12. Frequency vs Ome domain (1)
Normally signals appear in the time domain:
v(t) with t ∈ [0,T]
where T is the length of a record. If we assume that:
v(t +T) = v(t)
then the function is said to be periodic. Furthermore
if v(t) has a finite number of oscillations in [0,T] then
we can develop v(t) in a series:
v(t) = Ai
i=0
N/2
∑ cos(ωit)+ Bi sin(ωit)
which is known as a Fourier series and where Ai and Bi
denote the Euler coefficients.
21/03/18 12
17. Frequency versus Omeseries (5)
• EssenOally y=FFT(x) carries out a Fourier transform
• FFT algorithm input
– Real vector x(0..N-1) with N datasamples
– The record starts at 0 and is filled to N-1
• FFT algorithm output
– Euler coefficients are stored in the form of complex numbers
– Stored in y(i) are:
y(0) = A0 + I.B0 , y(1) = A1+ I.B1 , …, y(N/2) = AN/2 + I.BN/2
– I is a complex number:
– Be careful with scaling factors, check this always with a test funcOon of
which you now the Euler coefficients in advance
– Suitable test funcOons are for instance linear sin and cos expressions
• FFT algorithms exploit symmetries of sin and cos funcOons, please
use MATLAB because it is thoroughly debugged.
• Be aware that indices in MATLAB vectors start at 1 (and not 0)
I = −1
21/03/18 17
25. Aliasing = Folding
frequency
power
Watch how a part of the spectrum above the Nyquist frequency folds back
onto the lower part of the spectrum, this is what we call aliasing
Nyquist
frequency
34. Ocean tides (chapter 16)
• Purpose of this part is to describe motions of fluids
with a system of differential equations.
• We are only interested in representing currents and
water levels at time and space scales appropriate to
ocean tides. (several km to global, hours to days)
• Where is the input, what is the response, is it linear?
40. ConOnuity equaOon (5)
If we allow variations along x the u component
becomes:
dydzdx
x
u
udx
x
))((
∂
∂
+
∂
∂
+
ρ
ρ
When this is allowed for all components we get:
0
1
=⎥
⎦
⎤
⎢
⎣
⎡
∂
∂
+
∂
∂
+
∂
∂
+
z
w
y
v
x
u
dt
dρ
ρ
For non-compressible fluids the first term will vanish
42. Momentum equaOon (2)
Why is the pressure gradient like shown? (Pressure = force / area,
Force = pressure times area)
dx
dy
dz
P P+dP
Force effect x direction: zyx
x
P
zydPzydPPzyP δδδδδδδδδ .....).(..
∂
∂
−=−=+−
44. Momentum equaOon (4)
How to get the Coriolis term? Answer: consider an inertial
(i) and a rotating coordinate system (a).
a
a
a
a
a
a
i
i
a
a
a
a
i
i
a
a
i
i
xxxx
xxx
xx
eeeex
eeex
eex
!!!!!!!!!!
!!!!
++==
+==
==
2
Unit vectors coordinate basis: e
Point ordinates on a chosen basis: x
45. Momentum equaOon (6)
For a rotating basis we have:
aaa
aa
eωωeωe
eωe
××+×=
×=
!!!
!
And the consequence is:
aaaaI xωxωωxωxx ×+××+×+= !!!!!! 2
“frame”“centrifugal”
46. Momentum equaOon (7)
Du
Dt
=
−1
ρ
∇P − 2ω × u + g + F
The final result is:
Note:
• pressure and density symbols (watch the symbols)
• g does contain a centrifugal term!
• rotation is considered constant
• this equation holds in an Earth fixed frame
• contains Friction and ForcingF
47. Navier Stokes equaOons
€
Du
Dt
=
−1
ρ
∂p
∂x
+ 2Ωsinφv − 2Ωcosφ w + Fx
Dv
Dt
=
−1
ρ
∂p
∂y
− 2Ωsinφ u + Fy
Dw
Dt
=
−1
ρ
∂p
∂z
+ 2Ωcosφ u − g + Fz
Notes:
- w equation: largest terms result in hydrostatic equation
- D./Dt includes local derivative and advection (later)
For a suitable local coordinate frame we find:
48. HydrostaOc equaOon
For the w-equation we get:
€
∂p
∂z
= −gρ
p(z) = − gρdzH
−η
∫ = gρ(H +η)
Example:
• At 1000 m the pressure is approximately 100 bar
• Density changes in the oceans are relatively small (3%)
49. Velocity depth averaged equaOon
f = 2Ωsinϕ
€
Du
Dt
= −g
∂η
∂x
+ fv + Fx
Dv
Dt
= −g
∂η
∂y
− fu + Fy
Dη
Dt
= −(H +η)
∂u
∂x
+
∂v
∂y
%
&
'
(
)
*
In these equations u and v are averaged over the entire water
column. The Coriolis term
50. Laplace Odal equaOons
€
Du
Dt
= −g
∂η
∂x
+ fv + Fx
Dv
Dt
= −g
∂η
∂y
− fu + Fy
Dη
Dt
= −(H +η)
∂u
∂x
+
∂v
∂y
%
&
'
(
)
*
€
Fx
Fy
"
#
$
%
&
' =
∂Γ ∂x
∂Γ ∂y
"
#
$
%
&
' +
rx
ry
"
#
$
%
&
'
Γ = Hω
ω
∑ cos(χω (t) − Gω )
The terms u,v and η represent velocity and water level, f is a
Coriolis term, H is the depth of the ocean, Γ is a forcing
function, r represents dissipative forcing. (Chapter 16)
55. Helmholtz equaOon (4)
If you substitute the characteristic solutions for all dependent
variables in the shallow water equations:
(ω2
− f 2
) ˆη + c2 ∂2
ˆη
∂x2
+
∂2
ˆη
∂y2
⎛
⎝⎜
⎞
⎠⎟ ≈ 0
c = gH
In this relation c is the surface speed of the tide. You can
obtain it by substitution of the test function in the Laplace
tidal equations.
59. Forcing
• Question: what are Fx and Fy in the momentum equation?
• Answer: this depends on the type of problem
• Some examples are:
• Gradients tide generating potential
• Friction
• (Advection is not a forcing term)
60. Linearity
Ocean tides show a linear behavior when sea bottom friction
is linear and when advection is ignored:
rx =
Cu
H
and ry =
Cv
H
€
Γ(t) = ˆΓexp( jωt)
In reality sea bottom friction is quadratic:
r = Cu u
64. FricOon
• This is described by
• Fx acts on a molecular level.
• The term ν describes kinemaOc molecular viscosity, ν is
about 10-6 m2/s
• FricOon always introduces dissipaOon
• To apply fricOon in Ode models you have to re-scale fricOon
by means of horizontal and verOcal eddy viscosity terms
• Scale factor depends on numerical consideraOons
Fx =ν
∂2
u
∂x2
+
∂2
u
∂y2
+
∂2
u
∂z2
"
#
$
%
&
'
68. Dispersion relaOon
u(t) = ˆuexp( j(ωt − kx − ly))
v(t) = ˆvexp( j(ωt − kx − ly))
η(t) = ˆηexp( j(ωt − kx − ly))
∇F = 0
Unforced or free waves fulfill the following relation:
Show that this results in the following dispersion relation
)()( 2222
lkgHf +=−ω
Do free waves exist for all frequencies at all latitudes?
69. Simple Ode model
€
h(φ,λ,t) = Hi φ,λ( )
i
∑ cos wi t − t0( )− Gi φ,λ( )( )
Hi φ,λ( ): tidal amplitude
Gi φ,λ( ): Greenwich phase
Deep ocean tides can be predicted by the model:
The tidal frequency is astronomically determined, for a
location at which the tides are predicted it is sufficient to
have a series of H and G values available, t0 is a reference
time, t is the time of observation.
70. M2 tide with phase lines in hours.
North Sea Tides
90. Local DissipaOon (1)
∂u
∂t
+ f ×u = −g∇ζ + ∇Γ − F
∂ζ
∂t
= −∇.(uH )
W − P = D
W = ρH < u.∇Γ >
P = gρH∇.< uζ >
D = ρH < u.F >
W: Work
P: Divergence Energy Flux
D: Dissipation
∫
+=
=
=
Ttt
tt
dtFu
T
D
0
0
).(
1
100. Conclusions (1)
• Global dissipaOon:
– There are consistent values for most models,
• The M2 dissipaOon converges at 2.42 TW to within 2%
– Independent methods to determine the rate of energy
dissipaOon (LLR, satellite geodesy). LLR arrives at 2.5 TW
for M2
– Comparison to astronomic/geodeOc values:
• 0.2 TW at S2 for dissipaOon in the atmosphere
• 0.1 TW at M2 for dissipaOon in the solid earth
• gravimetric confirmaOon of the 0.1 TW is very challenging
– History of Earth rotaOon relies of dissipaOon esOmates from paleo-
oceanographic ocean Ode models.