a similar way as the depth dependence of the Stokes drift (i.e. assume
potential flow with the usual exp(−2kz) factor inside the integral), but I would expect that the main action of
wave dissipation is in a layer of thickness of the wave height HS . However, it is emphasized that there are still a
number of open questions regarding the nature of surface wave dissipation. The suggested causes of the wave
dissipation range from large scale wave breaking to microscale breaking or even by ocean eddies generated by
unsteady large scale waves. Each different process will have a different penetration depth and for simplicity
it is assumed here that these lengthscales can all be lumped together to one wave height scale. Therefore the
following depth dependence for I(z) is suggested:
1
Iw (z) = + δ pδ w = Iw (0) × Iˆw , Iˆw = e−z/z0 (21)
ρw
where the depth scale z0 ∼ HS will play the role of a roughness length. The surface value of Iw may be obtained
from Eq. (19), realizing that by definition Iw is negative, hence
Iw (0) = −α w3∗ (22)
Using Eq. (21) in (16) the TKE equation becomes
√
∂e ∂ ∂e ∂ Iw (z) 2 ∂ US
 
2 2 3/2
= lqSq + + νm S − w∗
2
− νh N −
2
e . (23)
∂t ∂z ∂z ∂z ∂z Bl(z)
At the surface there is no direct conversion of mechanical energy to turbulent energy and therefore the flux of
turbulent energy is assumed to vanish. Hence the boundary conditions become
∂e
lqSq =0 for z = 0, (24)
∂z
∂e
=0 for z → ∞. (25)
∂z
The values used in the empirical constants are from the Mellor-Yamada model. They are
(SM , Sq , B) = (0.39, 0.2, 16.6) (26)
Note that in order to agree with the turbulence results in case there is a balance between production and dissi-
3/4
pation of kinetic energy the parameters SM and B satisfy the relation B1/4 SM = 1.

Technical Memorandum No. 634 11

 Ocean Wave effects on the daily cycle in SST

3.3.1 Buoyancy and the mixing length l(z).

The description of the TKE equation is concluded by means of a discussion of buoyancy effects and the choice
of the mixing length. In the upper ocean effects of stratification are important. In this paper the present mixed
layer model will be applied to the prediction of the diurnal cycle in SST. Extreme events typically arise for low
winds. At sunrise the upper ocean is usually neutrally stably stratified and the temperature profile is almost
uniform. When the sun starts shining the top layer of the ocean gets heated up resulting in stable conditions
which reduce the heat transport to the layers below. As a consequence a considerable amount of heat is retained
in the top layer which may have a thickness of a few decimeters only. In the course of the day more and more
heat is added to this top layer with the consequence that the layer becomes more and more stable, reducing
heat transport to the layers below even more. In the extreme circumstances of low winds of 1 m/s the Obukhov
length may go down to a few centimetres, which is much smaller then what is encountered in the atmospheric
case. An adequate modelling of these extremely stable cases is clearly of the utmost importance, but little
empirical evidence is available for these extreme circumstances. Notable exceptions are the works of Cheng
and Brutsaert (2005) and of Grachev et al. (2007).
In the presence of stable stratification it may be argued that buoyancy gives rise to a reduction of momentum
and heat transport, because when the Richardson number would pass 1/4 then fluid motion will be damped.
Following Csanady (1964), Deardorff (1980), Britter et al. (1983) and Wyngaard (1985), this means that there
is an additional parameter which may determine the transport properties of the upper ocean, namely the Brunt-
Väisälä frequency N. Under very stable conditions one would expect that most of the ’turbulent’ energy is
concentrated near N which suggests that the mixing length is limited by an additional length scale lb = q/N.
The eddy viscosity can then be estimated by
−1/2
ν ∼ qlb ∼ qlRit (27)

where

Rit = (Nl/q)2 (28)

is the Richardson number for turbulent eddies and the mixing length l is chosen as the usual one for neutrally
stable flow, i.e.

l(z) = κ z (29)

with κ = 0.4 the von Kármán constant. On the basis of Eq. (27) which is valid at large Rit , Noh and Kim
(1999)2 suggested that the dimensionless parameters SM,H can be represented by

SM,H /S0 = fM,H (Rit ); fM,H = aM,H (1 + bM,H Rit )−1/2 + cM,H (30)

with aM,H , bM,H and cM,H empirical constants. In fact, Noh and Kim (1999) have chosen zero values of cM,H ,
but a number of studies have suggested that at least SM should have a finite value of cM in order to represent
effects of internal waves on momentum transport (Pacanowski and Philander, 1981; Strang and Fernando, 2001;
Sukoriansky et al. 2005). Finite cM has important consequences for the turbulent transport properties: while
for zero cM there is a critical value of the gradient Richardson number above which there is no transport, in
case of finite values of cM a critical Richardson number does not exist in agreement with the notion that also
internal waves may give rise to momentum transport. Furthermore, also the diffusion of turbulent kinetic energy
2 Baas et al. (2008) followed a similar idea but rather than modifying the parameters S
M,H they modified the mixing length directly
by assuming l = l(κ z, lb ). But by inspection of (12) it is realized that this amounts to the same thing.

12 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

is expected to be affected by effects of stratification as the size of the eddies is limited under strongly stable
circumstances. And the same applies to the coefficient B in the dissipation. As a consequence

Sq /Sq0 = B/B0 = fM (Rit )

thus under stable conditions the TKE transport enjoys the same reduction as the momentum transport. The
coefficients S0 , Sq0 and B0 assume the values as given in Eq. (26).
Finally, the case of unstable stratification (Rit < 0) needs to be modelled properly as well. It is assumed that
also in this case the relevant parameters depend on the turbulent Richardson number Rit but the functional
dependence is different. In this paper the following form is chosen for fM,H if Rit < 0:

dM,H Rit
fM,H = (aM,H + cM,H )(1 +
1 + dM,H Rit

where dM,H = −20 and fM,H is continuous at Rit = 0 while for Rit → −∞ the dimensionless parameter fM,H
is twice as large as its value at the origin. Although not shown explicitely here, this choice results in good
agreement with the parametrizations of dimensionless shear function and virtual potential temperature gradient
obtained from the Kansas field campaign (Businger et al., 1971). However, the experience from simulations of
the diurnal cycle suggests that the evolution of sea surface temperature and surface current is fairly insensitive
to details of how transport in unstable circumstances is represented.

4 Some properties of the TKE equation.

In §3 the mixed layer model has been described and it is straightforward to solve these equations numerically
(see e.g. Kondo et al., 1979, Mellor and Yamada, 1982, and Noh and Kim, 1999). Here, some interesting
properties of the TKE equation will be discussed, in particular regarding effects of ocean waves on turbulent
transport and effects of buoyancy. The discussion will be restricted to the steady state case.
Consider the steady state version of the TKE equation and eliminate the shear S and the buoyancy frequency N
using the equations for momentum (11) and heat (14). From (11) one obtains for the shear

νm S = −w2∗ 1 − T̂

Similarly, integrating (14) once with respect to depth z and prescibing the heat flux Qh at the surface one finds

∂T Qh + R(0) − R(z)
νh =−
∂z ρw c p

In order to eliminate the buoyancy frequency N = gρ ′ /ρ it is assumed that the water density is a function
of temperature only, hence ρ = ρ (T ) and therefore the vertical gradient in density can be connected to the
temperature gradient through the thermal expansion coefficient αw , i.e.

1 ∂ρ ∂T
= −αw .
ρ ∂z ∂z

Next, one introduces the dimensionless turbulent velocity Q,

 1/4
B
q = w∗ Q and w = Q3 . (31)
SM

Technical Memorandum No. 634 13

 Ocean Wave effects on the daily cycle in SST

Furthermore, introduce a new length scale x,

dz dz
Z
dx = q ⇒ x= q (32)
l 13 Sq B l 13 Sq B

where it is noted that the range of the new variable x is from −∞ to ∞ because the turbulent mixing length
l(z) = κ z vanishes at the surface. The TKE equation (23) then assumes the simple form

d2w
2 −1/3
− w + 1 − T̂ w − ζ fM = S(x), (33)
dx2
where the source function reads

d Iˆw
dÛs
S(x) = Φ0 + µ La−2 1 − T̂ . (34)
dx dx

with Φ0 = µα , µ = 3/Sq0 B0 and La = (w∗ /US (0))1/2 is the turbulent Langmuir number. Here, the left-hand
p

side of the dimensionless form of the TKE equation contains the processes which are usually encountered in
the atmospheric surface layer, namely diffusion, dissipation, turbulence production by shear and buoyancy. The
stability parameter ζ is defined as ζ = z/L where L is the Obukhov length scale

ρ w3∗
L= . (35)
κ gνh d ρ /dz

which is the height where shear production and buoyance balance. Making use of the temperature profile and
the relation between density gradient and temperature gradient, the Obukhov length becomes

ρw c p w3∗
L= . (36)
κ gαw (Qh + R(0) − R(z))

and, because of the local definition of the Obukhov length, radiative forcing is included in a natural way in the
expression for L (cf. Large et al., 1994). The right-hand side of (33) gives the effects of ocean waves on the
mixing in the upper ocean: the first term represents effects of wave dissipation which affect mixing close to the
ocean surface, while the second term (which depends on the turbulent Langmuir number) represents the effect
of Langmuir circulation which transports heat and momentum to the deeper parts of the ocean.
The differential equation for w, Eq. (33), has of course to be supplemented by boundary conditions. They are
given in Eqns. (24)-(25). In terms of the unknown w they become

dw
→ 0 for x → −∞; w → 1 for x → ∞, (37)
dx

4.1 The Local Approximation.

It is, as far as I know, not possible to obtain for the general case the exact solution of the nonlinear boundary
value problem (33), (37). Therefore, an approximate solution is obtained by showing that in the present context
effects of diffusion can be ignored and the search for the turbulent velocity reduces then to solving an algebraic
equation. Note that this approach is not feasible in the original Graig-Banner problem because diffusion is
essential in order to transport the turbulent kinetic energy through the surface layer. However, here a different
route is followed as the pressure vertical velocity correlation term in the TKE equation has been explicitely
modelled in terms of the energy flux and the profile function T̂ .

14 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

Let us now study the solution of the boundary value problem (33), (37). It should be noted that (33) is a
nonlinear differential equation which only in cases where a first integral can be found may be solved exactly.
An example is given in Janssen et al. (2004) who solved the Craig-Banner problem exactly. However, when
effects of buoyancy are present or when Langmuir turbulence and wave dissipation (in the form modelled here)
is important it is not possible to find a first integral. Therefore an alternative approach will be followed which
was suggested by Øyvind Saetra. In fact, this approach was also followed by Craig (1996) although it is not
mentioned explicitely in his paper. Inspecting the differential equation for w it is realized that the nonlinearity
only comes from the w−1/3 term and therefore the nonlinearity is fairly weak. It is therefore suggested to
replace the w−1/3 term by its equilibrium value for large x. Far away from the sea surface the diffusion term
vanishes while also the wave dissipation and Langmuir circulation term S(x) becomes small. However, it is not
known how the buoyancy terms behaves for large x so in the present discussion effects of buoyancy are ignored.
The equilibrium value for w then follows from the balance of shear production and dissipation (which is the
’typical’ situation in the atmospheric surface layer), hence w = (1 − T̂ )3/2 . Therefore, the nonlinear differential
equation for w becomes approximately

d2w
− w = −(1 − T̂ (x))3/2 + S(x) (38)
dx2
It is straightforward to solve the linear boundary value (37)-(38) by means of the Green function technique.
The solution becomes
Z ∞ h i
w= dx0 G(x, x0 ) −(1 − T̂ (x0 ))3/2 + S(x0 ) . (39)
−∞

where the Green function G(x, x0 ) is given by

 1 x−x
 − 2 e 0 , x < x0 ,
G(x, x0 ) = (40)
− 12 ex0 −x , x > x0 .


The current profile can easily be obtained by rewriting Eq. (13) in terms of an integration over the x-variable.
The result is
Z x
dx
u(z)/w∗ = −τ0
1/2

1 − T̂ , (41)
xh w1/3
1/2
where τ0 = (Sq0 B0 /3)1/2 and where x = xh corresponds to the depth H where the current profile vanishes. In
this Section H = 5HS is chosen.
The solution (39-41) is readily evaluated on the computer. In order to do this the decay length scale z0 of the
wave-induced stress needs to be specified, i.e.

z0 = 0.5HS , (42)
1/2
and HS is the significant wave height, HS = 4m0 , with m0 the zeroth moment of the wave spectrum. The
windspeed is 2.5 m/s, the turbulent Langmuir number is 1/4 and the dimensionless energy flux α is equal to
100, which is a typical value in the Tropics (see Fig. 3). This low wind speed example has been chosen because
under these circumstances a diurnal cycle in the sea surface temperature and in the surface drift may be present.
In this example it is assumed that there is only windsea present, so the significant wave height follows from
HS = β U102 /g, with β = 0.22. The Stokes drift decay length scale then follows from k = g/U 2 . For U =
s 10 10
2.5 m/s the significant wave height is only 14 cm so that the ’roughness’ length is about 7 cm. The air friction
velocity u∗ is 8 cm while the water friction velocity w∗ is about 0.3 cm. Finally, the Stokes wavenumber ks is
about 1.6 rad/m.

Technical Memorandum No. 634 15

 Ocean Wave effects on the daily cycle in SST

0.01

0.1
log(z/H_S)

w(z)~q^3 (exact)
u(z)/w* (exact)
10 w(z)~q^3 (approx)
u*(z)/w* (approx)

100
1 10 100
w(z) or u(z)/w*

Figure 4: Profile of w = Q3 and current profile in the ocean column near the surface. The approximate solution (43),
based on the slowly varying wave dissipation source function and the gradient in the Stokes drift, are shown as well.

Results for w and current u(z)/w∗ as function of dimensionless depth z/HS are displayed in Fig. 4. Note that
there are important differences between this solution and the results of Graig and Banner (1994). While in their
approach w obtains its maximum value at the surface z = 0, this is evidently not the case in the present approach
as the maximum is now at about a depth of the order of the significant wave height (which makes by the way
perfect sense).
The solution (39) although elegant is still awkward to deal with in practical applications because an integral
needs to be evaluated. However, the Green’s function (40) looks like a δ -function, therefore assuming that the
bracket term in (39) varies slowly compared to the Green’s function one finds the approximate solution

w ≈ (1 − T̂ (x))3/2 − S(x), (43)

and the approximate solution for w(z) and u(z)/w∗ is shown in Fig. 4 as well. The agreement between approxi-
mate and exact solution seems fair. Note that the approximate solution, which from now on will be referred to as
the local approximation, is based on the assumption that the production and dissipation terms are slowly vary-
ing with respect to the diffusion. In fact, (43) follows immediately from the neglect of the turbulent diffusion
term in the kinetic energy budget (33).
In terms of the depth variable z the solution (43) can be written explicitely as

d Iˆw dÛS
w ≈ (1 − T̂ )3/2 − ακ z fM − La−2 κ z fM (1 − T̂ ) , (44)
dz dz
and it is now straightforward to estimate the respective contributions of shear production, wave dissipation
and Langmuir turbulence to the dimensionless turbulent velocity Q = w1/3 . This will be done by taking the
maximum of the individual terms. The maximum of the shear production term is 1, while the maximum of the

16 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

wave dissipation contribution is ακ e−1 ≈ 15 at z = z0 and the maximum contribution by Langmuir turbulence is
La−2 κ e−1 ≈ 2 (at z = 1/2ks ). Based on these estimates it seems that near the surface the most relevant process
for mixing is wave dissipation because it is an order of magnitude bigger than the other two terms, however the
turbulent velocity is only enhanced by a factor 2.5 because the sum of the contributions is raised to the power
1/3. Nevertheless, Langmuir turbulence should be relevant as well as this process penetrates into the deeper
layers of the ocean. This is illustrated in Fig. 5 for the special case of low wind of this §. For comparison

0.01

0.1
z/H_S

Total
shear + breaking
10 shear + Langmuir
Monin-Obukhov similarity

100
1 10 100
w(z)

Figure 5: Profile of w = Q3 according to the local approximation in the ocean column near the surface. The contributions
by wave dissipation (red line) and Langmuir turbulence (green line) are shown as well. Finally, the w-profile according
to Monin-Obukhov similarity, which is basically the balance between shear production and dissipation, is shown as the
blue line.

purposes I have shown the w-profile in case of Monin-Obukhov similarity, which consists of a balance between
shear production and dissipation, and I have shown the impact of switching off Langmuir turbulence and wave
dissipation. The Figure shows that indeed the maximum in w by wave dissipation is close to the sea surface at
a depth z = z0 while the maximum by Langmuir turbulence is at larger depth of 1/2ks . These scales are widely
different because ocean waves are weakly nonlinear which means that their ’typical’ steepness ks HS << 1.
As a consequence the ratio of the penetration depths by wave dissipation and Langmuir turbulence, given by
2ks z0 = ks HS , is small as well.
Therefore it is evident that there are two regimes. The first one is close to the surface and is dominated by wave
dissipation. Around 4 times the roughness length a transition to a different regime is to be noted, namely one
dominated by the production of Langmuir turbulence. Hence, it is seen that there are two transport mechanisms
operating in the surface layer of the ocean. Up to a few wave heights wave dissipation is dominant in the
diffusion of momentum and heat and the transport of these quantities is taken over by Langmuir turbulence in
the deeper part of the surface layer. The enhanced transport by wave processes gives rise to much flatter profiles
near the surface. This may be inferred from Fig. 6 where current profiles from the Monin-Obukhov similarity
model are compared with current profiles when wave dissipation and Langmuir turbulence play a role. The

Technical Memorandum No. 634 17

 Ocean Wave effects on the daily cycle in SST

0.01

0.1
z/H_S

Total
shear + breaking
10 shear + Langmuir
Monin-Obukhov similarity

100
0 1 2 3 4 5 6 7 8 9 10
u(z)/w*

Figure 6: Current profile near the surface. The impact of wave dissipation and Langmuir turbulence is shown as well.
The Monin-Obukhov similarity gives the usual logarithmic profile.

surface current reduces from about3 7w∗ to 2.5w∗ , which is a considerable reduction. As a consequence, it
is expected that these two processes will play an important role in the determination of the amplitude of the
diurnal cycle. Finally, it is also concluded that a mixed layer model which has only a representation of Langmuir
turbulence is not sufficient as it will overestimate the amplitude of the diurnal cycle. If one is interested in
modelling the diurnal cycle then probably only the first few metres of the upper ocean need to be considered. In
that event wave dissipation is seen to be the dominant process for heat transport, however there is no reason to
disregard effects of Langmuir turbulence from the outset as it is very straightforward to take both into account.
In addition, during the duirnal cycle there will also be episodes when the flow is neutrally stable or unstable.
Langmuir turbulence will then play a pronounced role.
This Section is concluded with the following comment. So far two things have been learned. First, if one
describes the effect of wave dissipation through the correlation term of pressure and vertical velocity it seems
a valid assumption to neglect the effects of the diffusion of turbulent kinetic energy. Second, it seems possible
to combine in a simple way several physical processes that affect the mixing in the upper-ocean. From the
previous discussion it appears that if one has, apart from shear production SP , several processes P1 , P2 , P2 , ...
that contribute to turbulent mixing then the turbulent velocity q(z) of the combination of all those processes is,
following Eq. (43), given by
n o1/3
3/4
q = SP + P1 + P2 + P3 + ... .

The reason that processes can be added via an ’1/3’- rule is because dissipation is proportional to the third
power of q, while the shear production term has been linearized by replacing w1/3 by its equilibrium value and
3 Using the approximate solution given in the Remark on page 10 one may estimate the surface current in case of Monin-Obukhov

similarity. One finds u(0)/w∗ ≈ log(H/y0 )/κ = 7.23, with H = 5HS . Using a higher resolution version of my software a perfect match
with the approximate result is found.

18 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

the other processes are assumed to be independent of the turbulent velocity q. Because of the ’1/3’- rule it
makes sense, as done in the present work, to make plots involving w = Q3 as for w different processes may be
added.
The ’1/3’- rule also gives rise to consistent scaling behaviour in case of Monin Obukhov similarity. This is the
case that there is no wave dissipation and no generation of Langmuir turbulence. In that event, w from Eq. (43)
becomes w = (1 − T̂ )3/2 and the current profile becomes (cf Eq. (41)
Z x
dx
1/2
u0 (z) = −τ 1/2 1 − T̂
xh w1/3

therefore the current scales with the square root of 1 − T̂ which agrees with the scaling behaviour mentioned in
the Remark on page 10. Nevertheless, it should be pointed out that the ’1/3’- rule is not always appropriate. In
particular, the buoyancy term has so far not been considered but this effect is expected to play an important role
far away from the surface, thus making it difficult to give an estimate of the equilibrium value of q. In addition,
the buoyancy term is a fairly sensitive function of q and therefore it is not easy to linearize it.
In the remainder of this Section I will therefore refrain from linearizing the shear production term, but I will
disregard the effect of diffusion of turbulent kinetic energy. Therefore, the TKE equation becomes, neglecting
diffusion in (23),
√
∂ e ∂ Iw (z) 2 ∂ US 2 2 3/2
= + νm S − w∗
2
− νh N −
2
e . (45)
∂t ∂z ∂z Bl(z)
and in terms of the dimensionless variables introduced in this Section one has, with dimensionless time τ =
2 w t/l(z),
SM ∗

1 ∂ 2 1
Q4 − α (Q)Q − β ,


Q =− (46)
2 ∂τ Q
where

2
α (Q) = −ζ fM − S(x), β = 1 − T̂ .

and α still depends on Q through the buoyancy term −ζ fM (Rit ). For completeness the source function, which
represents effects of wave dissipation and Langmuir turbulence, is repeated from Eq. (34):

d Iˆw
dÛs
S(x) = Φ0 + µ La−2 1 − T̂ .
dx dx
In the local approximation the TKE equation has now been simplified considerably, and the search for its
equilibrium solution has been reduced to the solution of an almost quartic problem. For example, for the
neutrally stable case the equilibrium solution to Eq. (46) follows from the real, positive root of the quartic
equation Q4 − α Q − β = 0 and it can readily be shown that the ’1/3’-rule is a good approximation to this root.
In the next section (46) is used in a discussion on stratification effects.

4.2 Effects of stratification.

First, effects of stratification in the atmospheric context will be studied and the findings will be applied to
the mixed layer of the upper ocean. In the atmosphere close to the surface there is a balance between shear
production, buoyancy and dissipation, as forcing is usually absent. This will be called the case of Monin-
Obukhov similarity.

Technical Memorandum No. 634 19

 Ocean Wave effects on the daily cycle in SST

4.2.1 Monin-Obukhov similarity.

In the atmosphere, stability effects are usually studied in terms of the dimensionless shear function φm and the
dimensionless virtual potential temperature gradient φh . These dimensionless functions are defined as

κz ∂ u κ z ∂ θv
φm = , φh = , (47)
u∗ ∂ z θ∗ ∂ z

where u∗ is the air friction velocity and θ∗ = −w′ θv′ /u∗ is a turbulent temperature scale. The dimensionless
shear function measures deviations from neutral circumstances as for the logarithmic wind profile φm = 1, and
similarly φh measures deviations from the logarithmic virtual temperature profile. Using the local scaling theory
of Nieuwstadt (1984) it can be argued that the profile functions are only a function of the stability parameter
ζ = z/L, where L is the local Obukhov length defined as

u3∗ θv
L=− (48)
κ gΦh

Here θv is the virtual potential temperature and Φh = δ wδ θv is the virtual potential temperature flux. The shape
of the φ functions is usually determined from observations acquired during field campaigns, but high measure-
ment accuracy is required because the fluxes become weak during strongly stable conditions. Alternatively, a
realistic theoretical model of turbulent flows with stable stratification has been developed by Sukoriansky et al.
(2005) providing additional information on how to model stratification effects.
The Kansas field campaign (Businger et al., 1971)4 was one of the first experiments to propose realistic
parametrizations for the φ functions. For stable conditions it was found that φm and φh varies essentially
linearly with ζ over the observed stability range between 0 and 1. A fit gives

φm = 1 + 4.7ζ , φh = 0.74 + 4.7ζ , for 0 < ζ < 1. (49)

On the other hand, for unstable conditions a good fit was found to be

φm = (1 − 15ζ )−1/4 , −2 < ζ < 0. (50)

A similarly looking fit was found for φh . However, in the upper ocean strongly stable conditions occur with
ζ of the order 10 or even larger. These conditions are much more extreme than typically encountered for the
atmospheric surface layer except perhaps for air flow over ice. Therefore, relatively little is known in these
extreme circumstances, and in fact conflicting conclusions about properties of strongly stable turbulence have
been reached in the past. The problem is best illustrated by the behaviour of the Prandtl number Pr defined as
νm φh
Pr = = ,
νh φm
as function of the gradient Richardson number Ri given by

N2
Ri = .
S2
A vast number of studies (see e.g. Kondo et al., (1978); Kim and Mahrt (1992), Strang and Fernando (2001),
Sukoriansky et al. (2005), Zilitinkevich et al. (2007) and many others) suggest that for strongly stable flow,
hence for a Richardson number larger than the critical value of 1/4, the Prandtl number is larger than 1, while
4 Note that in order that φ (ζ = 0) = 1 the authors had to choose a von Kármán constant of 0.35, which does not agree with the
m
accepted value of 0.4.

20 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

for small Ri (the neutral limit) the Prandtl number is smaller than 1 (as is evident from Eq. (49)). In other words,
for strongly stable flow, momentum is mixed more efficiently than heat. This is thought to be an indication of
internal gravity wave activity which can produce transfer of momentum but only little heat transfer (as long as
the waves do not break).
However, in sharp contrast to these findings, Cheng and Brutsaert (2005) and Grachev et al. (2007) conclude
from the SHEBA observations, which were obtained for strongly stable flow over ice, that heat transport is more
efficient than momentum transport hence Pr < 1. Grachev et al. (2007b) have analyzed their findings in some
detail, but no physical explanation has been offered. They point out that there is a spurious correlation between
Pr = φh /φm = Ri/Ri f and measures of stability such as the local Richardson number Ri = ζ φh /φm2 , the flux
Richardson number Ri f = ζ /φm and the stability parameter ζ = z/L because Pr and these stability parameters
share parameters such as vertical gradients in mean wind speed and potential temperature and the corresponding
fluxes. But the argument of spurious correlation is not really convincing, as theoretical developments such as
given by Sukoriansky et al. (2005), who applied renormalization techniques to find the Richardson number
dependence of eddy viscosity νm and eddy diffusitivity νh , show that Pr > 1 for large Ri. Also direct numerical
simulation results by Shih et al. (2000) and the observations of Strang and Fernando (2001) confirm this.
Furthermore, both Cheng and Brutsaert (2005) and Grachev et al. (2007) find a levelling-off of the similarity
functions φm and φh as a function of the stability parameter ζ which is so large that it conflicts with the steady
state TKE equation. In order to see this, apply now the TKE equation (46) to the atmospheric problem where
forcing is absent. In the steady state one then finds

Q4 + ζ fM Q − 1 = 0. (51)

The similarity functions can be written in terms of the present dimensionless variables and the result is

1 1
φm = ; φh = (52)
Q fM Q fH

Expressing Q in terms of φm and substituting the result into (51) gives

φm4 − ζ φm3 − fM−4 = 0. (53)

For neutrally stable conditions, Eq. (53) reduces to the well-known KEYPS formula φm4 − ζ φm3 − 1 = 0 (Panof-
sky, 1963). This is usually regarded as an equation for the dimensionless shear function. It is advantageous,
however, to turn things around, i.e. to regard (53) as an equation for fM because φm is known from the observa-
tions. Rearranging (53) one finds for fM
−1/4
fM = φm3 (φm − ζ )

. (54)

From (54) it immediately evident that there is only a real solution for fM when φm > ζ . If the TKE equation
(53) holds then the condition φm > ζ has important implications for parametrizations of the dimensionless
shear function. For example, the Grachev et al. (2007) parametrization for φm becomes according to (54)
unrealistic because it will cross the line φm = ζ for ζ ≈ 17, which is well inside the stability range that the
dimensionless shear has been observed. This result can also be understood in physical terms. The present TKE
equation expresses a balance between shear production, buoyancy and dissipation. Now dissipation is always
positive therefore buoyancy can never exceed production, or in terms of the present dimensionless variables,
φm > ζ . Clearly, in the context of the present TKE formulation it is not possible to model a levelling off of
the dimensionless shear function as found in the SHEBA data set by Grachev et al. (2007a) and by Cheng
and Brutsaert (2005). A possible resolution of the conflict between the SHEBA data set (strongly stable flow
over ice) and the standard TKE equation, which is based on a balance between shear production, buoyancy and

Technical Memorandum No. 634 21

 Ocean Wave effects on the daily cycle in SST

2 12

1.8 Monin-Obukhov: momentum

Monin-Obukhov: heat
10
1.6

1.4
8
1.2
ν, κ

Pr
1 6

0.8
4
0.6

0.4
2
0.2

0 0
0.01 0.1 1 10 0.01 0.1 1 10
Ri Ri

Figure 7: Eddy viscosity νm and heat diffusivity νh as function of the local Richardson number Ri. In the right panel the
Prandtl number Pr is shown as function of Ri.

dissipation, might be that during SHEBA there was an additional source for the production of turbulent kinetic
energy. Thus, the SHEBA results cannot be used as a guideline for the present modelling work.
In order to be specific a choice has to be made for the coefficients in the parametrization (30) of fM and fH . This
choice will be based on the one hand on the Kansas field results in the weakly stable limit, while for the strongly
stable limit guidance from the renormalization work of Sukoriansky et al. (2005) is taken. In particular, the
following choice for fM and fH is made:

fM = aM (1 + bM Rit )−1/2 + cM , fH = aH (1 + bH Rit )−1/2 (55)

where aM = 0.8, bM = 100, cM = 0.2, aH = 1.4 and bH = 80. From (55) it is seen that fH vanishes for large Rit
while fM asymptotes to a finite value of cM = 0.2. For small turbulent Richardson numbers fH is larger than
fM , hence, with Pr = φh /φm = fM / fH , it is found that Pr ≈ 0.71 < 1 for Rit → 0 in agreement with results from
the Kansas field campaign. In order to determine some of the coefficients an approximate solution was used. In
fact, an approximate solution for the dimensionless turbulent velocity Q may be found for small values of the
stability parameter ζ . One finds Ri ≈ ζ /aH and the eventual result for the dimensionless shear function is
1 
φm ≈ 1 + ζ 1 + 2(1 − cM )bM S02 /aH .

4
but this approximation is only valid for a relatively small range of the stability parameter, ζ < 0.1. The choice
of coefficients given below Eq. (55) together with S0 = .39 gives a value of the slope of 4.6 which is close to
the value reported by the Kansas field campaign given in Eq. (49). In addition, anticipating results discussed
below the right panel of Fig. 7 shows that up to a gradient Richardson number of 0.1 the Prandtl number is a
constant so that in agreement with the Kansas data φh has the same slope as φm for small ζ .
On the other hand, for large turbulent Richardson number, Ri > 0.2, the Prandtl number is larger than 1,
indicating that in this domain momentum transfer is more efficient than heat transport, in agreement with
Sukoriansky et al. (2005) and the observations of Strang and Fernando (2001). In order to show explicitely the
effect of buoyancy on the transport properties, the eddy viscosity νm and heat diffusivity νh are normalized with
the eddy viscosity ν = κ u∗ z for neutrally stable flow. In terms of the present dimensionless variables one finds
νm /ν = fM Q while νh /ν = fh Q, hence the normalized viscosities are simply the inverse of φm and φh . Using
(55) in (51) and solving for Q by iteration the resulting transport coefficients as function of the Richardson
number Ri = ζ φh /φm2 are shown in the left panel of Fig. 7, while the Prandtl number Pr as function of Ri is
shown in the right panel. Comparing this Figure with Figs. 8 and 9 of Sukoriansky et al. (2005) it is seen that

22 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

there is good qualitative agreement with the results using renormalisation techniques to obtain the transport
coefficients. In particular, as already pointed out, a finite value of cM in (55) does not give rise to a critical
value of the gradient Richardson number as a finite cM represents additional diffusion by e.g. internal gravity
waves and/or intermittency. At the same time, the consequence is that for large Richardson number momentum
transport dominates heat transport.
4 4

Monin-Obukhov: momentum Monin-Obukhov: momentum

3.5 Monin-Obukhov: heat 3.5 Monin-Obukhov: heat
With Waves: momentum With Waves: momentum
With Waves: heat With Waves: heat
3 3

2.5 2.5

ν, κ
ν, κ

2 2

1.5 1.5

1 1

0.5 0.5

0 0
0.01 0.1 1 10 0.01 0.1 1 10
Ri ζ

Figure 8: Eddy viscosity νm and heat diffusivity νh as function of the local Richardson number Ri, showing the effects of
wave dissipation and Langmuir turbulence. In the right panel the same parameters are shown as function of the stability
parameter ζ .

4.2.2 Wave effects and buoyancy.

In this section the combined effects of wave dissipation, Langmuir turbulence and buoyancy on the properties
of turbulence in the mixed layer are studied. It is assumed that the parametrization of stratification for the
atmosphere (cf Eq. (55)) also holds for the oceanic case. The set of equations to be solved consists of the steady
state version of (46) together with (28), (30), and (36). This set of equations does not have an exact solution
because owing to effects of stability α in (46) depends strongly on the dimensionless turbulent velocity Q. The
set of equations was therefore solved by means of an iteration scheme using starting values Q = 1, fM = fH = 1.
Because the stability effects are modelled in terms of the turbulent Richardson number Rit = Nl(z)/q, the Brunt-
Väisälä frequency needs to be expressed in terms of Q. Introducing N∗ = l(z)N/w∗ one finds N∗2 = ζ / fH Q.

In Fig. 8 effects of wave dissipation and Langmuir turbulence on transport coefficients for momentum and
heat is shown. In the left panel these coefficients are plotted as function of the gradient Richardson number Ri,
while in the right panel they are shown as function of the stability parameter ζ . Of course, waves gives rise
to enhanced transport, but, remarkably, in the presence of this additional forcing the transport coefficients are
not a single-valued function of Ri. However, in terms of the turbulent Richardson number Rit or the stability
parameter ζ (as shown in the right panel of Fig. 8) the transport coefficients are unique functions.
In Fig. 9 effects of stratification on the profile for w = Q3 as function of dimensionless depth z/HS are shown.
For this plot the parameters from the example in §4.1 are used and, in addition, the heat flux Qh was 100 W/m2
while the water temperature T was 303 K. In order to vary the Obukhov length scale L, as determined by Eq.
(36), wind speed values of 2.5, 1.25 and 1 m/s were used respectively. It is instructive to compare results
for w(z) with the case of no stratification. It is then immediately seen that, as expected, buoyancy has the
biggest impact on the turbulent velocity Q in the deeper layers of the ocean (note that L = 0.76 corresponds to

Technical Memorandum No. 634 23

 Ocean Wave effects on the daily cycle in SST

0.01

0.1
z/H_S

Neutral
Stable (L = 0.76 m)
10 Stable (L = 0.10 m)
Stable (L = 0.05 m)

100
0.1 1 10 100
w(z)

Figure 9: Dependence of w(z)-profile on effects of buoyancy.

L ≈ 5HS ). This means that according to this model the impact of Langmuir turbulence on upper ocean mixing
is considerably reduced in stable circumstances. For these particular examples the maximum in w, caused by
wave dissipation, is hardly affected by stability effects. Surpringly perhaps, this is a fairly general result. Only
when the heat flux was increased by a factor of 10 an appreciable reduction of the impact of wave dissipation
on the mixing was found (not shown). This apparent robustness of the wave dissipation impact on mixing can
be understood by once more noting that the maximum of w(z) occurs at z = z0 , where according to the present
model the roughness length scales with the square of the friction velocity. A significant impact of buoyancy on
the maximum is expected when L ≤ z0 . Using the definitions for L and z0 one finds U10 ≤ 3 × 10−4 Qh which,
even with a large value of Qh of 1000 W/m2 , is still a small wind speed.
Finally, in Fig. 10 the impact of stability on the equilibrium current is shown. The increase of the surface
current for increasing stability is mainly caused by the reduction of the effects of Langmuir turbulence. The
Figure illustrates that also in the surface current a diurnal cycle is to be expected. As a general remark it
is noted that under unsteady circumstances the impact of effects of stability, wave dissipation and Langmuir
circulation is somewhat reduced, while the temperature and current profile may occasionally be convex rather
then concave as in the steady state case. This will be shown in more detail in the next Section during a discussion
of the simulation of the diurnal cycle in SST.

4.2.3 A qualitative validation.

The present experimental knowledge of turbulence in the ocean surface layer is summarized by the works of
Terray et al. (1996), Drennan et al. (1996) and Anis and Moun (1995). Here, dimensionless dissipation,
defined as ε∗ = ε HS /F0 with F0 the energy flux into the ocean, is found to be a function of dimensionless depth
(z + z0 )/HS . In the case of Monin-Obukhov similarity one would expect that the ’Law of the Wall’ holds which
states that dissipation scales with z−1 as the turbulent velocity is constant. However, according to observations
of turbulence near the surface dissipation depends in a more sensitive manner on depth. Based on work Terray
et al. (1999) and of Burchard (2001), who summarised the observational knowledge, one finds near the surface

24 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

0.01

0.1
z/H_S

Neutral
Stable (L = 0.76 m)
10 Stable (L = 0.10 m)
Stable (L = 0.05 m)

100
0 2 4 6 8 10
u(z)/w*

Figure 10: Dependence of near-surface current profile on effects of buoyancy.

the fit

ε∗ = 0.78Z −2.78 , Z = (z + z0 )/HS .

which is valid for ε∗ > 0.01. These observations are quite useful to determine an important parameter in the
mixed layer scheme, namely the roughness length z0 or the corresponding gradient length scale of the wave
dissipation source function. Burkhard finds an optimal fit (however using a somewhat different turbulence
model) when z0 = 0.5HS . This finding has been confirmed here. In order to illustrate that the present model
indeed gives the correct scaling behaviour, Fig. 11 shows dimensionless dissipation versus (z + z0 )/HS for
the strongly stable case and for neutral stability and compares the model results with the above power law.
The agreement between the neutrally stable case and the fit to the data seems fair. Also note that according
to the present mixed-layer model there is a transition from wave dissipation driven turbulence to shear driven
turbulence, giving the ’Law of the Wall’ in the deeper layers of the ocean, while in the transition layer turbulence
is controlled by production of Langmuir circulation.

5 Numerical simulation of the diurnal cycle in SST and surface current.

In this section the mixed layer model described in §3 is applied to a simulation of the diurnal cycle in sea
surface temperature (SST) and the surface current. The relevant equations are (11), (12), (14), (15) and (23).
The boundary condition for the momentum equation is vanishing turbulent stress at the surface, while for the
temperature equation the turbulent heat flux at the surface is given by Qh /(ρw c p ). The turbulent kinetic energy
flux at the surface vanishes as well. At depth z = D, where in the present application D is of the order of 3 m,
current velocity u(z) and temperature T (z) are assumed to be given.
The equations for momentum, heat and turbulent kinetic energy are discretized in the vertical in such a way
that the fluxes are conserved, while the relevant quantities are advanced in time using an explicit scheme. The

Technical Memorandum No. 634 25

 Ocean Wave effects on the daily cycle in SST

1
(z+z_0)/H_S

10

Neutral
Stable (L = 0.76 m)
Fit to Obs

100
0.0001 0.001 0.01 0.1 1 10
eps H_S/w*^3

Figure 11: Dimensionless dissipation ε∗ = ε HS /F0 versus (z + z0 )/HS

time step was chosen to be 2 seconds and in order to garantee stability a limitation on the size of the diffusion
coefficients was imposed. With n labelling a particular layer and N the total number of layers, the vertical
discretization is obtained using a logarithmic transformation of the type

z(n) = zs eξ (n) − 1 , n ≤ N,
 

where ξ (n) = n∆ is discretized in a uniform manner and ∆ = log(D/zs + 1)/N. Typically, zs is of the order of
a few centimetres thus giving high resolution near the surface, which is needed to resolve the solar absorption
profile (15) appropriately, while away from the surface resolution degrades. For the simple example of constant
wind and sea state discussed below zs is chosen to be one-third of the roughness length z0 as it can be easily
shown that the mixing scales with z/z0 . In that case depth D becomes a multiple of the roughness length,
D ≈ 110z0 . However, in the general case of varying winds the transformation for z(n) would become time-
dependent. Although it is straight-forward to deal with a time-dependent coordinate transformation, it was
decided to choose for the general case a constant zs with zs = 0.025 m. Then the depth D is a constant as well,
D = 3.5 m as observations of temperature at that depth are available. In all applications the number of layers N
is equal to 8.
Finally, when integrating the TKE equation forward in time numerical errors may introduce small negative
turbulent kinetic energy so that determination of the turbulent velocity would fail because of taking the square
root of the energy. For security reasons, therefore, a minimum value of turbulent kinetic energy is introduced
being a small fraction of the equilibrium turbulent kinetic energy, emin = 0.0001w2∗ /2.

5.1 Synthetic example.

As a first test a five day simulation was performed with constant fluxes of momentum τ and heat flux Q while
the solar radiation followed a daily cycle according to R = R0 max[sin(ω t), 0] where ω = 2π /(24 × 3600). The
intention is to generate a steady daily oscillation in SST without drift in the temperature and to study effects

26 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

296 2

295 1.5

u(0)/u*
SST

294 1

293 0.5

With Waves With Waves

No Waves No Waves
292 0
0 24 48 72 96 120 0 24 48 72 96 120
Time (hrs) Time (hrs)

Figure 12: The left panel shows for pure windsea time series of SST for a constant wind speed of 2.5 m/s and a heat flux
of -150 W/m2 , while the daily average solar insolation is 350 W/m2 . The impact of disregarding ocean wave effects is
shown as well. The right panel shows the surface current normalized with the air friction velocity.

of ocean waves on shape and amplitude of the daily cycle. In order to achieve a steady oscillation values of
daily average insolation, heat and momentum flux have to be chosen appropriately. The momentum flux τ was
chosen equal to 0.0069 m2 /s2 , which, with a drag coefficient of 1.11 × 10−3 , corresponds to a wind speed of
2.5 m/s, while the heatflux was given the value - 150 W/m2 typical for the Arabian sea in May. Hence, in the
absence of radiative forcing the ocean would cool down. The constant R0 in the formula for the solar insolation
was given the value 350 × π so that the daily average irradiation is 350 W/m2 and the maximum irradiation is
1099 W/m2 . All other parameters such as the turbulent Langmuir number, the Stokes drift decay length scale
and the water friction velocity were chosen as in §4.1. Note that for these particular cases the decay length scale
z0 of the wave-induced stress and energy flux is assumed to be given by one-half the wind sea wave height.
In Fig. 12 time series of SST are shown over the five day period and are compared with a simulation without
wave effects. Note that in the simulations without wave effects the wave dissipation term and the Langmuir
term are switched off in the TKE equation (23), while also the wave-induced stress in the momentum equation
(11) is switched off. The boundary condition for momentum flux at the surface is then, of course, replaced
by the usual one, namely τ = −w2∗ . Surprisingly, even for a low wind speed case of 2.5 m/s, sea state effects
on the simulation of the diurnal cycle in SST are clearly visible. As expected, wave dissipation and Langmuir
turbulence give rise to an enhanced mixing and therefore a reduction in the diurnal cycle amplitude compared
to the case without wave effects. From the right panel of Fig. 12 a similar conclusion also follows for the
diurnal cycle in the surface current. The corresponding amplitude is fairly substantial. Furthermore, note that
while the time series for SST shows no drift in temperature or surface current in the simulation with waves,
a drift is clearly visible in the simulation without waves. Presumably, in the simulation with waves there is a
more efficient transport towards the deeper layers of the ocean.
In Fig. 13 profiles for turbulent velocity Q(z), temperature T (z) and velocity u(z) are shown. Four hours into
the simulation the ocean is warming up producing a stable layer as is evident from the fact that the turbulent
velocity is less than 1 in the deeper parts of the ocean. Temperature and velocity profile are not in equilibrium
because they have an S-shape. Eight hours later, at sunset, the upper part of the ocean is already turning
unstable because the ocean is cooling off as the heatflux, given by Q = −150 W/m2 is directed from ocean
to atmosphere. Therefore, in the upper part of the ocean the temperature profile is well-mixed and is slightly
lower at the surface than at z/HS ≈ 6 where the maximum temperature is found. The shape of the surface
current is now concave and it looks similar to the equilibrium profiles shown in Fig. 10. Finally, at sunrise, 24
hours into the simulation the temperature in the whole column is almost uniform and equal to its value at the
bottom of the domain. The reason is that during the night the whole ocean column becomes unstable giving
an efficient transfer of heat towards the atmosphere and towards the deeper parts of the ocean. The efficient

Technical Memorandum No. 634 27

 Ocean Wave effects on the daily cycle in SST

0.1 0.1

1 1
z/H_S

z/H_S
Q(t = 4 hrs) Q(t = 12 hrs)
10 T(t = 4 hrs) 10 T(t = 12 hrs)
U(t = 4 hrs) U(t = 12 hrs)

100 100
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Q, T, U Q, T, U
0.1

1
z/H_S

Q(t = 24hrs)
10 T(t = 24 hrs)
U(t = 24 hrs)

100
0 0.5 1 1.5 2 2.5 3
Q, T, U

Figure 13: Profile of turbulent velocity, temperature and current after 4 (left), 12 (middle) and 24 (right) hours from the
start of the simulation.

transfer is reflected by the observation that now the turbulent velocity is everywhere larger than 1. Furthermore,
the current is now the smallest because during the night also momentum has been transferred efficiently towards
the deep ocean.
In order to give an impression of the overall behaviour of the present mixed layer model a one-day simulation
was performed for different wind speed and solar insolation. The results are summarized in Fig. 14. The plot
shows that the amplitude in the diurnal cycle is a sensitive function of wind speed and the magnitude of the
solar insolation. For comparison also shown is the wind speed dependence of the amplitude in diurnal cycle
of SST when wave effects are switched off. In relative terms it is found that for low wind speed switching off
the wave effects increases the diurnal amplitude by about 20% while for larger wind speed (U10 > 5 m/s) the
increase is about 50%.

5.2 Simulation of buoy observations.

Next, a simulation with the mixed layer scheme is performed and validated against buoy observations of the
Arabian Sea Mixed Layer Dynamics Experiment at 15o 30’ N, 61o 30’ E during a 3-month period from March
to May 1995 (Baumgartner et al., 1997; Weller et al., 2002). The diurnal cycle of SST in the Arabian Sea
can be quite profound. The mixed-layer model is driven by hourly surface fluxes computed with the COARE
flux algorithm (Fairall et al., 1996) using Improved Meteorology (IMET) buoy observations. Temperature

28 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

<R> = 350; Q = -150

6 <R> = 300; Q = -150
<R> = 250; Q = -125
<R> = 200; Q = -100
5 <R> = 150; Q = -75
Diurnal Amplitude SST

<R> = 100; Q = -50

No Waves: <R> = 350; Q = -150
4

0
0 2 4 6 8 10
U10 (m/s)
Figure 14: Diurnal amplitude in SST as function of wind speed for different solar insolation and heat flux as indicated
in the legend. The sea state is pure windsea. For the most extreme case of insolation also the result in absence of wave
effects is shown.In relative terms the difference is largest for large wind speed.

observations and flux data were downloaded from the Woods Hole Oceanographic Institution web page. For
verification purposes observed temperature at a depth of 0.17 m are compared with the model counterpart. As
a boundary condition the observed temperature at a depth of 3.5 m is prescribed, while the current at that depth
is assumed to vanish. Sea state parameters such as significant wave height HS , its wind sea part HS,ws , the
mean wavenumber kS and the components of the Stokes drift are obtained from archived wave spectra from
the ERA-Interim (wave) analysis (Simmons et al., 2007). The 6-hourly wave parameters are interpolated in
time and supplied to the mixed-layer scheme. However, as explained in §2, it is not straightforward to obtain
the energy flux parameter α from archived spectra because the implicit factors of the integration scheme with
which the energy balance equation (1) is solved need to be known. For this reason a parametrisation of Terray
et al. (1996) is used. It is given by
α = 15χ exp −(0.04χ )4 ,
 

where χ = c p /u∗ is the wave age which characterizes the stage of development of the sea state. It is straight-
forward to obtain the wave age from archived spectra.
A number of experiments were performed with the present mixed-layer scheme. The first set of experiments
were done to decide what is, in the context of the present model, the most appropriate penetration depth and/or
roughness length z0 that represents the transfer of ocean wave motion to ocean turbulence. A number of choices
were tried, namely

1. relate z0 to the inverse of a typical wavenumber such as the mean wavenumber kS . This is the depth scale
one would expect when the conversion from wave motion to ocean turbulence is described by potential
theory.

Technical Memorandum No. 634 29

 Ocean Wave effects on the daily cycle in SST

2. relate z0 to the wave height of the wind waves. This expresses the nonlinear character of the wave
dissipation process.

3. relate z0 to the significant wave height including swell. This reflects that the dissipating ocean waves are
transported in the vertical by the longer waves.

The statistics from the comparison with the temperature observations are shown in Table 1 and it is clear that
the third option performs best as the bias is very small while in particular the standard deviation of error in SST
is only 0.12 K. Therefore, from now on the decay length scale will be given by z0 = 0.5HS where HS is the
significant wave height which represents both windsea and swell. For this case a 20 day section of the timeseries
for ∆T = T (0.17) − T (3.5) is shown in Fig. 15 while in Fig. 16 the modelled Diurnal SST Amplitude (DSA)
is compared with the observed one. The mixed-layer model seems to perform remarkably well, and it is noted
that
√ the standard deviation of the difference between observed and modelled DSA is larger by about a factor of
2 since DSA is the difference between the daily maximum and minimum in SST.
Some additional experiments were performed. In §4.1 it was argued that the diffusion term in the TKE equation
may probably be neglected. In order to verify this the mixed-layer model was run without diffusion in the TKE
equation and the verification statistics were found to be almost identical to the case with diffusion (not shown)
therefore confirming that neglect of diffusion in the TKE equation is a valid assumption. Furthermore, it is
of interest to study the importance of Langmuir turbulence in the simulation of SST. Therefore, Langmuir
turbulence was switched off and the resulting verification statistics are shown in Table 1 as well. It is seen
that Langmuir turbulence has a relatively small impact on the simulation of the diurnal cycle for the present
case. This can be understood by noting that for this particular example the average wavenumber over the three
month period was found to be < kS >≈ 0.066 so that the maximum contribution by Langmuir turbulence is

Modelled
2.5 Observed

2
SST

1.5

0.5

0
190 195 200 205 210
Days

Figure 15: Observed and simulated ocean temperature ∆T = T (0.17) − T(3.5) at 15o30’ N, 61o 30’ E in the Arabian Sea
for 20 days from the 23rd of April.

30 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

3
Modelled Diurnal Amplitude

0
0 1 2 3 4
Observed Diurnal Amplitude
Figure 16: Comparison of simulated and observed diurnal amplitude at 15o 30’ N, 61o 30’ E in the Arabian Sea for the
3-month period starting from 1rst of March 1995.

at z = 1/(2kS ) ≈ 7.5 m which is outside the domain that was modelled (recall that the boundary condition for
temperature was provided at a depth of 3.5 m). A factor that has much more impact on the simulation results
is how stratification effects are modelled. In order to illustrate the sensitivity to the shape of the stratification
function fM an experiment was performed where cM in Eq. (55) is set to zero. In that event there is a critical
Richardson number and, just like heat, momentum transport vanishes for large gradient Richardson number. As
can be seen from Table 1 this change has a significant impact on the verification statistics, with a large increase
in bias, standard deviation of error and normalized variability.
The final set of experiments explores the possible impact of waves on the simulation of the diurnal cycle. To
that end, wave effects were switched off in the TKE equation while also the wave-induced stress term in the
momentum equation was switched off. In this simulation, which is supposed to ignore wave effects, it makes no
sense to relate the roughness length to the sea state. Therefore, after some trial and error optimizing statistics,
the roughness length was given the value z0 = 1.5 m which is close to a suggestion by Graig and Banner
(1994). It is seen that also wave effects (and to be definite mainly wave dissipation) have a considerable impact
on the SST simulation as the standard deviation errors increase by about 40-50 %. Additional evidence of the
sensitivity of the diurnal cycle to the sea state may be found in the last two experiments. In the first one the
average value of energy flux parameter α over the three month period is used in the wave dissipation term of
the TKE equation. From Table 1 it is seen that this experiment gives almost the same statistics as the default
experiment (indicated by the boldface numbers). However, if one would take in stead the global average of
α , equal to 148, then the standard deviation of error in the DSA is seen to increase by 40% while variability
is reduced by 25%. Hence, for an accurate simulation of the diurnal cycle an accurate representation of wave
dissipation in space (and probably also in time) seems to be important. Returning to Fig. 3, which shows a
measure of the normalized energy flux, it is seen that regarding wave dissipation the Arabian Sea is a very

Technical Memorandum No. 634 31

 Ocean Wave effects on the daily cycle in SST

Table 1: Summary of Statistics of a number of experiments. Here, DSA is the Diurnal SST Amplitude, SD is the standard
deviation and VAR is the variability normalised with the observed variability. The number of hourly SST observations is
2040, while the number of daily cycles is 85.

Exp Bias DSA SD DSA SST Bias SD SST VAR

z0 = 0.5HS,ws +0.241 0.25 +0.078 0.19 1.32
z0 = 1/2kS +0.153 0.20 +0.046 0.16 1.16
z0 = 0.5HS +0.022 0.18 +0.009 0.12 1.00

No Langmuir +0.029 0.18 +0.011 0.12 1.00

cM = 0 +0.328 0.34 +0.110 0.24 1.41
No Waves +0.053 0.25 -0.104 0.19 1.07
hα i = 34 +0.025 0.20 +0.008 0.13 1.01
hα i = 148 -0.172 0.26 -0.048 0.14 0.76

interesting area. In the western part high values of normalized wave dissipation are found, related to an active
Somali jet generating steep waves, while in the eastern part of the Arabian Sea the sea state is much more gentle
giving low values of α . The buoy used in the present simulation exercize was just located at the border of high
and low values of normalized dissipation. Note that the Somali jet is intrinsically linked to the the onset of the
Asian monsoon as it supplies the necessary moisture.

6 Conclusions.

The main purpose of this paper was to investigate the role of wave dissipation (e.g. wave breaking) and Lang-
muir turbulence on the mixing of the upper ocean. As an interesting first application the impact of ocean waves
dynamics on the simulation of the diurnal cycle in SST was studied. The wave effects were studied in the con-
text of the Mellor-Yamada (1982) scheme where the TKE equation was extended to allow for effects of wave
dissipation and following Grant and Belcher (2009) effects of Langmuir turbulence. Following Janssen et al.
(2004) effects of wave dissipation on turbulence production in the ocean column were incorporated by mod-
elling the wave-induced energy flux δ pδ w while wave dissipation also affects the ocean momentum through
the wave-induced stress. Particular attention was paid to modelling of stratification effects on the turbulent
exchange coefficients for momentum, heat and turbulent kinetic energy, since, apart from solar insolation, the
main reason for the existence of the diurnal cycle is the reduction of the turbulent transport by buoyancy effects.
For low winds and strong solar forcing stratification in the ocean can become quite extreme, but unfortunately
observations under these extreme circumstances are rare. Therefore, at least in the atmospheric context, there
is no consensus on how the turbulent exchange coefficients behave in strongly stable conditions. First, it may
be argued that turbulent motion is damped when the gradient Richardson number exceeds a critical value, say
of the order of 1/4. A prominent example of this approach is the Mellor-Yamada scheme (1982). Many others
argue, on the other hand, that beyond the critical Richardson number there is still transport possible related to
internal gravity waves and intermittency. Presently, two directions may then be distinguished. One approach,
which is based on observations, direct numerical simulations and group normalization methods argues that due
to internal wave activity and intermittency momentum transfer will be more efficient than heat transport, while,
on the other hand, from the SHEBA data there is compelling evidence that the opposite is true. A choice has
therefore to be made, and in the main text arguments have been presented why I have chosen for the option of
a more efficient momentum transport for the strongly stable case. At the same time, for the weakly stable case
the proposed model for stratification agrees with the Kansas field experiment.

32 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

Properties of the resulting model for mixing in the upper ocean have been studied extensively. Under neutral
circumstances it can be shown that the turbulent velocity may be obtained from a ’1/3’-rule (see Eq. (44). This
rule is important in understanding the sensitivity of upper ocean transport to variability in the sea state. When
determining the energy flux from dissipating waves it is found that there is high variability in the dimensionless
flux α in particular near the passage of a front (see Fig. 1). As the turbulent velocity depends, according to
the ’1/3’-rule only on α 1/3 , its variability, and the variability in the turbulent transport, is much reduced. The
’1/3’-rule also explains that when only Langmuir turbulence is taking into account the turbulent velocity scales
with La−2/3 in agreement with the scaling arguments of Grant and Belcher (2009).
Results from a simulation with the present mixed layer model of the diurnal cycle in SST over a three month
period for a location in the Arabian Sea are compared with in-situ observations and judged by statistical pa-
rameters such as bias, standard deviation and simulated variability there is an excellent agreement. It has also
been shown that, as expected, results depend in a sensitive manner on the way stratification is modelled. For
example, neglect of the contribution to turbulent transport by intermittency and internal gravity waves gives a
large increase in error. In a similar spirit, it can be shown that wave effects play an important role in the mixing
in the upper ocean. No sensitivity to Langmuir turbulence was found in the simulation results for the diurnal
cycle, presumably because on average the maximum of the Langmuir production term was at a larger depth
than the depth where the boundary condition for ocean temperature was given.
Nevertheless, the model still needs to be validated more extensively against satellite observations from geosta-
tionary satellites and polar orbiters. This work is left for the future.

Acknowledgements. Discussions with Yuhei Takaya and Anton Beljaars are greatly appreciated, while Jean
Bidlot provided me with the sea state parameters from the ERA-interim reanalysis. I would also like to thank
the Woods Hole Oceanographic Institution for making available the in-situ observations which were of great
value in this model development.

Technical Memorandum No. 634 33

 Ocean Wave effects on the daily cycle in SST

7 References.

Abramowitz, M. and I.A. Stegun, 1964. Handbook of Mathematical Functions, Dover Publications, Inc., New
York.
Anis, A. and J.N. Moun, 1995. Surface wave-turbulence interactions: Scaling ε (z) near the sea surface. J.
Phys. Oceanogr. 25, 2025-2045.
Baas, P., S.R. de Roode and G. Lenderink, 2008. The Scaling Behaviour of a Turbulent Kinetic Energy Closure
Model for Stably Stratified Conditions. Boundary-Layer Meteorol. 127, 17-36.
Banner, M.L., 1990b. Equilibrium spectra of wind waves. J. Phys. Oceanogr. 20, 966-984.
Baumgartner, M.F., N.J. Brink, W.M. Ostrom, R.P. Trask and R.A. Weller, 1997. Arabian Sea Mixed Layer
Dynamics Experiment data report. Woods Hole Oceanographic Institution Technical report WHOI-97-08, UOP
Technical Report 97-03, 157 pp.
Bidlot, J.-R., D.J. Holmes, P.A. Wittmann, R. Lalbeharry and H.S. Chen, 2002. Intercomparison of the Per-
formance of Operational Ocean Wave Forecasting Systems with Buoy Data. Weather and Forecasting 17,
287-310.
Birch, K.G. and J.A. Ewing, 1986. Observations of wind waves on a reservoir, IOS-rep. No. 234, Wormley,
37p.
Britter, R.E., J.C.R. Hunt, G.L. Marsh and W.H. Snyder, 1983. The effects of stable stratification on turbulent
diffusion and the decay of grid turbulence. J. Fluid Mech. 127, 27-44.
Burchard, H., 2001. Simulating the Wave-Enhanced Layer under Breaking Surface Waves with Two-Equation
Turbulence Models. J. Phys. Oceanogr. 31, 3133-3145.
Businger, J.A., J.C. Wyngaard, Y. Izumi and E.F. Bradley, 1971. Flux-Profile relationships in the Atmospheric
Surface Layer. J. Atmos. Sci. 28, 181-189.
Burgers, G.J.H., P.A.E.M. Janssen, and D.L.T. Anderson, 1995. Impact of sea-state dependent fluxes on the
tropical ocean circulation. International Scientific Conference on the Tropical Ocean’s Global Atmosphere
(TOGAS), 2-7 April 1995, Melbourne, 295-297.
Cheng, Y. and W. Brutsaert, 2005. Flux-profile relationships for wind speed and temperature in the stable
atmospheric boundary layer, Boundary-Layer Meteorol. 114, 519-538.
Craig, P.D., 1996. Velocity profiles and surface roughness under breaking waves. J. Geophys. Res. 101,
1265-1277.
Craig, P.D. and M.L. Banner, 1994. Modeling wave-enhanced turbulence in the ocean surface layer. J. Phys.
Oceanogr. 24, 2546-2559.
Csanady, G.T., 1964. Turbulent diffusion in a stratified fluid. J. Atmos. Sci. 21, 439-447.
Deardorff, J.W., 1980. Stratocumulus-capped mixed layers derived from a three-dimensional model. Boundary-
Layer Meteorol. 18 495-527.
Drennan, W.M., M.A. Donelan, E.A. Terray and K.B. Katsaros, 1996. Ocean turbulence dissipation rate mea-
surements in SWADE. J. Phys. Oceanogr. 26, 808-815.
Fairall, C.W., E.F. Bradley, D.P. Rogers, J.B. Edson and G.S. Young, 1996. Bulk parametrization of air-sea

34 Technical Memorandum No. 634

Ocean Wave effects on the daily cycle in SST

fluxes for Tropical Ocean-Global Atmosphere Coupled-Ocean Atmosphere Response Experiment. J. Geophys.
Res. 101, 3747-3764.
Grachev A.A., E.L. Andreas, C.W. Fairall, P.S. Guest and P.O.G. Persson, (2007a). SHEBA flux-profile rela-
tionships in the stable atmospheric boundary layer. Boundary-Layer Meteorol. 124 315-333.
Grachev A.A., E.L. Andreas, C.W. Fairall, P.S. Guest and P.O.G. Persson, (2007b). On the turbulent Prandtl
number in the stable atmospheric boundary layer. Boundary-Layer Meteorol. 125 329-341.
Grant A.L.M. and S.E. Belcher, 2009. Characteristics of Langmuir Turbulence in the Ocean Mixed layer, J.
Phys. Oceanogr. 39, 1871-1887.
Hara T, and A.V. Karachintsev, 2003. Observation of Nonlinear Effects in Ocean Surface Wave frequency
Spectra. J. Phys. Oceanogr. 33, 422-430.
Hasselmann, K., T.P. Barnett, E. Bouws, H. Carlson, D.E. Cartwright, K. Enke, J.A. Ewing, H. Gienapp, D.E.
Hasselmann, P. Kruseman, A. Meerburg, P. Müller, D.J. Olbers, K. Richter, W. Sell and H. Walden, 1973.
Measurements of wind-wave growth and swell decay during the Joint North Sea Wave Project (JONSWAP),
Dtsch. Hydrogr. Z. Suppl. A 8(12), 95p.
Janssen, P.A.E.M., 1999. On the effect of ocean waves on the kinetic energy balance and consequences for the
inertial dissipation technique. J. Phys. Oceanogr. 29, 530-534.
Janssen, P.A.E.M., O. Saetra, C. Wettre, H. Hersbach and J. Bidlot, 2004. Impact of the sea state on the
atmosphere and ocean. Annales Hydrographiques 6e série, 3 (772), 31323.
Kim J. and L. Mahrt (1992) Simple formulation of turbulent mixing in the stable free atmosphere and nocturnal
boundary layer. Tellus 44, 381-394.
Komen, G.J., 1987. Energy and momentum fluxes through the sea surface. Dynamics of the Ocean Surface
Mixed Layer, P. Müller and D. Henderson, Eds., Hawaii Institute of Geophysics Special Publications, 207-217.
Komen, G.J., L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann, and P.A.E.M. Janssen, 1994: Dynamics
and Modelling of Ocean waves (Cambridge University Press, Cambridge)
Kondo J., O. Kanechika, N. Yasuda, 1978. Heat and momentum transfers under strong stability in the atmo-
spheric surface layer. J. Atmos. Sci. 35, 1012-1021.
Kondo, J., Y. Sasano and T. Ishii, 1979. On Wind-Driven Current and Temperature Profiles with Diurnal Period
in the Oceanic Planetary Boundary Layer. J. Phys. Oceanogr. 9, 360-372.
Large, W.G., J.C. McWilliams, and S.C. Doney, 1994. Oceanic vertical mixing: A review and a model with a
nonlocal boundary parameterization. Rev. Geophys. 32, 363-403.
Mellor, G.L. and T. Yamada, 1982. Development of a turbulence closure model for geophysical fluid problems.
Rev. Geophys. Space Phys. 20, 851-875.
Nieuwstadt, F.T.M., 1984. The turbulent structure of the stable, nocturnal boundary layer. J. Atmos. Sci. 41
2202-2216.
Noh, Y. and H.J. Kim, 1999. Simulations of temperature and turbulence structure of the oceanic boundary layer
with the improved near-surface process. J. Geophys. Res. C104, 15,621-15,633.
Panofsky, H.A., 1963. Determination of stress from wind and temperature measurements. Quart. J. Roy.
Meteor. Soc., 89, 85-94.
Pacanowski, R.C. and S.G.H. Philander, 1981. Parameterization of Vertical Mixing in Numerical Models of

Technical Memorandum No. 634 35

 Ocean Wave effects on the daily cycle in SST

Tropical Oceans. J. Phys. Oceanogr. 11, 1443-1451.

Phillips, O.M., 1977. The dynamics of the upper ocean, Cambridge University Press, Cambridge, 336p.
Plant, W.J., 1982. A relation between wind stress and wave slope. J. Geophys. Res. C87, 1961-1967.
Shih, L.H., J.R. Koseff, J.H. Ferziger and C.R. Rehmann, 2000. Scaling and parameterization of stratified
homogeneous turbulent shear flow. J. Fluid Mech. 412, 1.
Simmons, A., S. Uppala, D. Dee and S. Kobayashi, 2007. ERA-Interim: New ECMWF reanalysis products
from 1989 onwards. ECMWF Newsletter 110, pp. 25-35. European Centre for Medium-Range Weather Fore-
casts, Reading, U.K.
Soloviev, A.V., 1982. On the vertical structure of the ocean thin surface layer at light wind. Dokl. Acad. Sci.
USSR, Earth Sci. Ser., Engl. Transl. 18, 751-760.
Strang, E.J. and H.J.S. Fernando, 2001. Vertical mixing and transports through a stratified shear layer. J. Phys.
Oceanogr. 31, 2026-2048.
Sukoriansky, S., B. Galperin and I. Staroselsky, 2005. A quasinormal scale elimination model of turbulent
flows with stable stratification. Phys. Fluids 17 085107-1-28.
Terray, E.A., M.A. Donelan, Y.C. Agrawal, W.M. Drennan, K.K. Kahma, A.J. Williams, P.A. Hwang, S.A., and
Kitaigorodskii, 1996. Estimates of Kinetic Energy Dissipation under Breaking Waves. J. Phys. Oceanogr. 26,
792-807.
Terray, E.A., W.M. Drennan, and M.A. Donelan, 1999. The vertical structure of shear and dissipation in
the ocean surface layer. The wind-driven air-sea interface, M.L. Banner, Ed., School of Mathematics, The
University of New South Wales, Sydney, 239-245.
Weber, S.L., 1994. Statistics of the air-sea fluxes of Momentum and Mechanical energy in a Coupled wave-
atmosphere model. J. Phys. Oceanogr. 24, 1388-1398.
Weller, R.A., A.S. Fisher, D.L. Rudnick, C.C. Eriksen, T.D. Dickey, J. Marra, C. Fox and R. Leben, 2002.
Moored observations of upper-ocean response to the monsoons in the Arabian Sea during 19941995, Deep Sea
Res., Part II, 49, 2195-2230.
Wyngaard, J.C., 1985. Structure of the planetary boundary layer and implications for its modeling. J. Clim.
Appl. Meterorol. 24, 1131-1142.
Zilitinkevich, S.S., T. Elperin, N. Kleeorin and I. Rogachevskii, 2007. Energy- and flux-budget (EFB) turbu-
lence closure model for stably stratified flows. Part I: steady-state, homogeneous regimes. Boundary-Layer
Meteorol. 125 167-191.

36 Technical Memorandum No. 634