Geoscience Reference
In-Depth Information
We represent the fluctuating horizontal velocity vector u = (u, v) at a point x =
(x, y) in the plane and at time t by a sum of Fourier components of horizontal vector
wavenumber κ n =
1 n 2 n ) with random vector coefficients a n =
(a 1 n ,a 2 n ) ,
N
a n ( κ n ,z,t,α)e i κ n · x ,
u ( x ,z,t
;
α)
=
(10.54)
n
=−
N
with α the realization index. It follows that ∂u/∂x , for example, is
N
∂u
∂x =
1 n a 1 n e i κ n · x .
(10.55)
n
=−
N
With (10.53) - (10.55) we can now write w as
N
e i κ n · x
z
κ 2 n a 2 n ) dz
w( x ,z,t
;
α)
=−
i (κ 1 n a 1 n +
0
n
=−
N
(10.56)
κ 1 n z
0
a 2 n dz e i κ n · x .
κ 2 n z
0
N
a 1 n dz +
=−
i
n =− N
At horizontal scales much larger than the height z (i.e., for κ n z
1, with
κ n =| κ n |
) we expect the Fourier coefficients a n to vary only weakly with z because
turbulent eddies with large horizontal scales (small κ ) are unlikely to have small
vertical scales. Thus, for κ n z
1 we can approximate the integrals in (10.56) as
z
z
a 1 n dz
a 2 n dz
a 1 n z,
a 2 n z,
(10.57)
0
0
so that Eq. (10.56) becomes
N
( κ 1 n a 1 n + κ 2 n a 2 n ) e i κ n · x
w( x ,z,t ; α)
iz
n
=−
N
(10.58)
N
iz a n · κ n e i κ n · x n z
=−
1 .
n
=−
N
Thus, the Fourier coefficients of w are those of (u, v) dotted with
iz κ n .Since
the spectrum at any wavenumber is proportional to the mean-square value of the
Fourier coefficients at that wavenumber, it follows that
spectrum of w n z) 2
×
spectrum of u, v,
κ n z
1 .
(10.59)
Search WWH ::




Custom Search