You are on page 1of 23

Notes on stratified flows

Contents
1 Wakes in a stratified turbulent flow 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Effect of stratification on the wakes . . . . . . . . . . . . . . . . . . . 3
1.3 Different stages in the lifecyle of wakes in stratified flows: [1] . . . . 7
1.4 DNS of stratified wake flows: [2] . . . . . . . . . . . . . . . . . . . . . 10

2 A Kolmogorov-like turbulence for the strongly-stratified flows 11


2.1 Strong-stratification hypothesis . . . . . . . . . . . . . . . . . . . . . 11
2.2 Geophysical interpretation . . . . . . . . . . . . . . . . . . . . . . . . 15

3 DNS of stratified wake turbulence - Large scale characteristics [3] 17


3.1 Wakes from a slender-body [5] . . . . . . . . . . . . . . . . . . . . . 18

1 Wakes in a stratified turbulent flow


1.1 Introduction
A wake is defined to be the non-propagating disturbance produced by a moving
body. Wakes have been of interest to many fluid dynamicists and engineers
because they are a basic flow phenomenon associated with fluid flowing over an
obstacle [6]. Wakes are associated with ships, submarines, aircrafts, unmanned aerial
vehicles, wind turbines, marine swimmers, aerial flyers, windflow over islands and
mountains [2].
The term stratified turbulence was introduced by Lilly (1983) to describe the
dynamics of flows dominated by stable density stratification. Such flows can consist
of internal waves, but also of quasi-horizontal, meandering motions; they possess
potential vorticity and are strongly nonlinear. It is believed that several seemingly
anomalous ocean and atmospheric data can be explained by understanding stratified
turbulence [7]. In oceans and atmosphere, flows at various scales are commonly
described in terms of the dimensionless parameters:

UL U U
Re = , Fr = , Ro = (1.1)
ν NL fL

1
fig:wake.png

Figure 1: Schematic of wake behind a sphere

which describes the balance of time-scales1 between the inertial forces of the fluid
U
( ) and other competing forces, namely:
L
ν
1. For the Reynolds number, Re it is the viscous forcing-timescale 2 .
L
2. For the Froude number, Fr it is buoyancy force given from the Brunt-Väisälä
frequency N 2 = − ρg0 dρ
dz
.

3. For the Rossby number, Ro it is the Coriolis force frequency f = 2Ω sin(φ) at


lattitude φ and planetary rotation rate Ω.

The principal requirement for the existence of stratified turbulence is that the local
Froude number satisfies Fr ≪ 1. Despite our understanding, a general theory of
wake pattern formation is lacking and would have to span many orders of magnitude
in Reynolds number [8].
The evolution of the turbulent wake of a towed sphere in a stable background
density gradient is a convenient experimental model and numerical model for
investigating fundamental problems in the decay of isolated patches of turbulence
in stratified oceans or atmospheres, or in application to more practical problems of
turbulent wake generation and detection for slender bodies of various kinds [1].
1
It is the ratio of frequencies, to be exact
1.2 Effect of stratification on the wakes
A well documented understanding of wakes created by towed sphere on a
stratified fluid via numerical experiments is given in the Refs. [1, 9, 10] by Spedding
et. al.,

1.2.1 Qualitative picture of what difference does stratification induce on


the wakes.
In a stratified fluid, vertical motion causes vertical transport of heat or mass,
which in turn converts kinetic energy to potential energy. As a result, vertical motion
is suppressed and the vertical heat or mass transport becomes weakened [6]. With
the turbulent mixing converting kinetic energy to potential energy the stratified
fluid would decay faster than in a non-stratified fluid. In a stratified fluid, the
turbulent wake cannot expand indefinitely in the vertical direction as it would in
a non stratified fluid. This inhibited growth of the vertical height is called wake
collapse. Consequently, the horizontal width of the stratified wake expands faster
than that of the non stratified wake. Associated with the wake collapse, internal
waves are also generated [6].
After the wake collapses, the mean motion in the vertical direction diminishes
and the horizontal motion dominates the wake. As a result, the wake flow tends
to be quasi-two-dimensional. Since a two-dimensional wake flow is very unstable,
the wake could therefore meander horizontally and break into horizontal vortices.
The horizontal vortices look like pancakes as the ratio of their vertical dimension to
horizontal dimension is about 1 to 5 [6].
The stand-out feature of the wakes in stratified environment is the presence of
stable patches of vertical vorticity having large horizontal length scales compared
with any vertical structure, and its long life times.

1.2.2 Defining turbulence in the wakes


How to define turbulent wake energy, enstropy and dissipation?
The boundary of the wake has to be assumed (reasonably) first. Here the body will
move, so the free-stream velocity si 0. Ux (y) is a velocity of the wake averaged over
a small portion over the whole wake length ∆x ≪ X. A Gaussian fit can be made
fig:wake_width.png

Figure 2: Schematic of wake measurement

for the averaged streamwise velocity Ux (y) along y as2


1 y
 2 !
Ux (y) = U0 exp − (1.2)
2 Lσ
where Lσ is the measure of wake half-width L1/2 , and U0 is the maximum of Ux over
y, the centerline velocity deficit. The wake region can be defined as the horizontal
band in y satisfying Ux (y) > 0.2U0 [9]. The wake width from this definition will be
denoted ad Lw .
The wake averaged values, that we are interested are
1. The inner wake kinetic energy is
1D 2 E
E= u + v2 .
2
2. Vertical vorticity and mean enstrophy are
1 D 2E
ωz = ∂x v − ∂y u, W = ω
2 z
3. Strain on the horizontal plane (or dissipation) as
S = ⟨ϵL ⟩ = 2ν ⟨sij sij ⟩ , i = 1, 2.

2
The flow statistics are not axisymmetric, unlike the unstratified case. Hence cartesian variables
are used.
Ozmidov scale:
The largest vertical size attainable by a turbulent eddy is of the order of Ozmidov
!1/2
u3
scale, with lO = with u the fluctuating velocity at integral scale l.
lN 3
(Will disuss this scale later.)

1.2.3 Long time evolution of the initally turbulent wake in stable


stratification [9] and similarity scaling of it [10]
Following results are from Spedding et. al., [9, 10]. Now to study the scaling in
general, a functional form can be defined as
 B1
L1/2 x
∼B0
D D
 C1
U0 x
∼C0 (1.3)
U D

Inferences: Experimental results found for Re ∼ [5E3, 1E4], Fr ∈ [1, 15]:

• B1 was ≈ 0.36, close to the classical value of 1/3.

• But C1 was ≈ −0.88, larger than the classical value of −2/3.

• The proportional factor C0 ≈ 4.2 implies a significantly higher center-line


velocity, due to the persistence of the coherent structures (see Fig. 3).

For small Reynolds number ∼ 1E3, the wake width increases faster (almost like
a 2-D case), that is B1 ∼ 0.5 with U0 scaling remaining the same.
Later in 2002, Bonnier and Eiff also experimentally shown an accelerated
collapse phase where there is an abrupt increase in the centerline velocity defect
U0 , at N t ≈ 3.
fig:vel_width_scaling.png

Figure 3: Wake width and defect velocity, normalized. Symbols are from
stratified measurements, thin dotted lines are least-squares fit, and other lines
are from unstratified measurements [10].

One can make the usual equivalence between observations that vary over time
in a stationary laboratory frame of reference and an effective wake length in x as
U t = x. Then in parameters that involve stratification or buoyancy (with Fr = N2UD ),
we can write
x Fr
= Nt (1.4)
D 2
The classical prediction for the scaling of the non-dimensionalized turbulent
quantities are

E 1/2 x −2/3 q 2/3


 
≈ or Fr ∼ (N t)−2/3 (1.5)
U D U0
W 1/2
 −1
x ω
U/D
≈ or Fr ∼ (N t)−1 (1.6)
D U/D
 −7/3
S x ⟨ϵ⟩ 7/3
3
U /D
≈ or 3 Fr ∼ (N t)−7/3 (1.7)
D U /D
Similarly the turbulence quantities E, W, S surprisingly scale almost as per classical
predictions, except for the S, which scales slower than the expected x−7/3 .
In the late wakes, with q = {u, v}, the magnitudes ωz = ∇ × q and ∆z = ∇ · q
can be regarded as approximately proportional to the amplitudes of vortex and wave
motions (internal gravity waves), respectively, in a plane of constant z. Wave motions
have timescales of O(N−1 ) and z−amplitude of O(1), whereas turbulent motions
have O(N −1 /Fr) and O Fr2 respectively.
Reports by Lin and Pao (1979) [11], on experiments before 1979 on the wake
studies. The inference is that the size of the wakes grow with t1/3 law (classical
theory) until N t ≈ 2, and then for later times it increases as t1/2 . Note these were
based on shadowgraph visualizations.

1.3 Different stages in the lifecyle of wakes in stratified flows:


[1]
Claim of Universality:
It was particularly surprising to see that variations in Fr were unimportant in
determining the wake properties, but the reason is because all wakes, no matter
how energetic and turbulent their initial starting point, will eventually become
dominated by the background stratification, and now the local, effective Fr is
very small.

An important question would be whether coherent vortex wakes are always


formed even at high Froude number. If they are not, then a critical value of Fr might
be identifiable, separating regimes that generate persistent wakes from those that
do not [1]. The study by Spedding et al., in Ref. [1] concluded that the unstratified
result will surely be recovered as Fr → ∞, is therefore false, because all flows at
finite Fr are different from the particular case where F = ∞.

Spedding [1] raised two important questions following the above inferences.

1. If scaling is same for stratified and unstratified, then how come there is
an order of magnitude difference in the U0 ?

2. If all results in the long-times are independent of Fr, is there a Frcr.


where we recover unstratified classical turbulent sphere wake?

Experiments in the 90ś (by Hopfinger, 1991; Sysoeva & Chashechkin, 1991; Lin,
1992; Chomaz, Bonneton & Hopfinger, 1993;) led to a qualitative observation of
fig:vel_x_NEQ.png

Figure 4: Scaling showing the existence of the intermediate regime [1].

wakes in stratified turbulent flows, illustrating the pancake structures. Spedding,


Browand & Fincham, 1996, showed that the late-wake mean and turbulence profiles
decayed with exponents that were similar to those found for turbulent sphere wakes in
unstratified homogeneous fluids [1]. Further, the energetic mean wake flow had higher
degree of order, coherent structure and organization as compared to the homogeneous
case.
So with improved experimental techniques, Spedding was able to observe and
record the wake properties at an even initial stage. The wake-averaged half-width,
L1/2 , shows a 1/3 power law increase x/D = 100 → 10000, for all values of Fr from
10 to 240. The downstream evolution of the mean centreline velocity, by contrast,
differs significantly from previous results, and shows a systematic variation with F.
There lies an intermediate range with the scaling exponent for U0 with decay rates
∼ (x/D)−0.25±0.05 , which is much slower.
The intermediate regime is extrapolated back (in (x/D)) to intersect with the
three-dimensional result, namely xI (Fr) and similarly forward to intersect with the
far-wake results obtained in [10], calle it xI I(Fr). When the positions xI , xII are
converted to the non-dimensional times, the observations was

N tI = 1.7 ± 0.3, N tII = 50 ± 15 (1.8)

Now if we observe the turbulence quantities E, W, S in the early regime, both E 1/2
and W 1/2 show lower initial rates until N tII = 50, which is consistent with U0 scaling.
By contrast, the wake-averaged dissipation rate, S, shows no such breakpoint, and
continues with more or less the same power-law behaviour, right through the critical
value of N tII .
1.3.1 Universal characteristics of stratified wakes
Spedding in Ref. [1] divided the wake life cycle into three regimes:

1. A three-dimensional regime for N t ∈ [0, 2] where the scaling of the


wave-averaged half-width shows a power law of L1/2 ∼ x1/3 . Similarly the
velocity defect scales as x−2/3 for small distances. Then the stratification effect
comes into picture, changing the power-law behaviour. Larger the Froude
number, longer the time it takes for the stratification to affect the flow. But
when rescaled appropriately, it collapses.
U0 2/3
Fr ∼ (N t)−2/3
U

2. A non-equilibrium (NEQ) regime for N t ∈ [2, 50] where U0 ∼ t−1/4 . Physically,


it corresponds to an adjustment, or non-equilibrium (NEQ) period, when
potential energy of vertically displaced fluid is partly reconverted back to
kinetic energy close to the wake centreline. It is also when the turbulent kinetic
energy is not sufficient to overcome the buoyancy force, and some that was
able to in the initial stages gets stratified and gains momentum back. The
low decay rates in the NEQ regime derive from the conversion of potential
to kinetic energy, and will be associated with negative buoyancy flux as the
turbulence collapses, and restratification occurs. At high Re, the NEQ regime
has enhanced turbulence owing to the secondary Kelvin-Helmholtz instabilities.
By quantification of the mean wake energetics, Brucker & Sarkar (2010) found
that, owing to the buoyancy-induced reduction of Reynolds shear stress of the
vertical velocity, the turbulent production responsible for the transfer of energy
from mean to fluctuations is reduced, thus leading to defect velocity in the NEQ
regime that is substantially higher than in the unstratified wake.

3. A quasi two-dimensional regime for N t ∈ [50, ∞] where U0 ∼ t−0.76 . At


sufficiently large times, or sufficiently far downstream of the original turbulent
source, all freely decaying stratified turbulent flows, with arbitrarily large initial
Fr, eventually reach a state where the local Frl is small, and the dynamics are
dominated by buoyancy effects. The final asymptotic state is the Q2D phase in
Fig. 5. The answer to whether the unstratified result will surely be recovered
as F → ∞ is therefore no!, because all flows at finite Fr are different from the
particular case where truly Fr → ∞.
Although numerical simulations (as in Ref. [12]) show a prolonged NEQ regime
when compared to the simulations done by Spedding, owing to the physical space
limitations of the laboratory. Numerical simulations have the capability to study the
Reynolds number variation on the internal waves emitted by the wake turbulence,
turbulent and non-turbulent interfaces [3]. At larger Reynolds number, secondary
Kelvin-Helmholtz instabilities starts to appear in the NEQ regime [12]. The
structures that are of interest are the ones that attains an order unity local Froude
number. These behave in a quasi two-dimensional fashion, constrained by the
ambient stratification. When the initial conditions are fully turbulent, the transition
between a disorganized three-dimensional flow and the final bouyancy-dominated
state has not yielded to simple analysis [1].

1.4 DNS of stratified wake flows: [2]


Numerical simulations that have been performed in the past two decades to study
stratified sphere wakes have primarily used the temporally evolving model, wherein
periodic boundary conditions are used in the streamwise direction and the wake
statistics are a function of time. The simulation is initialized with a synthetic
velocity field that satisfies assumed profiles for the mean and variance. These
profiles are typically taken to approximate unstratified wake measurements a few
body diameters downstream of the sphere. In contrast, the spatially evolving model
allows streamwise evolution of the flow statistics and thus has the ability to include
the body and near-wake dynamics.
A temporal approximation of the initial fluctuations requires knowledge from
the near wake region flows, to show the NEQ transition. Typical initial conditions

fig:lifecycle_wakes.png

Figure 5: Universal characterization of different lifecycles of the wake [1].


used in unstratified flows will not work. The temporal model is inexpensive, as the
boundary layer and its separation is not modelled or resolved.
In body inclusive, spatial simulations curvilinear grids has to be employed.
Cylindrical co-ordinates with z pointed along the streamwise direction. The sphere
inside the domain is represented by an immersed boundary method.

2 A Kolmogorov-like turbulence for the strongly-stratified


flows
It has been observed that the horizontal wavenumber spectra for the kinetic and
−5/3
potential energy measured in the upper atmosphere follows a E(kh ) ∼ kh , similar
to the isotropic turbulence and the vertical spectra to show a ∼ kv−3 behaviour (see
Fig. ??). It has often been debated and speculated about the direction and nature
of cascade that can explain the observations.
While some arguments are along the lines for an existence of an inverse-cascade
of energy (like in 2D turbulence). Much of which comes mainly from the fact that
the potential vorticity (Π = ω · (∇φ + N êz )) integrated over the domain is conserved
in the inviscid limit of the Boussinesq equation. While to a large extent, it is
believed and advocated that there is a forward cascade of energy in the stratified high
Reynolds number turbulence. The latter sticks with an argument that as N → 0, we
get a homogeneous turbulence with a passive scalar, still possessing the conservation
of the potential vorticity.

2.1 Strong-stratification hypothesis


The Boussinesq equations in a non-rotating frame of reference for the stratified
flow can be written as

∂t u + u · ∇u = −∇p + ν∇2 u + N êz (2.1)


gT ′
∂t φ + u · ∇φ = κ∇2 φ − N u · êz , ∇ · u = 0, φ= . (2.2)
N T0
Based on this equation, in Ref. [13] a forward cascade similarity hypothesis is
developed. Starting with a statistically stationary homogeneous system obtained by
injection of energy at integral scales, say at the rate of P . Denoting the mean kinetic
and potential energy as EK = 21 ⟨u · u⟩ and EP = 12 ⟨φ2 ⟩ respectively, whose evolution
fig:atmospheric_spectrum.png

can be derived from the equations (2.2).

P = −ϵK + N ⟨φw⟩ (2.3)


0 = −ϵP − N ⟨φw⟩ (2.4)

Here ϵK &ϵP are the dissipation rates of kinetic and potential energies respectively.

Billant & Chomaz (2001) have shown that there is a fundamental scale invariance
in the Boussinesq equations in the limit of strong stratification, and that this
invariance implies an approximate equipartition of energy between kinetic and
potential energy [13].

As a starting point, assume that there is an approximate equipartition in the


fully-developed system [13], that is u ∼ φ, EP ∼ EK , ϵP ∼ ϵK . Adopting the
classical estimates from Taylor (1935), u ∼ (lh ϵ)1/3 , EK ∼ (lh ϵ)2/3 . With the
continuity equation, the active component of the velocity field will scale as w ∼
ulv /lh . If an estimate of a vertical lengthscale is to be provided for the system with
parameters {ε, lh , N } based on the existence of equipartition, it would be the Froude
number, Fr.
lv ϵ1/3
∼ 2/3
∼ Frh (2.5)
lh N lh
The horizontal Froude number is the ratio of vertical to the horizontal lengthscales.
Now the how small this vertical scale can be, for this hypothesis to hold is the
question. Independent of the large scale, a small-scale devised only from ϵ, N (based
on dimensional analysis) is the Ozmidov scale lO .

1/2
ϵK

lO = (2.6)
N3

And the ratio of it to the verticcal scale is


lv −1/2
∼ Frh .
lO
This relation determines that the smallest scale for the strongly stratified hypothesis
to hold is the lO . In other words, it is the largest scale having sufficient energy to
overturn the buoyancy effects.
The nonlinear inertial forces are what causes the shear instability, that breaks
the larger structures of order l into small scales of motion, inducing a forward
energy cascade. Although the dynamics of this energy cascade are controlled
by the strong stratification, the process has some important features in common
with the three-dimensional Kolmogorov cascade. The ratio between inertial and
buoyancy forces is measured by the vertical Froude number Frv = u/(N lv ). For
large enough Reynolds numbers, strongly stratified flows can instantaneously develop
strong vertical shearing of the horizontal velocity, leading to vertical differential
length scales of order lv ∼ Nu . This shearing allows a local vertical Froude number to
be of order unity. Eventually at scales much smaller than lO the strongly stratified
turbulent cascade can no longer prevail, and there will be a transition to more
classical three-dimensional turbulence.
Therefore, with a strong separation between the vertical and horizontal length
scales, that is lh ≫ lv , the horizontal energy spectrum will be dependent on kh only
and the vertical will be on kv only.
−5/3
E(kh ) ∼ ϵ2/3 kh (2.7)
E(kv ) ∼ N 2 kv−3 (2.8)
The self-similar form given in the equation (2.8) will be obeyed in the range of
1
≪ kh ≪ khmax (2.9)
lh
1
≪ kv ≪ kvmax (2.10)
lv
1/2
The maximum vertical wavenumber kvmax will be ∼ Frh,crit. lO
−1
whereas for the
3/2
horizontal case khmax ∼ Frh,crit lO
−1
.

2.1.1 Order of magnitude estimates


To test the hypothesis, a realistic parameter value estimate is given below. With
the choice of

lh = 500km, lv = 1km (2.11)


N = 2.0 × 10−2 s−1 , T0 = 225K (2.12)
(2.13)

and dissipation value of ϵ = 8.0 × 10−5 m2 s−3 from airplane data from Lindborg
(2001) we find the horizontal Froude number to be close enough to our estimate.
lv
Frh = 3.6 × 10−4 ≈ 0.2
lh
The typical velocity and temperature fluctuations are

u ∼ (ϵlh )1/3 ≈ 3m s−1 , T ′ = To (ϵlh )1/3 N/g ≈ 1K

for the eddy turnover time of τ ∼ 40h, which all seem reasonable. The Ozmidov
scale for this values is lO ≈ 10m.

2.1.2 Constraints in the simulations


To simulate a low Fr number cascade in a periodic box (which could be larger
in the horizontal and smaller in vertical direction respectively) and resolve all the
necessary scales a scaling of the computational power with respect to the Re and Fr
is needed. If Lx = Ly denotes the horizontal sides and Lz the vertical side, then
Lz
≈ Frh
Lx
Also to resolve the viscous scales
lv
≫1
ηv
where ηv is the vertical diffusion scale. We know the classical relation that lh /η ∼
3/4
Reh , with η is the Kolmogorov scale. If nh points are needed in the horizontal
direction, then nv ∼ nh Frh is needed in the vertical direction. Hence the total
9/4
number of computational grid points scale as n3 hFrh ∼ Reh Frh .

Since Frh < 1, it might seem like the computation had just became easier. But,
NO!.
Remember, the lh ∼ 100km which is huge compared to the scales involved in the
homogeneous DNS. To put it in perspective, consider writing in terms of vertical
−5/4
Reynolds number, Rev = Reh Frh then the no of grid points become Rev9/4 Frh .
It is almost impossible to simulate a physically realistic values in DNS. The
get-around in this case is to modify the diffustion terms in the Navier-Stokes
equation. With hyperviscous terms, ∇8 instead of ∇2 narrows the dissipation
wavenumber band. Also νh > νv allows us to use coarser grids in the horizontal
scale than the vertical scale.

2.2 Geophysical interpretation


As we saw in the previous section, the Ozmidov scale marks the transition
between those scales of motion that are strongly influenced by stratification and
those that are not. Such transition is witnessed in certain geophysical observations,
like Gargett et al. (1981). If we look at the vertical shear spectrum, Φ(kv ) = kv2 E(kv )
it takes a minimum around kO ∼ 1/lO , that marks a transition between those scales of
motion that are strongly influenced by stratification and those that are not. To the
right of the minimum the influence of stratification becomes asymptotically weaker
with increasing wavenumber and the spectrum approaches the spectrum of isotropic
Kolmogorov turbulence. The fact that Φ(kv ) ∼ kv− 1 for k ⩽ kO is also consistent
with the strong-stratification hypothesis there we expected a E(kv ) ∼ kv−3
fig:Ozmedov_separation.png

Given a wide range of scales present, the inertial range becomes analogous to the
K41 inertial range. The horizontal kinetic and potential energy spectrum are given
by
2/3 −5/3
EK (kh ) = CK ϵK kh
−2/3 −5/3
EP (kh ) = CP ϵK ϵP kh (2.14)

which are verified later by Lindborg. As a kind note, interestingly the choice of
−1/3
ϵK ϵP in the potential energies gives the constants CK , CP a more universal nature
when tested for different Fr, rather than simply using ϵ.
3 DNS of stratified wake turbulence - Large scale
characteristics [3]
An instructive way to classify the different regimes of stratified turbulence is to
look at the flow in the phase-space of (Reh , Fr−1
h ), h denotes the horizontal turbulent
motions. A buoyancy Reynolds number defined as

R = Reh Fr2h (3.1)

The two conditions for strongly stratified turbulence (SST) is that:


• Stratification is strong, that is Frh ≪ 1.
• Layered structures at the vertical scale U/N are not affected by viscosity, that
is
R≫1
.

fig:phase_diagram_stratification.png

Figure 6: Phase diagram in Reh and Fr−1


h .

There have been scaling arguments, and numerical evidence as to the existence
of a threshold value Rc , above which the flow regime can indeed be interpreted as
SST.
A significant feature of the SST flow is that the vertical turbulent Froude number
U
Frv =
N lv
becomes order unity, backed from numerical simulations denotes the entry to the
SST. The exit characterized by the event where the viscous effects start to have
−1/2
significant impact on the layered structure. This is when lv/lh ∼ Reh , reminiscent
of a laminar boundary layer scaling.

Numerical simulations by Zhou and Diamessis, involves implicit large-eddy


simulations (ILESs) using an incompressible Navier-Stokes solver based on
a spectral multidomain penalty method invoking Boussinesq approximation.
This solver employs Fourier discretizations in both horizontal directions x
(streamwise) and y (spanwise) and a Legendre-polynomial-based spectral
multidomain scheme in the vertical direction z. Spectral filtering and a penalty
scheme ensure the numerical stability of the simulations without resolving the
full spectrum of turbulent motions.

Inferences:
1. The emerging horizontal length scales grow in time lh/D ∼ (N t)1/2 .

2. Vertical Froud number crossing unity and testing of viscous scaling by lv


lh

−1/2
Reh .

3. Scaling of horizontal and vertical turbulent velocities and their decay.

4. Travel in the turbulent Reynolds-Froude number phase space.

3.1 Wakes from a slender-body [5]


As a canonical shear flow that can serve as a building block to study more complex
flows, turbulent wakes have been studied extensively in the past. However, an
overall wake decay theory is still missing and recent studies have shown a richer
and more complex problem than that portrayed by classical theory. The classic
solutions for the decay of the axisymmetric turbulent wake assume that the flow
evolves in a self-similar manner. They are based on the hypothesis that the profiles
of the velocities and the Reynolds stresses become invariant with respect to x (the
streamwise coordinate) when expressed in terms of local characteristic scales. The
self-similarity concept is often combined with the notion of universality. According
to the universality hypothesis, far from their generators, all wakes evolve into a
universal state, where the chaotic mixing of eddies has erased the memory of the
initial conditions. In the self-similar and universal state, an axisymmetric wake
obeys a single set of power-law exponents.

Ud ∼ x−2/3 , L ∼ x1/3 (3.2)

At low-Reynolds the laws were modified to

Ud ∼ x−1/2 , L ∼ x1/2 (3.3)

Despite the use of slender bodies in many engineering applications, the majority
of wake studies are devoted to the wake of bluff bodies. There are some
particularities that make the study of high-Reynolds-number slender-body wakes
especially challenging. Computationally, it is very costly to simulate turbulent
boundary layers over long bodies and also resolve their far wake. The problem is
tremendously stiff and the required spatial and temporal resolution is enormous. In
a wind tunnel or in a tank, if the body is long, a significant amount of the measuring
section can be taken by the wake generator. Furthermore, slender-body wakes are
thinner than bluff-body wakes and potentially harder to probe and measure. These
constraints have significantly limited the available studies on the topic.

Temporal numerical simulations: Instead of introducing the wake generator,


an initial condition is chosen to approximate the flow at some distance from
the body. The use of a temporal model reduces the computational cost very
significantly but requires a good initial condition to achieve a quantitative match
with experiments.

Non-equilibrium dissipation concept: The turbulent dissipation scales as


!n
Re k3/2/L
ϵ∼
ReL

where ReL = Us L/ν is the local Reynolds number. In the non-equilibrium decay,
Us ∼ x−1 (not the classical exponent −2/3) was seen for a flat plate.
Computationally, it is very costly to simulate turbulent boundary layers over long
bodies and also resolve their far wake. The problem is tremendously stiff and the
required spatial and temporal resolution is enormous.
Some of the questions that we address are as follows.
1. Does the decay of a slender-body wake at high Reynolds number conform to
classical theory?

2. How is the transition to self-similarity different from the wake of a bluff body?

3. Why have some previous experiments and simulations shown wake exponents
close to low-Re values even at very high Reynolds numbers?

4. Is there a possible relation between coherent structures and the wake decay?

Body-inclusive (BI) and Body-exclusive (BE) numerical methods.

Temporal evolution neglects streamwise flow evolution, thereby allowing the use
of periodic boundary conditions which significantly reduces the computational cost
of a simulation. The main drawback is that an approximation of initial fluctuations
(often taken from measurements of the unstratified downstream wake) is necessary
which, thus, may not be able to accurately capture buoyancy effects in the near
wake. This is BE type of numerical simulation. The grid cell in the BE simulation,
which has to be sufficiently small to adequately resolve wake turbulence, is still much
larger than that required to resolve the boundary layer. The flow is assumed periodic
in longitudinal direction and the initial conditions, for velocity and temperature,
correspond to mean profiles with superimposition of an ad hoc noise, issued from first
and second order statistics of some experiments. In this case, no coherent structures
are present in the flow at the initial time and so the validity of the temporal study
may be criticized [14].
An alternative is spatially-evolving simulations including the body that resolve
the boundary layer, flow separation and the near wake. This is BI type and there
is a limitation to their practicality. Because a large number of grid points are
required near the body to resolve the boundary layer and flow separation, the
simulations are computationally expensive and unable to extend far downstream
without prohibitive cost. The use of the BI simulation avoids the drawback of
regular temporal simulations, for which the choice of initial conditions introduces
considerable variability in the subsequent wake evolution.
Hybrid method:
Here, data from a body-inclusive spatially-evolving simulation was used to initialize
a separate temporally-evolving simulation. In this model, cross-sectional data planes
are extracted from a spatially-evolving body-inclusive simulation at some point
downstream of the body and over a time interval that starts after the flow has
achieved statistical steadystate. Data at a chosen downstream location are used as
inlet conditions in a spatially-evolving simulation without a body which has a coarser
grid than that of the body-inclusive simulation.

fig:hybrid_model.png

Figure 7: Illustration of the setup for the hybrid simulation showing a


three-dimensional view of the body-inclusive domain and planes at
x/D = 3, xD = 6, and x/D = 10 indicating the different starting point
choices for the hybrid model simulation domain. Image taken from [15]

The choice of inflow location requires careful consideration. x = 3D is not idea,


whereas x = 6 − 10D are apt choices [15].
References
[1] G. R. Spedding, Journal of Fluid Mechanics 337, 283 (1997), publisher:
Cambridge University Press.

[2] A. Pal, S. Sarkar, A. Posa, and E. Balaras, Journal of Fluid Mechanics 826, 5
(2017), publisher: Cambridge University Press.

[3] Q. Zhou and P. J. Diamessis, Phys. Rev. Fluids 4, 084802 (2019), publisher:
American Physical Society.

[4] A. M. Abdilghanie and P. J. Diamessis, Journal of Fluid Mechanics 720, 104


(2013), publisher: Cambridge University Press.

[5] J. L. Ortiz-Tarin, S. Nidhan, and S. Sarkar, Journal of Fluid Mechanics 918,


A30 (2021), publisher: Cambridge University Press.

[6] J. Lin and Y. Pao, Annual Review of Fluid Mechanics 11, 317 (2003).

[7] J. J. Riley and E. Lindborg, Journal of the Atmospheric Sciences 65, 2416
(2008), publisher: American Meteorological Society Section: Journal of the
Atmospheric Sciences.

[8] G. R. Spedding, Annual Review of Fluid Mechanics 46, 273 (2014), _eprint:
https://doi.org/10.1146/annurev-fluid-011212-140747.

[9] G. R. Spedding, F. K. Browand, and A. M. Fincham, Dynamics of Atmospheres


and Oceans Stratified flows, 23, 171 (1996).

[10] G. R. Spedding, F. K. Browand, and A. M. Fincham, Journal of Fluid Mechanics


314, 53 (1996), publisher: Cambridge University Press.

[11] J. T. Lin and Y. H. Pao, Annual Review of Fluid Mechanics 11, 317 (1979),
_eprint: https://doi.org/10.1146/annurev.fl.11.010179.001533.

[12] P. J. Diamessis, G. R. Spedding, and J. A. Domaradzki, Journal of Fluid


Mechanics 671, 52 (2011), publisher: Cambridge University Press.

[13] E. Lindborg, Journal of Fluid Mechanics 550, 207 (2006), publisher: Cambridge
University Press.

[14] R. Pasquetti, Computers & Fluids 40, 179 (2011).


[15] A. VanDine, K. Chongsiripinyo, and S. Sarkar, Computers & Fluids 171, 41
(2018).

You might also like