Vous aimerez aussi
and
lab
or
of
to
oratory
Preprin
exp
w
erimen
y
ts
o
of
It
sidew
v
all
tion
heating
of
thermohaline
stratied
as
v
tanks
ection
high
C.
hanism
Sabbah,
diusiv
R.
y
P
t
asquetti,
the
R.
b
P
ted
eyret
y
L
ab
ab.
under
J.A.
[12,13
Dieudonn?,
The
UNSA,
to
UMR
b
CNRS
terfaces
6621,
lead
e,
in
F
ection,
r
w
forces
e
t
V.
of
Levitsky
y
,
densit
Y.D.
la
y
hkin
es
Institute
observ
for
Pr
a
oblems
b
in
horizon
Me
heat
slots
Mosc
pap
ow,
so
R
some
ussia
w
width
Sp
tial
ectral
Elsevier
y
of
t
the
e.g.
o
no
w
one
whic
h
h
the
dev
h
elops
is
near
a
a
heated
e
v
en
y
w
a
all
in
with
a
This
uid
v
stably
itself
stratied
splitting
b
smo
y
prole
a
system
salinit
ers
y
high
gradien
ts.
t
in
are
are
℄
out.
in
The
erimen
ecting
of
form
w
tin
eak
solution
stratica
heating
tion
v
and
heating
oin
due
[6
to
heating
a
℄
in
t
in
heat
℄
ux
as
is
erheating
v
The
b
n
slots
umerical
to
results
resp
are
p
of
to
submitted
exp
t
erimen
28,
tal
thin
ones.
gradien
Both
in
on
(see
qualitativ
℄
e
is
and
w
quan
that
titativ
e
whic
argumen
ts,
to
a
formation
satisfactory
agreemen
sharp
t
terfaces
is
double
obtained.
e
The
v
i.e.
formation
pro
ectiv
o
is
driv
discussed
b
b
buo
y
fo
in
t
ing
o
on
onen
the
medium
o
dieren
diusivities.
of
kind
terrotativ
ec
e
manifests
b
and
the
on
of
the
initial
instabilit
oth
y
y
phenomenon
in
of
a
the
of
sim
y
ultaneous
separated
y
of
densit
gradien
The
Reviews
high
early
sensitivit
v
y
tigations
of
presen
in
h
As
phenomena
ed
to
man
sligh
exp
t
ts,
disturbances
v
at
la
the
ers
initial
in
time
is
uously
p
salt
oin
under
ted
ottom
out.
℄
1
o
In
e
tro
tal
℄
Thermohaline
p
exp
t
erimen
ts
℄
w
side
ere
in
mainly
[8,9,10,11
stim
or,
ulated
b
this
y
er,
the
tanks
disco
,14,15
v
as
ery
on
of
the
la
v
y
ered
alue.
in
the
et
o
een
and
or
refer
in
their
the
with
atmosphere,
ect
where
the
thic
oten
k
rise
homogeneous
an
la
t
y
to
ers
Preprin
are
Septem
separated
er
b
2007Le = κ /κ ≈ 100 κT S T
κS
5≈ 610
in
the
a
here
e
domain
this
of
y
o
description
used
y
n
but
tially
also
form
in
metho
other
domain
elds,
Th
lik
erimen
e
safet
in
y
es
in
to
gasstorage
the
(o
in
v
errolling
their
problem
er,
in
v
gas
t
reserv
metho
oirs)
o
[18].
Thermohaline
hallenge,
w
v
(
ection
e
tal
a
In
hallenge
they
for
n
n
the
umerical
or
metho
the
ds
results
:
e
due
ell
to
allo
the
ysical
high
e
v
Cheb
alue
and
of
hnique
the
adapted
Lewis
ectral
n
in
um
b
in
er,
e
not
ev
applications
as
e
size
v
t
ha
rather
studies
of
ection
er,
v
v
satisfactory
e
een
diusiv
in
double
the
,
w
where
the
that
terrotating
also
erimen
and
y
and
particle.
Moreo
are
a
the
heater,
thermal
the
and
,
salt
diusivities
hes
resp
:
ectiv
only
ely
of
,
the
the
are
salt
b
is
ac
essen
a
tially
the
transp
Moreo
orted,
simple
so
the
that
es
v
metho
ery
attractiv
sharp
use
in
ultidomain
terfaces
to
form
w
b
parallel
et
the
w
ulti
een
describ
is
v
ectiv
ectiv
e
observ
oratory
of
that
nearly
v
uniform
Ho
salt
p
ulation
tration.
real
Of
to
the
the
grids
the
had
and
to
ortan
b
b
e
oin
used
v
in
men
the
3D
earlier
but
whic
t
h
et
exp
the
n
estigations
of
thermohaline
appro
pap
v
are
ection
terested
in
formation
slots
us,
[19,20
are
,21].
e
Later,
,
thanks
predicted
to
linear
the
[9,31
to
ically
of
er,
the
formation
it
of
has
b
ously
ecome
w
p
ossible
already
to
Concerning
pro
umerical
in
y
tanks
disagreemen
and
n
for
[21,23
higher
w
v
essive
alues
v
of
rst
the
ectral
Ra
ds
yleigh
w
n
suited
um
ecause
b
high
er
[22,23
ws
℄
ne
Usually
of
the
ph
temp
phenomena.
erature
v
is
the
shap
ed
of
at
the
mak
heating
the
w
yshev
all
d
but
ery
in
e
some
the
studies
of
it
m
is
the
leads
heat
ux
algorithms
densit
ell
y
to
whic
h
us,
is
sp
imp
m
osed
domain
[23,24
d
℄
ed
Also,
℄
at
here
the
to
initial
the
time
v
the
e
uid
w
is
ed
generally
lab
assumed
exp
isothermal,
ts
but
w
some
ha
pap
e
ers
out.
in
w
v
er,
estigate
erforming
the
sim
situation
w
where
a
the
uid
due
is
the
initially
of
b
oth
to
stratied
heigh
in
of
uid
heating
only
all,
)
a
b
imp
used
t
p
um
a
er
2D
gridp
umerical
ts
erheated
w
temp
o
erature
and
has
salinit
een
y
to
[24
erform
,25].
reliable
More
n
often,
lo
Ho
w
ev
order
as
metho
tioned
ds
[30],
lik
in
e
estigations
nite
b
dierence
desirable,
[19
the
℄
agreemen
nite
obtained
v
b
olume
w
[22,24,23
our
℄
erimen
or
and
nite
umerical
elemen
v
t
giv
[20]
a
metho
of
ds
2D
are
ximation.
used,
this
but
er
in
e
[21
essen
,27]
in
sp
in
ectral
pro
are
Th
out.
Ho
dicult
w
observ
ev
exp
er,
tally
in
whereas
[23]
are
an
b
adaptativ
the
e
stabilit
t
analyses
w
℄
oblo
easier
obtain
k
umer
.
osite
v
grid
after
strategy
earlier
is
of
used
few
in
at
order
extremities
to
the
the
ulate
grid
simultane
p
along
oin
heated
ts
all
near
the
essively
heating
as
w
discussed
all,
℄
on
this
the
trary
n
of
what
ma
is
b
done
in
in
t
℄
the
where
umerical
equidistan
of
t
,27]
grids
sho
are
the
used.
F
or
ectiv
h
at
sp
250cm 50cm 15cm
ϕ
Tb
N Λ
2NT = 2π N Λ = gb
δS
βδS =H/Λ β H
simultane
ous
the
er
alls
is
eing
generally
length
observ
ed
h
(see
Then
e.g.
w
[8
erimen
℄
F
heat
or
w
the
v
rst
time,
are
on
top
our
elsewhere,
kno
ely
wledge,
The
the
surface
presen
the
t
an
to
sho
heat
w
tank
that
the
o
a
y
of
the
the
[22
later
essive
or
The
simultane
width
ous
after
On
pro
the
the
is
ered
strongly
the
link
o
ed
a
to
the
the
ximativ
amplitude
temp
of
initial
disturbances,
used
at
stratied
least
y
when
wn
high
thanks
order
electric
metho
the
ds
of
are
y
in
d
v
angular
olv
℄
ed.
ely
In
in
a
.
same
et
w
ottom
a
from
y
,
e
it
in
is
heigh
horizon
p
resp
oin
formation
ted
side,
out
and
that
seems
ottom
terrotating
isolated
are
upp
m
uc
y
h
A
more
ter
in
t
tense
t
when
e
no
aterheater
disturbances
heater.
are
aterheater
used
hanger)
at
p
the
ose
initial
whereas
time,
heater
so
trol
that
densit
some
A
disagreemen
system
ts
ll
b
a
et
i.e.
w
the
een
t
exp
and
erimen
ards.
tal
is
observ
an
a
sensor:
tions
y
and
y
or
t
n
uid
umerical
the
in
the
v
p
estigations
simultane
ma
Brun
y
the
b
e
the
b
resp
etter
the
understo
o
and
d.
the
The
salinit
pap
,
er
een
is
the
organized
the
as
e
follo
relation
w
w
:
hniques
b
2
found
is
e.g.
dev
[32].
oted
tank
to
t,
a
and
brief
tal
description
are
of
ectiv
the
of
exp
the
erimen
,
tal
tal
setup,
exp
to
the
its
obtained.
mo
.
delling
b
and
and
to
v
the
w
sp
are
ectralCheb
and
yshev
free
m
er
ultidomain
is
metho
v
d
b
used
a
to
sheet.
solv
t
e
the
of
go
tank
v
w
erning
dieren
equations.
heaters
In
b
used:
3
w
the
or
results
electric
of
Using
the
w
n
(at
umerical
sim
appro
ulation
ely
are
ermits
presen
imp
ted
the
and
erature,
with
with
electric
the
one
exp
erimen
the
tal
ux
results
y
are
.
pro
t
vided.
otank
In
is
to
4,
the
w
with
e
linearly
p
uid,
oin
t
as
out
salinit
the
gradien
is
of
t
initial
do
disturbances
w
on
Its
the
alue
measured
formation
to
pro
electric
y
and,
the
nally
,
b
in
to
salinit
5.
,
2
sligh
Exp
disturbance
erimen
the
tal
setup,
of
mo
delling
at
and
buo
n
umerical
erio
metho
ous
d
.
2.1
the
Exp
tV?is?l?
erimental
frequency
setup
and
Here
w
length
e
of
only
stratication
obtained,
the
ectiv
from
elemen
relations
ts
of
required
the
for
whereas
rst
one,
the
upp
,
at
the
The
dier
y
the
y
and
b
t
w
the
since
and
more
b
details
of
on
tank
the
b
exp
determined
erimen
the
tal
all
setup
er
and
a
lo
go
o
where
d
is
understanding
salinit
of
expansion
the
pap
and
er,
3N
g∂ρ2N =−
ρ∂y
N H ≪ Λ
∂ n nx
x ∂ ρ ρ nx
ϕ
30cm
10cm
K h =αδTΛa a
L = 12cm
25cm H =
50cm
2mm
erature
prob
es
is
o
are
and
lo
gradien
tak
inside
the
and
tank:
the
Esp
only
ecially
kno
,
On
the
en
is
vironmen
the
t
In
temp
isolated
erature,
w
the
Computational
mean
righ
temp
v
erature
uid
of
as
the
has
electric
e
heater
phenomenon
and
onditions:
the
The
temp
onditions:
era
mo
tures
h
of
,
the
the
v
v
w
the
alls
Mo
opp
ob
osite
the
to
of
the
in
heater
the
are
from
measured.
rest,
A
.
Sc
so
hlieren
,
system
ept
is
v
used
in
to
b
visualize
ultaneous
the
o
initial
w,
temp
b
t
y
ards.
pro
viding
The
an
a
image
densit
at
with
the
Ho
for
.
b
since
y
,
the
the
(iii)
partial
the
deriv
uid
ativ
rise
e
densit
of
[12].
the
ling
uid
F
refractiv
symmetry
e
part
index
on
uid,
hand
with
heater
resp
sho
ect
1.
to
to
the
horizon
that
tal
heater
axis
nearly
stratied
halfwidth
,
smaller,
itself
ev
than
with
alid
tially
onen
heigh
p
een
,
to
due
of
to
since
the
not
dep
in
endence
but
of
the
ex
the
on
of
an
ectiv
.
Initial
The
t
the
b
rest.
et
is
w
y
een
the
wn
n
Boundary
umerical
b
and
used
exp
er
erimen
heater
tal
b
results
enden
will
heat
b
,
e
anishes
done
heater
for
t
the
exp
erimen
reasons,
ts
e
lo
with
(ii)
the
that
electric
alls
heater,
are
for
and
whic
y
h
losses
the
o
b
erheated
oundary
particle
ould
is
in
w
initial
ell
y
t
trolled:
2.2
the
del
heat

ux
domain:
densit
or
y
vious
for
reasons,
ma
the
y
of
b
tank
e
the
assumed
t
uniform
side
on
the
the
is
heater
as
and
wn
the
Fig.
global
Moreo
ux
er,
result
from
v
and
oltage
wing
and
far
electric
the
the
t
is
measuremen
at
ts.
the
The
is
heigh
en
t
T
of
w
the
er,
electric
rather
heater
on
is
as
equal
.
to
the
t
trary
the
only
t
b
,
k
lea
equal
ving
v
a
assuming
free
alue
space
,
of
w
in
are
is
only
that
terested
lo
at
phenomena
the
also
top
y
and
global
at
of
the
sim
b
ottom
the
of
v
the
e
tank.

F
rom
A
an
the
exp
time
erimen
uid
t
at
one
The
erature
homogeneous.
sev
salinit
eral
gradien
is
t
parameters
do
of
w
the

o
w,
Neumann
lik
oundary
e
are
the
for
mean
temp
heigh
ature.
t
electric
and
is
width
delled
of
y
the
timeindep
t
as
uniform
w
ux
ell
y
emp
whic
on
v
h
regularly
are
the
to
extremities:
end.
agreemen
the
with
ux
heater
y
hnology
assumed
t
ull:
as
deriv
ed
side
quan
used
tities
this
Elsewhere,
h
heat
as
densit
the
is
ratio
n
us
(i)
er
symmetry
sur
ab
the
v
mean
and
heigh
e
t
w
to
heater,
the
b
heigh
assuming
t
the
Let
w
a
of
tank.
tank
the
p
of
t
also
heigh
b
to
neglecting
whic
thermal
h
at
an
upp
the
uid
of
41K
1
D V =−∇p+Ra (T −R S)e +ΔVt ϕ ρ y
Pr
∇.V =0
DT =ΔTt
1
DS= ΔSt
Le
e t V py
T S
Dt
2L L /κ κ /LT T
2ρνκ /L Lϕ/λ δS ρ λ νT
Pr =ν/κT
Le =κ /κT S
Ra Rϕ ρ
4αϕgL βδSλ Hλ
Ra = , R = =ϕ ρ
κ νλ αϕL ΛαϕLT
L
mean
v
alues
at
densit
y
y
the
,
initial
the
time.
dep
The
Prandtl
partial
the
dieren
t
tial
Boussinesq
equations
of
where
then
b
y
e
b
written
b
in
of
dimensionless
transp
form:
o
and
reference
heat
erature
the
on
and
and
motion
and
uid
The
ectiv
quations:
only
e
(
Governing
erature

.
Ra
y
for
diusion
elo
F
v
dened
the
es
for
v
used
us,
are
mo
y
noslip
,
and
the
y
v
salinit
enden
the
assumed
(1)
for
,
used
are
are
,
and
enetration)
y
(2)
.
(nop
in
Neumann
ariations
Homogeneous
and
.
the
out
salinit
ab
n
(3)
are
is
parameters
heater
whereas
the
n
on
temp
erheating
the
v
ratio
o
the
o
that
the
(4)
they
with
y:
wing
kno
Na
for
b
the
is
unit
the
v
ximation.
within
v
using
ector,
what
when
to
for
salinit
the
and
time,
temp
,
for
ariations
the
the
v
t
elo
,
linearly
y
and
,
term
are
,
for
y
the
buo
deviation
the
of
the
the
y
pressure
the
from
y
the
the
h
viscosit
ydrostatic
resp
one,
ely
assumptions
The
and
(
h
resp
are
ectiv
v
ely
)
for
Lewis
the
densit
deviations
(ii)
of
y
the
and
temp
)
erature
um
their
ers
(5)
the
that
of
of
uid,
as
the
yleigh
is
um
v
er
t
the
the
and
d
buo
but,
the
equations
ysical
are
oin
of
the
really
w.
for
or
t
presen
e
problem
mains
are
.
b
w
ort
er,
to
is
equations
a
vierStok
p
to
y
a
erned
v
go
t
w
length
uid
the
(i)
and
Th
to
appro
the
the
umerical
delled
b
and
salinit
Note
y
the
hoice
are
tration
from
their
length
mean
initial
enien
v
for
alues
mo
and
elling,
from
done
ph
our
p
of
t
n
view,
and
justied
erimen
slot
results.
yp
These
do
quan
only
tities
Ho
ha
ev
v
it
e
alw
b
ys
een
ossible
made
use
dimensionless
b
enien
y
reference
using
for
the
follo
then
wing
discuss
reference
n
v
results
alues:
y
transfers
other
This
lengths.
,
is
salinit
is
in
for
analysis
the
the
material
umerical
deriv
exp
ativ
tal
e.
5V.∇S
IP I xI,J
k¯J y ΩI,J
k k¯Ω 1≤ k ≤ K ΩI,J
2 k k¯ ¯k = 1,...,K V IP (Ω ) p T S IP (Ω )I,JI,J
kΔT −σ T =f in ΩT T I,J
∗ kΔS−LeV .∇S−σ S=f in ΩS S I,J
kΔV −σ V −∇p=f (T,S) in ΩV V I,J
k¯∇.V =0 in ΩI,J
+ BoundaryConditions.
k = 1,...,K−1
[[V ]] = [[T ]] = [[S]] =0k k k
[[∂ V ]] = [[∂ T ]] = [[∂ S]] = [[p]] =0y x k y k y k k
k k+1¯ ¯[·] I = Ω ∩Ωk k
elds.
The
partial
dieren
domain
w
,
tial
e
equations
uit
are
es
appro
a
ximated
with
n
the
whic
Cheb
ximation.
yshev
for
bined
sp
metho
of
d
y
using
v
a
equation,
strong
or
form
damsBashforth
ulation.
the
The
order
general
Discr
Stok
in
es
(6)
problem
is
the
solv

ed
use
with
on
the
this
same
an
v
(9)
ector
w
space
alue
is
follo
used
for
for
im
the
for
the
a
v
ectiv
elo
are
kw
y
y
The
onen
where
ts
the
and
eing
for
h
the
parallel
pressure.
is
Finally
,
with
e:
and
etization
y
salinit
an
the
for
timestep
the
space
implicitly
of
(8)
the
w
p
treated
olynomials
the
of
small
maxim
lead
um
b
degree
the
equation
large
.
with
e
(horizon
tal
salinit
axis)
.
and
is
yp
salinit
in
olation,
ectiondiusiont
order
(v
y
terms
axis),
(11)
adv
implicitly
steady
linear
a
Euler
(ii)
dierence
erature,
a
temp
discretized
the
deriv
set
in
of
d
the
v
gridp
es
oin
at
ts
Stok
of
b
the
sub
subdomain
the
for
to
equation
(
hnique
Helmholtz
osition
an
domain
(i)
ely
v
Along
)
ac
and
in
and
Discr
timecycle
(7)
h
pro
iterativ
at
of
the
the
subset
of
but,
the
the
inner
t
p
the
oin
strongly
ts,
term
at
Considering
.
h
a
timecycle
explicit
one
in
solv
as
es
term
the
if
follo
timesteps
wing
ery
problem:
to
F
ould
or
er
e
um
solv
Lewis
to
of
used
v
is
the
d
(10)
metho
the
ectral
wing
sp
tin
,
y
nd
ts:
pseudo
y
in
the
a
F
domain,
plicitly
sub
treated
h
h
y
In
the
osed.
imp
extrap
are
A
transmission
using
and
b
natural
explicitly
,
e
and
v
,
tial
and
in
treated
essen
terms
the
The
domains,
appro
sub
ard
the
een
nite
w
et
using
b
b
terfaces
are
ativ
h
time
as
time:
:
etization
in

the
(12)
t
elo
pressure
the
and
problem
y
is
jump
the
terface
asso
a
generalized
A
metho
(iii)
pro
a
al
to
2.3
in
6σ f
∗V
T S V
Δ−σS
IP −IPN N
[p] = 0k
0.5(Ra )ϕ
8Pr = 7 Le = 100 Ra = 3.46× 10 R =ϕ ρ
1.66 K = 34 I = 250
5J = 70 6×10
x
τ ≈ 0.2s
t = 30mnF
0.014K
imp
ose
space
with
the
transmission
,
These
y
nal
matrices,
a
whic
divided
h
wing
are
from
in
ulate
preliminary
w
initial
asso
ha
of
for
v
the
in
v
ond
n
ing
solv
homogeneous
t
problem
the
the
h
in
transform
terface
heater
v
v
alues
v
to
the
the
the
jumps.
Finally
Then,
men
the
o
full
whic
problem
alues
is
are
solv
b
ed
in
with
tal
the
done
the
v
at
alues
homogeneous
of
problem
matrices
and
,
.
use
and
of
inu
(7),
at
out
the
ts.
in
elo
terfaces.
oin
In
w
units
h
from
subdomain,
(S1
(i)
p
the
and
equation
ulations
for
temp
the
is
temp
to
erature
w
is
e
solv
that
ed
oid
with
v
a
dimensionless
metho
reference
d,
time
(ii)
for
ectiv
the
m
salinit
y
but
equation
.
one
exp
uses
Calculations
the
b
the
Conju
alues
gate
Residual
b
(see
unkno
e.g.
alues
[33])
ed
iterativ
is
e
olation.
pro
using
o
with
at
the
or
Helmholtz
w
op
elo
erator
domains,
using
v
y
b
in
determined
i.e.
as
in
preconditioner,
gridp
and
(iii)
used
the
to
generalized
mesh
Stok
near
es
the
problem
esp
is
ph
solv
timestep
ed
in
.
a
sim
S2)
w
b
a
till
y
,
using
These
a
from
P
for
oisson
for
equation
temp
for
disturb
the
amplitude
pressure.
expressions
The
,
pressure
e
b
v
oundary
to
v
tion
alue
to
is
v
to
b
small
y
alues
using
the
again
timestep,
an
h
matrix
the
v
hnique,
for
yielding
and
a
elo
div
y
ergencefree
resp
v
ely
elo
and
ultiplied
y
y
eld.
to
Nev
taking
ertheless,
terfaces,
due
the
to
3
the
and
use
erimen
of
results
the
ha
are
e
alues
een
v
with
terface
follo
in
v
of
the
t
dimensionless
yp
um
e
ers:
appro
wns
ximation,
the
the
of
pressure
v
is
with
not
,
unique
rst
and
(6)(10)
aected
The
with
extrap
the
a
socalled
timecycles,
spurious
previous
mo
w
des
the
of
y
pressure.
F
Consequen
the
tly
discretization
,
e
the
transmission
v
sub
on
with
the
alues
pressure,
the
Then,
and
is
oundary
equation
b
the
subdomain,
and
ab
terms
the
forcing
,
,
the
oin
t
A
to
ordinate
no
is
are
in
for
,
As
ected,
the
disturb
p
ts
yields
the
in
where
etter
o
t
is
the
ecially
erimen
In
data
ysical
y
the
imp
is
osing
The
them
the
w
result
eakly
T
,
o
using
ulations
as
and
trial
ha
functions
e
a
een
set
erformed,
of
the
p
time
olynomials
terms
orthogonal
ts
to
the
.
pressure
sim
space
dier
k
the
ernel.
The
the
details
erature:
of
S2
the
initial
algorithm
erature
are
randomly
giv
ed
en
an
in
equal
[28
of
℄
The
b
e
whereas
enforced
disturbances
strongly
used
.
S1.
The
exp
dicult
the
y
ed
has
S2
b
results
een
b
o
agreemen
v
with
er
exp
tal
b
7oS = 5( / )oo
Tf
ϕm
Λ Tf
ϕ Pr Le Ram ϕ
Rρ
o −2Λ(m) T ( C) ϕ (W.m ) Pr Le Ra Rf m ϕ ρ
8107.8 18 71.9 7.15 99.3 3.23×10 1.72
8108.6 19 74.2 7.08 100 3.51×10 1.57
8103.2 20 69.4 7 100 3.46×10 1.66
α
α αϕ
α
ϕ
o −4 −1α α(18 C) = 1.8510 K
o −4 −1 o −4 −1α(19 C) = 1.9610 K α(20 C) = 2.0610 K
),
Lewis
salinit
mean
(
er,
eing
1
b
thermal
),
for
Ra
only
yleigh
(
of
er
(
y
h
la
ha
)
is
n
in
um
so
b
ma
ers
of
and
er
buo
(at
y
the
ers
ratio
t
(
on
er
erature.
upp
)
since
for
w.
the
b
exp
that
erimen
ts
n
and
v
for
balanced
the
.
sim
used
ulations.
heat
Exp.
ysical
the
estimate
,
sit
y
used
salinit
uid
kground
Stratication
from
eak
w
,
a
dep
with
oth
done
and
een
w
b
e
e
v
v
w
ha
salinit
ts
ery
erimen
w
exp
ma
oratory
observ
lab
equation
of
app
set
the
A
exp
disturbances.
for
initial
of
in
of
E1
b
the
y
study
in
to
v
order
that
in
T
5,
densit
the
in
and
S2
uid
of
v
those
),
to
to
ux
e
the
b
(
will
erature
S1
for
ulation
sim
b
the
for
the
obtained
expansion
results
The
T
whic
this
strongly
in
ends
used
b
are
the
so
y
2:
temp
E2
Here
where
e
1
v
able
simply
T
its
in
alues
wn
fresh
sho
ater,
as
the
y
the
v
for
lo
used
Ho
those
ev
to
it
equal
y
ely
e
ximativ
ed,
appro
the
ers
(5),
b
ts).
um
ears
n
through
dimensionless
pro
the
erimen
of
,
alues
that
v
the
yield
umerical
h
a
whic
hange
S1,S2
the
E2),
alue
and
the
(E1
y
them
e
of
b
o
a
w
hange
t
during
on
The
T
alues
e
y
able
are
fresh
in
w
able
Note
are:
for
ux
sak
mean
of
for
y
).
same
parameters
are
ph
for
the
v
alues
with
the
without
,
the
ater
(
(
Prandtl
y
stands
den
latter
heat
T
),
go
lo
from
erature
to
temp
of
and
the
temp
dieren
),
t
w
dimensionless
length
n
um
b
1
men
in
(
W
en
giv
.
also
that
are
the
ulations
e
sim
umerical
,
n
notations
the
used
for
the
alues
used
or
alues
dimension,
v
dimension
the
eing
reference
tioned
alues
the
listed
y
o
in
easily
the
ones
determination
others,
the
main
The
v
main
are
dicult
in
fo
able
and
8−12 −1 −2L L κ κ L κ LT TT
5 −6 −1 −5 −10.12m ≈ 10 s 1.1810 m.s 0.9910 s
−1 −1 −1κ Lϕλ HΛ βT
−7 2 −1 o1.4210 m .s 14.1K ≈ 5 /oo
ω ψ T S
t ={t = 0.0048,t = 0.006,t =1 2 3
0.012,t (= t ) = 0.018} t = {8,10,20,30mn}4 F
∂ ρx
tF
•
• t ≈ t1
∂ ρ t≈tx 2
•
times
t
dieren
simultane
steep
the
the
temp
at
ariables.
ed,
of
y
v
displa
sho
ely
e
ectiv
together
resp
are
the
y
of
salinit
t
the
nal
and
temp
erature
the
temp
phenomena
the
,
egin
Afterw
the
on
,
y
A
of
v
ell
the
a
of
enlargemen
isolines
y
The
,
S2.
plexit
ulation
y
sim
the
the
the
with
,
obtained
b
results
forms
the
the
w
ectiv
sho
its
5
then
(
app

ones
2
This
Fig.
esults
tral
r
are
al
the
ation
3.1
Just
y
observ
salinit
earance,
erature
horizon
temp
from
able
y
of
and
v
at
in
time
ph
order
ysical
the
units).
of
In
and
Fig.
with
6
ts
is
y
giv
terfaces
en,
v
at
Essen
the
follo
same
v
times,
observ
the
upo
y
the
distortion
elo
y
elds,
whic
h
form
ma
rst
y
ottom
b
the
e
new
view
in
ed
the
as
h
the
gro
a
tures
pro
ation
vided
the
b
and
y
the
the
t
Sc
,
hlieren
system.
a
o
that
efore,
these
streamlines
gures
as
only
the
visualize
the
the
eld,
strip
e
a
part
heater.
of
Main
the
es
ts
domain:
the
elsewhere
erature
the
salinit
uid
elds
is
the
nearly
time
at
length
rest.
in
Ho
to
w
w
ev
er,
y
one
the
observ
erature
es,
salinit
a
elds,
w
the
a
gradien
y
of
from
salinit
v
at
a
in
of
of
e
ectiv
ed
along
tially
heater.
the
Then
wing
ha
is
e
b
een
the
ed:
of
An
doubling
w
along
heater,
a
elops
of
to
salinit
set
isolines,
t
with
o
v
in
e
b
later,
to
to
at
set
extremities,
three
at
b
a
and
system
at
of
top.
dissipativ
ards
e
gra
ear
vit
y
of
w
rst
a
whic
v
go
es
in
whic
wing.
h
is
propagate
phase
in
to
essive
the
e
undisturb
of
ed
In
uid.
It
part
is
near
in
heater,
teresting
streamlines
that
v
the
v
t
has
dieren
for
this
the
w
ous
a
e
v
of
ephenomenon
set
whic
h
w
b
as
analyzed
the
earlier,
are
b
ed
oth
w
on
as
its
app
the
in
oretical
alues
and
reference
exp
of
erimen
tal
tal
es,
asp
w
ects
y
[34,35,32
the
℄
A
Fig.
time
7
2
giv
the
pattern
v
ectiv
b
e
observ
all
and
the
T
the
heat
the
fron
width
t,
horizon
y
tal
mean
strip
a
es
pro
asso
:
h
with
dev
a
in
mo
a
dulation
of
along
w
the
small
v
ternal
and,
in
These
a
strip
of
es
9∂ ρx
•
•
δTm
h lm m
15cm hm
∂ ρ lx m
t =t1
ha
3αgδThaRa = , h =αδTΛa
νκT
Ka
h K =h /ha a m a
δT h l Km m m a
Ra
Ka
Ra
4αδT Ra ∝ (αδT)
ards
its
whic
the
tral
w
part.
the
egins
the
Within
a
sensitivit
h
yleigh
its
and
the
in
the
ternal
t
t
the
the
salt
the
since
tration
ed
tends
measured
to
o
b
[12
ecome
y
homogeneous.
ratio
This
t
to
for
h
,
ne
results
the
limited
the
b
orse
y
is
sti
ternal
salinit
to
y
gra
e
dien
e:
ts
er
at
v
the
dieren
particle
t
y
outer
e
and
the
inner
in
,
terfaces.
heigh
3.2
oten
Comp
h
arisons
.
of
the
the
E1
S2.
al
and
ecially
exp
.
erimental
go
data
erimen
Bey
of
ond
the
the
with
go
.
o
erimen
d
heat
agreemen
when
t
um
of
results
the
n
exp
then
erimen
the
tal
v
and
ha
n
umerical
tities,
o
the
w
um
pattern
on
ev
olutions,
and
esp
heigh
ecially
of
with
erheated
rst
reac
the
neutral
formation
lev
of
at
v
v
t
ectiv
onding
e
erio
the
at
eep
the
to
ex
the
tremities
of
of
the
the
rising
heater
out,
and
es
then
mean
with
ables
the
giv
sim
obtained
ultaneous
erimen
E2
of
sim
a
t
set
one
of
elds.
visible
v
,
ectiv
pro
e
n
ear
quan
d
titativ
the
e
E1,
heigh
ha
v
smaller
e
than
also
erimen
b
een
gro
out.
the
Th
E2,
us,
the
w
densit
e
esp
ha
v
yleigh
e
er
This
the
the
ev
of
olution
b
of
pro
the
heater
mean
of
temp
lo
erature
Moreo
rise
er,
b
e
rst
v
also
of
deriv
the
quan
heater
lik
as

w
Ra
ell
n
as
b
the
based
mean
the
heigh
(or
t
o
pro
erheating
This
on
and
rising
width
t
merging.
their
of
an
of
v
the
uid
to
for
h
the
new
exp
buo
er
imen
el
ts
℄
E1
and
ectiv
E2
and
of
for
heigh
the
to
sim
ulation
S2.
p
F
v
or

b
oth
k
the
equal
exp
the
erimen
of
ts
mean
and
t
the
the
sim
to
ulation,
p
these
tial
heigh
quan
but
tities
:
ha
stretc
v
strip
e
time,
b
the
een
T
obtained
3
b
5
y
e
fo
results
for
on
exp
the
ts
and
tral
and
part
the
of
ulation
the
A
heater,
dieren
i.e.
times,
along
ab
In
out
the
y
on
a
,
w
esp
the
is
.
and
The
wing
v
The
alues
umerical
of
app
y
in
b
o
are
agreemen
easily
with
obtained,
exp
e.g.
t
from
although
the
mean
t
is
the
is
w
little
.
for
determination
of
for
the
exp
of
ts,
is
less
on
straigh
ratio
tforw
This
ard
They
(see
w
e.g.
with
the
exp
dieren
t
t
for
elds
h
at
mean
t
ux
heigh
y
mean
higher,
The
ecially
).
fo
W
on
e
Ra
ha
n
v
b
e
used
.
the
mainly
tem
from
p
high
erature
y
eld
this
to
um
get
er
an
the
to
propagates
and
the
,
heat
the
fron
extremities
t
estimate
at
of
in
eld.
The
10