Wave spectra and statistics

A wave spectrum is the distribution of wave energy as a

function of frequency. It describes the total energy transmitted
by a wave-field at a given time. Formally

( ) ( ) d R S 2 cos 4
0

=

(1)

where is the frequency of the waves (defined previously) and
( ) R is the autocorrelation function of the water-surface time
series

( ) ( ) ( ) | | + = t x t x E R (2)

where is the time lag between samples.


Wave spectra are strongly influenced by the wave-producing
wind and its statistical/spatial characteristics. The spatial
variability is primarily encapsulated into the fetch. Fetch is the
length over which the wind blows to generate the waves.
Virtually all models assume a constant wind speed over the
fetch. Unfortunately, this is rarely the case.


Stochastic wave distributions

Another way to assess wave conditions is to describe the water
depth (or the perturbation from the mean water level, ) at one
point for all time. To do so, the mathematics of probability
density functions becomes important.

The most common distribution used is the Rayleigh distribution:

( )
|
|
.
|

\
|
=
2
2
2
exp

p (3)


where is the perturbation from the mean water surface and is
the standard deviation of water surface. The standard deviation
is defined by

( ) d p

=


2 2
(4)


Another popular model is the Weibull distribution. The Weibull
distribution was developed primarily to describe water flow (and
stage) in rivers. It is

( ) ( )

=

exp
1
p (5)

where and are constants to be determined. Massel uses
75 . 0 = and 75 . 0 = for shallow-water situations.

Occasionally, a log-normal distribution is also assumed.



Common wave-field descriptors

To describe the intensity of the wave-field, it is useful to define
moments. Moments are defined slightly differently in wave
analysis than for turbulent flows. In this case,

( ) d S m
n
n

=

0
(6)

For instance, you can show that the standard deviation of the
water surface
0
m = .

There are several quantities used to describe the strength of a
wave field. The most common is the significant wave height
H
s
. H
s
is the average height of the largest 1/3 of the waves.
However, it occasionally given the definition

0
4
0
m H
m
= (7)

The other common wave-field descriptor is the root-mean-
square wave height H
rms
. Since the root-mean-square is
equivalent to the standard deviation (of a zero-mean process),

0
2 2 m H
rms
= (8)

Typical statistical quantities can also be expressed in terms of
the zero-moment, if we assume a Rayleigh distribution

Mean
0
2 m H H = = , Median
0
2 m H = , Mode
0
2 m H =

Of particular interest to sedimentologists is the maximum wave
height Hmax for a given H .

The problem is that the Rayleigh distribution is flat.


Assuming a Rayleigh distribution (from Massel)

0 1 . 0
09 . 5 m H = ,
0 01 . 0
67 . 6 m H =

To counteract this problem, Glukhovskiy (1966) extended the
Rayleigh distribution to shallow water and cast the pdf described
in (3) to an exceedence pdf. His formulation is

( )
( )

|
|
.
|

\
|
+
=




2 1 4
exp
1 2
H H P (9)

where h H = .


Statistics of wave period

The temporal structure of waves (i.e., the period) is more
difficult to characterize. There are three different definitions,
which have three different results. They are:

Average period between increasing zero-crossings
2 0
2 m m T
z
=

Average wave period
1 0
2 m m = T

Average period between crests
4 2
2 m m
z
= T


Wind generation of waves

To identify some important factors, we perform a dimensional
analysis

( ) g h t U X f T H
s s
, , , , , = (10)

where H
s
and T
s
are the significant wave height and period, X is
the fetch over which U blows (the most common wind velocity
used is obtained at 10 m above the surface), t is time, h is the
water depth and g is the gravitational acceleration.




|
.
|

\
|
=
2 2 2
, , ,
U
gh
U
gt
U
gX
f
U
gT
U
gH
s s
(11)


Common models

SMB Sverdrup, Munk and Bretschneider

This model was developed during and after WWII from data in
the North Atlantic. SMB uses the dimensional analysis above to
derive empirical relationships between the dependent and
independent variables.

The SMB model is as follows

|
.
|

\
|
=
42 . 0
2 2
0125 . 0 tanh 283 . 0
U
gX
U
gH
s
(12)

|
.
|

\
|
=
25 . 0
2
077 . 0 tanh 2 . 1
2 U
gX
U
gT
s

(13)

|
.
|

\
|
+

+
|
.
|

\
|

|
|
.
|

\
|
|
.
|

\
|
=
2
5 . 0
2
2
2
ln ln ln exp
U
gX
D C
U
gX
B
U
gX
A K
U
gt

(14)


K = 6.5882, A = 0.0161, B = 0.3692, C = 2.2024, D = 0.8798


Fully-developed seas (FDS) occur when the fetch no longer
controls the development of the waves. In other words, it is at
the point when there is no net transport of energy from the wind
to the waves.

Of course, the water depth term is negligible (the data was
obtained in deep water mid-North Atlantic).


However, many applications (including sediment transport)
require more than H
s
. As a result, a second-generation of
models was developed that included the entire spectrum.


JONSWAP Joint North Sea Wave Project

It was noted (first by Phillips, 1958) that at higher frequencies
than the peak frequency, the energy in a given wave-field
saturated. This saturation produced relatively equivalent
energies for a given frequency regardless of virtually all other
parameters.


Considerable data taken off the western shore of Denmark was
used to produce a model of the wave spectrum (Hasselmann,
1973).

The model is

( ) ( )

|
|
.
|

\
|
=
4
25 . 1 exp
f
f
f E S
p
(15)

where

( )

=
2 2
2
2
exp
p
p
f
f f

; ( )
( )
5 4
2
2 f
g
f E

=

f is the frequency, f
p
is the peak frequency (frequency at which
S(f) is a maximum), is the Phillips constant (sometimes called
the equilibrium-range parameter), is the peak-enhancement
factor (usually taken to be 3.3), and 07 . 0 = for or
p
f f <
09 . 0 = for .
p
f f >

There is a slight dependence on the fetch in f
p
and .
Hasselmann (1973, 1976) used the nondimensional quantities
derived above to create two empirical relations. They are

22 . 0
2
076 . 0

|
.
|

\
|
=
U
gX
(16)

33 . 0
2
5 . 3

|
.
|

\
|
=
U
gX
g
Uf
p
(17)


Donelan

Donelan et al. (1985) proposed a popular alternative which
combines the relatively weak effects due to fetch into a single
formulation. The formulation also accounts for directionality of
the wind.

The model still uses Equation (15) as its basis, though they
suggest replacing f
-5
in the linear term ( ) f E with f
-4
f
p
-1
. The
differences lie in the treatment of , , and f
p
. Donelan et al.
(1985) suggest

55 . 0
cos
076 . 0
|
|
.
|

\
|
=
p
C
U
for 5 83 . 0 < <
p
C U (18)

where
p p
f g C 2 = and is the angle between the wind and the
waves. The peak-enhancement parameter also changes

7 . 1 = for 1 83 . 0 < <
p
C U
( )
p
C U ln 0 . 6 7 . 1 + = for 5 1 <
p
C U (19)

Finally, the directionally dependent, non-dimensional fetch
relation is

23 . 0
2
6 . 11
cos

|
.
|

\
|
=
U
gX
C
U
p

(20)

where X

is the fetch in the direction of propagation of the peak

waves.


Donelan, like SMB and conventional JONSWAP, does not
account for water depth. However, these models are
EXTREMELY popular and are often used in shallow-water
situations (incorrectly?). For a good review of the pluses and
minuses of these deep-water models, see:

Schwab, D. J., et al. 1984. Application of a simple numerical
wave model to Lake Erie. Journal of Geophysical Research, v.
89, p. 3586-3592.


TMA (Bouws et al., 1985)

To correct for depth-dependent effects, Bouws et al. (1985) also
manipulated the linear term ( ) f E in (15). They wanted it to
reflect the loss of energy due to enhanced dissipation of shallow
water.

They replace ( ) f E with ( ) H f E
k
, , where

( )
( )
( )
5 4
2
2
,
f
g
H f E
H k
k

= (21)

where the factor
k
is defined by

( )
( ) ( )
( )
( ) ( )
( )

f
k
k
f
H k
h k
H k
,
,
,
,
3
3

(22)

and g h f
H
2 = . To solve for the partial derivatives, consult
Kitaigorodskii et al. (1975).

Kitaigorodskii, S. A. et al. 1975. On Phillips theory of
equilibrium range in the spectra of wind-generated gravity
waves. Journal of Physical Oceanography, v. 5, p. 410-420.

Buows et al. (1985), however, provide the graphical solution



Which Komar uses to make the point that water depth is
important.




Third-generation wind-wave models

The second-generation models discussed above are all fetch-
limited. Recent research has focused on developing numerical
algorithms capable of describing wave growth in two-
dimensions. A number of canned codes are available that do
this. The WAM (WAve Modeling group) model is one of the
most popular of these models. It is described in:

Komen, G. J. et al. 1994. Dynamics and modeling of ocean
waves. Cambridge University Press.

These models are extremely good at predicting both the
temporal and spectral (in wavenumber space) characteristics of
wave fields in response to a given wind forcing. However, they
require substantial input and are probably too complex for most
geological applications.