Ocean wave model output parameters

Jean-Raymond Bidlot
ECMWF
[email protected]
February 27, 2020

1 Introduction
This document gives a short summary of the definition of ECMWF ocean wave model output param-
eters. Please refer to the official IFS documentation part VII for a full description of the wave model
(https://software.ecmwf.int/wiki/display/IFS/Official+IFS+Documentation). The data are encoded in grib,
edition 1. See table(1) for the full list of all available parameters

2 Wave spectra
There are two quantities that are actually computed at each grid point of the wave model, namely the two-
dimensional (2-D) wave spectrum F (f, θ) and the total atmospheric surface stress τa for a given atmospheric
forcing. In its continuous form, F (f, θ) describes how the wave energy is distributed as function of frequency f
and propagation direction θ. In the numerical implementation of the wave model, F is discretised using nf re
frequencies and nang directions. In the current analysis and deterministic forecast configuration nf re = 36
and nang = 36. Whenever possible, F (f, θ) is output and archived as parameter 251. It corresponds to
the full description of the wave field at any grid point. It is however a very cumbersome quantity to plot
as a full field since it consists of nf re × nang values at each grid point. Nevertheless, it can plotted for
specific locations. Figures 1 and 2 display the spectra at 6 different locations as indicated on Figure 3
(model bathymetry, parameter 219 is also shown). The two-dimensional wave spectrum is represented using
a polar plot reprsentation, where the radial coordinate represents frequency and the polar direction is the
propagation direction of each wave component. The oceanographic convention is used, such that upwards
indicates that waves are propagating to the North. The frequency spectrum, which indicates how wave
energy is distributed in frequency, is obtained by integrating over all directions. From Figure 1, it clear that
for many locations, the sea state is composed of many different wave systems.

3 Integral parameters describing the wave field

In order to simplify the study of wave conditions, integral parameters are computed from some weighted
integrals of F (f, θ) or its source terms. It is quite often customary to differentiate the wave components in
the spectrum as windsea (or wind waves) and swell. Here, windsea is defined as those wave components that
are still subject to wind forcing while the remaining part of the spectrum is termed swell, total swell if the
full remaining part is considered as one entity or first, second or third swell partition when it is split into
the 3 most energetic systems (see section on swell partitioning).
To a good approximation, spectral components are considered to be subject to forcing by the wind when

1.2 × 28(u∗ /c) cos(θ − φ) > 1 (1)

where u∗ is the friction velocity (u2∗ = ρτair

a
), ρair the surface air density, c = c(f ) is the phase speed as
derived from the linear theory of waves and φ is the wind direction.
The integrated parameters are therefore also computed for windsea and swell by only integrating over
the respective components of F (f, θ) that satisfies (1) or not (for windsea and total swell) or by integrating
over the part of the spectrum that has been identified to belong to the first, second or third swell partition.
Let us define the moment of order n of F , mn as the integral
Z Z
mn = df dθ f n F (f, θ) (2)

1
 NORMALISED 2-D SPECTRUM for 0001 wave od NORMALISED 2-D SPECTRUM for 0001 wave od
06:00Z on 27.03.2016 06:00Z on 27.03.2016
at P0011 ( 66.00 , -10.00 ) at P0001 ( 66.00 , 0.00 )
Hs= 4.17 m, Tm= 10.50 s, Tp= 12.29 s Hs= 3.66 m, Tm= 11.10 s, Tp= 13.51 s
Peakedness Qp = 1.23, Directional Spread = 0.99 Peakedness Qp = 1.53, Directional Spread = 0.68
MWD = 156 degrees PWD = 170 degrees MWD = 36 degrees PWD = 60 degrees
Propagation direction is with respect to North Propagation direction is with respect to North
North is pointing upwards North is pointing upwards
Concentric circles are every 0.05 Hz Concentric circles are every 0.05 Hz

1.00 1.00

0.96 0.96

0.64 0.64

0.43 0.43

Normalised spectrum

Normalised spectrum
0.28 0.28

0.19 0.19

0.13 0.13

0.08 0.08

0.06 0.06

0.04 0.04

0.00 0.00

21 15

18
12
E(f) (m^2 s)

E(f) (m^2 s)
15

9
12

9
6

6
3
3

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Frequency (Hz) Frequency (Hz)

(a) point 11 (b) point 1

NORMALISED 2-D SPECTRUM for 0001 wave od NORMALISED 2-D SPECTRUM for 0001 wave od
06:00Z on 27.03.2016 06:00Z on 27.03.2016
at P0012 ( 64.00 , -10.00 ) at P0002 ( 64.00 , 0.00 )
Hs= 4.09 m, Tm= 10.62 s, Tp= 12.29 s Hs= 4.35 m, Tm= 9.57 s, Tp= 13.51 s
Peakedness Qp = 0.99, Directional Spread = 0.92 Peakedness Qp = 0.94, Directional Spread = 0.68
MWD = 84 degrees PWD = 50 degrees MWD = 16 degrees PWD = 70 degrees
Propagation direction is with respect to North Propagation direction is with respect to North
North is pointing upwards North is pointing upwards
Concentric circles are every 0.05 Hz Concentric circles are every 0.05 Hz

1.00 1.00

0.96 0.96

0.64 0.64

0.43 0.43
Normalised spectrum

Normalised spectrum
0.28 0.28

0.19 0.19

0.13 0.13

0.08 0.08

0.06 0.06

0.04 0.04

0.00 0.00

18

12
15
E(f) (m^2 s)

E(f) (m^2 s)

12 9

9
6

3
3

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Frequency (Hz) Frequency (Hz)

(c) point 12 (d) point 2

NORMALISED 2-D SPECTRUM for 0001 wave od NORMALISED 2-D SPECTRUM for 0001 wave od
06:00Z on 27.03.2016 06:00Z on 27.03.2016
at P0013 ( 62.00 , -10.00 ) at P0003 ( 62.00 , 0.00 )
Hs= 3.98 m, Tm= 10.50 s, Tp= 13.51 s Hs= 6.08 m, Tm= 9.50 s, Tp= 10.15 s
Peakedness Qp = 1.19, Directional Spread = 0.74 Peakedness Qp = 1.18, Directional Spread = 0.64
MWD = 75 degrees PWD = 60 degrees MWD = 354 degrees PWD = 340 degrees
Propagation direction is with respect to North Propagation direction is with respect to North
North is pointing upwards North is pointing upwards
Concentric circles are every 0.05 Hz Concentric circles are every 0.05 Hz

1.00 1.00

0.96 0.96

0.64 0.64

0.43 0.43
Normalised spectrum

Normalised spectrum

0.28 0.28

0.19 0.19

0.13 0.13

0.08 0.08

0.06 0.06

0.04 0.04

0.00 0.00

40
15
E(f) (m^2 s)

E(f) (m^2 s)

12 30

9
20

10
3

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Frequency (Hz) Frequency (Hz)

(e) point 13 (f) point 3

Figure 1: Normalised 2d spectra (top of each panel) and frequency spectra (bottom of each panel) on 23 March
2016, 6 UTC, for locations shown in Figure 3. The 2-D spectra were normalised by their respective maximum value.
The concentric circles in the polar plots are spaced every 0.05 Hz

2
 NORMALISED 2-D SPECTRUM for 0001 wave od NORMALISED 2-D SPECTRUM for 0001 wave od
06:00Z on 27.03.2016 06:00Z on 27.03.2016
at P0014 ( 60.00 , -10.00 ) at P0004 ( 60.00 , 0.00 )
Hs= 4.24 m, Tm= 10.43 s, Tp= 13.51 s Hs= 6.48 m, Tm= 9.41 s, Tp= 11.17 s
Peakedness Qp = 1.19, Directional Spread = 0.72 Peakedness Qp = 1.71, Directional Spread = 0.49
MWD = 78 degrees PWD = 60 degrees MWD = 353 degrees PWD = 0 degrees
Propagation direction is with respect to North Propagation direction is with respect to North
North is pointing upwards North is pointing upwards
Concentric circles are every 0.05 Hz Concentric circles are every 0.05 Hz

1.00 1.00

0.96 0.96

0.64 0.64

0.43 0.43

Normalised spectrum

Normalised spectrum
0.28 0.28

0.19 0.19

0.13 0.13

0.08 0.08

0.06 0.06

0.04 0.04

0.00 0.00

18 50

15
E(f) (m^2 s)

E(f) (m^2 s)
40

12
30
9

20
6

10
3

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Frequency (Hz) Frequency (Hz)

(a) point 14 (b) point 4

NORMALISED 2-D SPECTRUM for 0001 wave od NORMALISED 2-D SPECTRUM for 0001 wave od
06:00Z on 27.03.2016 06:00Z on 27.03.2016
at P0015 ( 58.00 , -10.00 ) at P0005 ( 58.00 , 0.00 )
Hs= 4.45 m, Tm= 11.01 s, Tp= 13.51 s Hs= 5.23 m, Tm= 8.18 s, Tp= 9.23 s
Peakedness Qp = 1.32, Directional Spread = 0.82 Peakedness Qp = 1.52, Directional Spread = 0.52
MWD = 87 degrees PWD = 70 degrees MWD = 357 degrees PWD = 350 degrees
Propagation direction is with respect to North Propagation direction is with respect to North
North is pointing upwards North is pointing upwards
Concentric circles are every 0.05 Hz Concentric circles are every 0.05 Hz

1.00 1.00

0.96 0.96

0.64 0.64

0.43 0.43
Normalised spectrum

Normalised spectrum
0.28 0.28

0.19 0.19

0.13 0.13

0.08 0.08

0.06 0.06

0.04 0.04

0.00 0.00

21 30

18 25
E(f) (m^2 s)

E(f) (m^2 s)

15
20

12
15
9
10
6

5
3

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Frequency (Hz) Frequency (Hz)

(c) point 15 (d) point 5

NORMALISED 2-D SPECTRUM for 0001 wave od NORMALISED 2-D SPECTRUM for 0001 wave od
06:00Z on 27.03.2016 06:00Z on 27.03.2016
at P0016 ( 56.00 , -10.00 ) at P0006 ( 56.00 , 0.00 )
Hs= 4.19 m, Tm= 11.78 s, Tp= 13.51 s Hs= 3.09 m, Tm= 6.68 s, Tp= 7.63 s
Peakedness Qp = 1.65, Directional Spread = 0.66 Peakedness Qp = 1.34, Directional Spread = 0.57
MWD = 71 degrees PWD = 70 degrees MWD = 358 degrees PWD = 330 degrees
Propagation direction is with respect to North Propagation direction is with respect to North
North is pointing upwards North is pointing upwards
Concentric circles are every 0.05 Hz Concentric circles are every 0.05 Hz

1.00 1.00

0.96 0.96

0.64 0.64

0.43 0.43
Normalised spectrum

Normalised spectrum

0.28 0.28

0.19 0.19

0.13 0.13

0.08 0.08

0.06 0.06

0.04 0.04

0.00 0.00

24

6
18
E(f) (m^2 s)

E(f) (m^2 s)

4
12

6 2

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Frequency (Hz) Frequency (Hz)

(e) point 16 (f) point 6

Figure 2: Normalised 2d spectra (top of each panel) and frequency spectra (bottom of each panel) on 23 March
2016, 6 UTC, for locations shown in Figure 3. The 2-D spectra were normalised by their respective maximum value.
The concentric circles in the polar plots are spaced every 0.05 Hz

3
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Model bathymetry
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Model bathymetry
expver= 0001, Stand alone wave model,
Shading: Model bathymetry

11 1 10000
900

12 2 700

500
13 3
300

60°N 14 4 100

Model bathymetry
80
15 5
60

16 6 40

20

10°W 0°E

(a)

Figure 3: Model bathymetry (m) and locations for wave spectra (red numbers) in Figures 1 and 2.

4
and define the frequency spectrum E(f ) as
Z
E(f ) = dθ F (f, θ) (3)

The integrations are performed over all frequencies and directions or over a spectral sub-domain when the
spectrum is split between windsea and swell or partitioned into main components. In the high-frequency
range the usual Phillips spectral shape (f −5 ) is used where the Phillips parameter is determined by the
spectral level at the last discretised frequency bin. Then the relevant integral parameters are:

3.1 Significant Wave Height

By definition, the significant wave height Hs is defined as
√
Hs = 4 m0 (4)

Hence the definition of parameters 229. As an example, Figure 4 shows the significant wave height, mean
wave direction and energy mean wave period corresponding to the synoptic situation as shown in Figure 5.
When the spectrum is split between windsea and total swell using (1), the respective significant wave
height can be obtained, hence 234, 237 (Figure 9). If the total swell spectrum is partitioned into its 3 main
components, the integrals over the respective domain yield 121, 124, 127 (see below). A simpler approach
into the detection of low frequency waves is to integrate the spectum only for all spectral components with
frequency below 0.1 Hz (i.e. with periods above 10s) and to convert this into a corresponding significant
wave height for all waves components with period above 10 seconds, H10 , hence parameter 120. It is quite
common to plot the square of ratio of H10 to Hs (namely the ratio of the wave energy of all waves with
periods larger than 10s to the total wave energy) as shown in Figure 6.

3.2 Mean Periods

The mean period Tm−1 is based on the moment of order -1, that is

Tm−1 = m−1 /m0 (5)

Hence the definition of parameters 232, 236, 239, and the partitioned 123, 126, 129.
Tm−1 is also commonly known as the energy mean wave period. Together with Hs , it can be used to
determine the wave energy flux per unit of wave-crest length in deep water, also known as the wave power
per unit of wave-crest length P :
ρw g 2
P = Tm−1 Hs2 (6)
64π
where ρw is the water density and g the acceleration due to gravity.
In order to look at different aspects of the wave field, other moments can be used to define a mean period.
Periods can be based on the first moment Tm1 given by

Tm1 = m0 /m1 (7)

Hence the definition of parameters 220, 223, and 226. Tm1 is essentially the reciprocal of the mean frequency.
It can be used to estimate the magnitude of Stokes drift transport in deep water (see Stokes drift)
and periods based on the second moment Tm2 given by
p
Tm2 = m0 /m2 (8)

Hence the definition of parameters 221, 224, and 227. Tm2 is also known as the zero-crossing mean wave
period as it corresponds to the mean period that is determined from observations of the sea surface elevation
using the zero-crossing method.

3.3 Peak Period

The peak period is defined only for the total sea. It is defined as the reciprocal of the peak frequency. It
is obtained from a parabolic fit around the discretised maximum of two-dimensional wave spectrum, Hence
parameter 231 (Figure 7).

5
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Mean wave period/Mean wave direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell
expver= 0001, Stand alone wave model,
Shading: Significant Wave Height, Arrows: (intensity:Mean Energy Wave Period, direction:Mean Wave direction)

15 s
15.0

Significant Wave Height and Mean Energy Wave Period

13.0

11.0

60°N 9.0

7.0

5.0

10.0

8.0

6.0

4.0

2.0

0.0

10°W 0°E

(a)

Figure 4: Signifcant wave height (Hs ) (colour shading), Mean Wave Direction (arrow direction) and Mean Wave
Period (Tm−1 ) (arrow length and colour) on 23 March 2016, 6 UTC.

6
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m 10 metre wind speed/10 metre wind direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m 10 metre wind speed
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m 10 metre wind speed
expver= 0001, Stand alone wave model,
Shading: 10m neutral wind speed, Arrows: (intensity: 10m neutral wind speed, direction: 10m wind direction)

22 m/s
22
20

10m neutral wind speed and 10m neutral wind speed

16

12
60°N
8

0
22
20

16

12

10°W 0°E

(a)

Figure 5: 10m neutral Wind Speed (colour shading, arrow length and grey scale), 10m Wind Direction (arrow
direction) on 23 March 2016, 6 UTC.

7
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell/Mean wave direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant wave height of all waves with period larger than 10s
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant wave height of all waves with period larger than 10s
expver= 0001, Stand alone wave model,
Shading: Square of the ratio of H10 to Hs , Arrows: (intensity: Significant Wave Height (Hs), direction: Mean Wave Direction)

10 m

and Significant Wave Height (Hs)

10.0

8.0

6.0

60°N
4.0

2.0

0.0

Square of the ratio of H10 to Hs

0.9
0.8

0.6

0.4

0.2

0.0

10°W 0°E

(a)

Figure 6: Square of the ratio of H10 to Hs (colour shading), Significant Wave Height (Hs ) (arrow length and colour)
on 23 March 2016, 6 UTC.

8
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell/Mean wave direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Peak period of 1D spectra
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Peak period of 1D spectra
expver= 0001, Stand alone wave model,
Shading: Peak Period, Arrows: (intensity: Significant Wave Height, direction: Mean Wave direction)

11 m
11

Peak Period and Significant Wave Height

5
60°N
3

18

16

14

10

10°W 0°E

(a)

Figure 7: Peak Period (Tp ) (colour shading), Mean Wave Direction (arrow direction) and Significant Wave Height
(Hs ) (arrow length and colour) on 23 March 2016, 6 UTC.

9
3.4 Mean Wave Direction
By weighting F (f, θ), one can also define a mean direction hθi as

hθi = atan(SF/CF ) (9)

where SF is the integral of sin(θ) F (f, θ) over f and θ and CF is the integral of cos(θ) F (f, θ) over f and θ.
Hence the definition of parameters 230, 235, 238, and the partitioned 122, 125, 128. Note that in grib 1, the
direction parameters are encoded using the meteorological convention (0 means from North, 90 from East).

3.5 Wave Directional Spread

Information on the directional distribution of the different wave components can be obtained from the mean
directional spread σθ given by p
σθ = 2(1 − M1 ) (10)
where
For total sea:
M1 = I1 /E0 (11)
I1 is the integral of cos(θ −hθi(f )) F (f, θ) over f and θ, where hθi(f ) is the mean direction at frequency
f:
hθi(f ) = atan(sf (f )/cf (f )) (12)
with sf (f ) the integral of sin(θ) F (f, θ) over θ only and cf (f ) is the integral of cos(θ) F (f, θ) over θ
only. Hence the definition of parameter 222.

For wind waves and swell:

M1 = Ip /E(fp ) (13)
Ip is the integral of cos(θ − hθi(fp )) F (fp , θ) over θ only, fp is the frequency at the spectral peak and
hθi(fp ) is given by (12), where F (fp , θ) is still split in all calculations using (1) (including in E(fp )).
Hence the definition of parameters 225, 228.
√
Note: As defined by (10), the mean directional spread √ σθ takes values between 0 and 2, where 0
corresponds to a uni-directional spectrum (M1 = 1) and 2 to a uniform spectrum (M1 = 0).
Figure 8 shows the directional spread for total sea, mean wave direction and zero-crossing mean wave
period corresponding to the synoptic situation as shown in Figure 5. The plot highlights areas where the
sea state is composed of different wave systems as shown in Figures 1 and 2.

3.6 Spectral partitioning

Traditionally, the wave model has separated the 2D-spectrum into a windsea and a total swell part (see
above). Figure 9 shows the windsea and total swell significant wave height and mean wave direction corre-
sponding to the synoptic situation as shown in Figure 5. However, in many instances, the swell part might
actually be made up of different swell systems. Comparing the wave spectra in Figures 1 and 2 with the
simple decomposition shown in Figure 9, it is clear that for many locations, the total swell is made up of
more that one distinct wave system.
We have adapted and optimised the spectral partitioning algorithm of Hanson and Phillips (2001) to
decompose the SWELL spectrum into swell systems. It uses the fact that the spectra are model spectra
for which a high frequency tail has been imposed and it excludes from the search the windsea part (the
original partitioning method decomposes the full two-dimensional spectrum). The three most energetic swell
systems are retained (for most cases, up to 3 swell partitions was found to be enough) and the spectral
variance contained in the other partitions (if any) is redistributed proportionally to the spectral variance
of the three selected partitions. Because it is only the swell spectrum that is partitioned, some spectral
components can end up being unassigned to any swell partitions, their variance should be assigned to the
windsea if they are in the wind directional sector, but it is currently NOT done because the old definition
of the windsea and the total swell was not modified, otherwise they are redistributed proportionally to the
spectral variance of the three selected partitions.
Based on the partitioned spectrum, the corresponding significant wave height (4), mean wave direction
(9), and mean frequency (5) are computed.
So, by construct, the 2D-spectrum is decomposed into windsea (using the old defitinition) and up to three
swell partitions, each described by significant wave height, mean wave period and mean direction. Figure 10
shows how the partitioning has decomposed the spectra. Compared to Figure 9, it can be seen that the new

10
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Mean wave period based on second moment/Mean wave direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Wave spectral directional width
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Wave spectral directional width
expver= 0001, Stand alone wave model,
Shading: Wave spectral directional width, Arrows: (intensity: Mean zero-crossing Wave Period, direction: Mean Wave direction)

14 -

Wave spectral directional width and Mean zero-crossing Wave Period

14.0

12.0

10.0

8.0
60°N
6.0

4.0

1.4

1.2

1.0

0.8

0.6

0.4

0.2

0.0

10°W 0°E

(a)

Figure 8: Directional spread for total sea (colour shading), Mean Wave Direction (arrow direction) and Mean Wave
Period (Tm2 ) (arrow length and colour) on 23 March 2016, 6 UTC.

11
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of total swell/Mean direction of total swell
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of wind waves/Mean direction of wind waves
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell
Stand alone wave model, Contours: Significant wave height.
Arrows: significant wave height of windsea (crossed triangle), total swell (full triangle)

6
5.5 5 4
11 1 4
4.5

2.5 m
3.5
12 3.5 2 7

5
3 4

Significant Wave Height of each wave system (m)

3 6.5
13 3
6
6

5.5
6.5

2.5
3.5 5.5
60°N 14 4

4.5
5 5
5 .5
5
3.

4.5 4 6.5
4

4.5
15 5
3.5 3
3 4

2
16 6
2.5
3.5
3.5

3
1.5 1
2.5
3
1.5

1.5

1
4.5 4 2
2 0.5

10°W 0°E

(a)

Figure 9: Windsea Significant Wave Height (arrow length and colour with open crossed triangle head), TOTAL
swell Significant Wave Height (arrow length and colour with full triangle head), and significant wave height (black
contours) on 23 March 2016, 6 UTC. The red numbers are the locations for wave spectra in Figures 1 and 2.

12
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant wave height of third swell partition/Mean wave direction of third swell partition
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant wave height of second swell partition/Mean wave direction of second swell partition
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant wave height of first swell partition/Mean wave direction of first swell partition
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of wind waves/Mean direction of wind waves
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell
Stand alone wave model, Contours: Significant wave height.
Arrows: significant wave height of windsea (crossed triangle), swell 1 (full triangle), swell 2 (open triangle), swell3 (chevron)

6
5.5 5 4
11 1 4

4.5
2.5 m
3.5
12 3.5 2 7

5
3

Significant Wave Height of each wave system (m)

3 6.5
13 3
6
6

5.5
6.5

2.5
3.5 5.5
60°N 14 4

4.5
5 5
5 .5

5
3.
4.5 4 6.5
4

4.5
15 5
3.5 3
3 4

2
16 6

2.5
3.5
3.5

3
1.5 1
2.5
3

1.5
2

1.5

1
4.5 4 2
2 0.5

10°W 0°E

(a)

Figure 10: Windsea Significant Wave Height (arrow length and colour with open crossed triangle head), Primary
swell Significant Wave Height (arrow length and colour with full triangle head), Secondary swell Significant Wave
Height (arrow length and colour with open triangle head), Tertiary swell Significant Wave Height (arrow length and
colour with chevron head), and significant wave height (black contours) on 23 March 2016, 6 UTC. The red numbers
are the locations for wave spectra in Figures 1 and 2.

partitioning scheme gives a better representation of the full 2-D spectra as given in Figures 1 and 2. Note
that the swell partitions are labelled first, second and third based on their respective wave height. Therefore,
there isn’t any guarantee of spatial coherence (first might be from one system at one location and another
one at the neighbouring location). It is ONLY by taking the windsea and the 3 partitioned swell systems
that one can reconstruct the main feature of the 2D-spectrum! This is obviously an approximation as the
true sea state is only entirely described by the 2-D spectrum.

3.7 Mean Square Slope

An integrated parameter which can be related to the average slope of the waves is the mean square slope
which is only defined for the total sea as the integral of k 2 F (f, θ) over f and θ, where k is the wave number
as given by the linear dispersion relation. Hence parameter 244 (Figure 11).

13
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m 10 metre wind speed/10 metre wind direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Mean square slope of waves
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Mean square slope of waves
expver= 0001, Stand alone wave model,
Shading: Mean square slope, Arrows: (intensity: 10m neutral wind speed, direction: 10m wind direction)

22 m/s
22.00

18.00

Mean square slope and 10m neutral wind speed

14.00

10.00

60°N 6.00

2.00

0.08

0.07

0.06

0.05

0.04

0.03

0.02

0.01

0.00

10°W 0°E

(a)

Figure 11: Mean Square Slope (colour shading), 10m Wind (arrows) on 23 March 2016, 6 UTC.

14
4 Forcing fields
4.1 10 m Neutral Wind Speed
Due to different spatial grids, the forcing 10m neutral winds are interpolated to the wave model grid.
Furthermore, in case of analysed fields, the radar altimeter data assimilation scheme is such that it produces
increments for wave heights but also for wind speeds. Hence, the wind speed which is actually seen by the
wave model (U10 ) is different than the 10 m neutral wind speed provided by the atmospheric model.

By definition, the air-side friction velocity u∗ is related to the norm of the atmospheric surface stress
k~τa k
q
k~τa k = τx2 + τy2 = ρair u2∗ (14)

where ρair is the surface air density, and (τx , τy ) are the x- and y-components of the atmospheric surface
stress.
Hence,
h i 41
τx2 + τy2
u∗ = √ (15)
ρair
The norm of the vertical neutral wind profile (Uz , Vz ) is defined as

~ u∗  z 
kU(z)k = ln (16)
κ z0
where κ is the von Kármán constant, and z0 the surface roughness length scale for momentum.

τx , τy , z0 are prognostic variables in the IFS, whereas the surface air density ρair can be determined from
values of pressure, temperature and humidity at the lowest model level (see 21).

By definition, the neutral winds are in the direction of the surface stress, namely

~ τx ~ τy
Uz = kU(z)k , Vz = kU(z)k (17)
k~τa k k~τa k

Using (15), with (16) and (14) yields

τy
√ τx √
1 ρair
z 1 ρair
z
Uz = h 1 ln , Vz = h ln
1 (18)
κ i4 z0 κ i 4 z0
τx2 + τy 2 τx2 + τy 2

This can be re-written as

τx τy
1 ρair
z 1 ρair
z
Uz = h 1 ln , Vz = h 1 ln (19)
κ τx 2

τy
2 i 4 z0 κ τx 2

τy
2 i
4 z0
ρair + ρair ρair + ρair

This last form is what is coded in the IFS.

At every coupling time, (19) is used for z = 10m (U10 , V10 ) with the updated values for τx , τy , z0 and
ρair to provide the neutral 10m wind that is used to force ECWAM. These components are interpolated onto
the ECWAM grid and archived as wave model parameters as magnitude and direction (parameters 245 and
249) (Figure 5).
Over the ocean, z0 is itself a function of the sea state (see the IFS Documentation). Because the IFS
and ECWAM do not share the same grid, nor the same land-sea mask, there will be some values of τx , τy ,
z0 and ρair that correspond to land values. Nevertheless, the contribution from the Turbulent Orographic
Form Drag (TOFD) parametrisation is substracted from the total value of the surface stress prior to apply
(19) because it is intended to only be valid over land.

4.2 Drag Coefficient

In the wave model, the surface stress depends on the waves. This feature is archived via the drag coefficient
Cd (parameter 233) which relates the surface stress to the square of the neutral wind speed, u2∗ = Cd kU10 k2
(Figure 12). Note that currently, u∗ is not archived as such and needs to be computed with the previous
relation.

15
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m 10 metre wind speed/10 metre wind direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m Coefficient of drag with waves
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m Coefficient of drag with waves
expver= 0001, Stand alone wave model,
Shading: Drag Coefficient, Arrows: (intensity: 10m neutral wind speed, direction: 10m wind direction)

22 m/s
22.000
20.000

16.000

Drag Coefficient and 10m neutral wind speed

12.000
60°N
8.000

4.000

0.000
0.005
0.005

0.004

0.003

0.002

0.001

0.000

10°W 0°E

(a)

Figure 12: Wave modifed drag coefficient (Cd ) (colour shading), and 10m neutral winds (arrows) on 23 March 2016,
6 UTC.

16
4.3 Free convective velocity scale and air density
Strickly speaking the free convective velocity (w∗ ) (parameter 208) and the air density (ρair ) (parameter 209)
are not wave model parameters, they are part of the atmospheric forcing. Nevertheless, they are archived on
the same model grid as the wave model, and are only available over the oceans, as defined by the wave model
land/sea mask. The free convection velocity scale w∗ is used to parameterise the impact of wind gustiness
on the wave growth which is also propotional to the ratio of air density to water density (currently assumed
constant).
The free convection velocity scale w∗ is computed using
  1/3
1 zi
w∗ = u∗ f or L < 0 and w∗ = 0 f or L >= 0 (20)
κ −L

where u∗ is the friction velocity (u2∗ = ρτair

a
), κ is the von Kármán constant, zi is the height of the lowest
inversion, L is the Monin-Obukhov length. The quantity zi /L, which is a measure for the atmospheric
stability, is readily available from the atmospheric model, and the surface air density ρair is given by
P
ρair = (21)
RTv

where P is the atmospheric pressure, R ' 287.04 J kg−1 K−1 is a constant defined as R = R+ /ma , with
R+ the universal gas constant (R+ ' 8314.36 J kmol−1 K−1 ) and ma is the molecular weight of the dry air
(' 28.966 kg kmol−1 ), and Tv is the virtual temperature. The virtual temperature can be related to the
actual air temperature, T , and the specific humidity, q, by: Tv ' (1 + 0.6078q)T . To avoid using diagnostic
variables, the pressure, the temperature and humidity at the lowest model level are now used.
Figure 13 shows a combined map of the different fields that make up the forcing to the ECMWF wave
model (U10 , ρair , w∗ ) .

17
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Free convective velocity over the oceans
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC 10 m 10 metre wind speed/10 metre wind direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Air density over the oceans
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Air density over the oceans
expver= 0001, Coupled wave model,
Shading: Air density, Countours : Free convective velocity, Arrows: (intensity: 10m neutral wind speed, direction: 10m wind direction)

1.2
0.4 0.2 0.6
1 0.6
0.8
0.6 0.8
0.8 0.8
0.6
1 0.2 0.4 24 m/s
24.00
1.4 22.50

19.50
1

Air density and 10m neutral wind speed

1.2
16.50
2
1.
1

60°N
13.50
1.4
0.

10.50
8

0.8 7.50

1 4.50
0.6

1.50

1.30

1.25

0.6 0.8
1.20
1
0.8
1.15

1.10

10°W 0°E

(a)

Figure 13: Surface air density (ρair ) (colour shading), free convection velocity scale (w∗ ) (black contours), and 10m
neutral winds (arrows) on 23 March 2016, 6 UTC.

18
 Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea U-component stokes drift/V-component stokes drift
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea U-component stokes drift
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea U-component stokes drift
expver= 0001, Stand alone wave model,
Shading: Ratio of Surface Stokes Drift to 10m Wind Speed(%), Arrows: (intensity: Surface Stokes Drift )

Ratio of Surface Stokes Drift to 10m Wind Speed(%) and Surface Stokes Drift
1.2 m/s
1.2
1.1

0.9

0.7

60°N 0.5

0.3

0.1

6.0

5.0

4.0

3.0

2.0

1.0

0.0

10°W 0°E

(a)

Figure 14: Ratio (in percentage) of the surface Stokes drift magnitude to the 10m wind speed (colour shading),
surface Stokes drift (arrow length and colour) on 23 March 2016, 6 UTC.

5 Interaction with the ocean circulation

5.1 Stokes drift
The surface Stokes drift ~ust is defined by the following integral expression
Z
2gk ~kF (f, θ)
~ust = df dθ (22)
ω tanh(2kD)

The integration is performed over all frequencies and directions. In the high-frequency range the usual
Phillips spectral shape is used where the Phillips parameter is determined by the spectral level at the last
frequency bin, while it is tacitly assumed that these frequencies are so high that shallow water effects are
unimportant. This defines parameters 215 and 216.
Figure 14 shows the ratio (in percentage) of the surface Stokes drift magnitude to the 10m wind speed
(colour shading) and the actual surface Stokes drift (colour arrows) corresponding to the synoptic situation
as shown in Figure 5 for wind. This ratio shows that the surface Stokes drift cannot easily be represented
as a fixed ratio of the 10m wind speed.

5.2 Momentum and energy flux into ocean

In order to be able to give a realistic representation of the mixing processes in the surface layer of the
ocean, a reliable estimate of energy and momentum fluxes to the ocean column is required. As energy and
momentum flux depend on the spectral shape, the solution of the energy balance equation is required. It

19
reads
∂ ∂
F+ · (~vg F ) = Sin + Snl + Sdiss + Sbot , (23)
∂t ∂~x
where F = F (ω, θ) is the two-dimensional wave spectrum which gives the energy distribution of the ocean
waves over angular frequency ω and propagation direction θ. Furthermore, ~vg is the group velocity and on
the right hand side there are four source terms. The first one, Sin describes the generation of ocean waves
by wind and therefore represents the momentum and energy transfer from air to ocean waves. The third
and fourth term describe the dissipation of waves by processes such as white-capping, large scale breaking
eddy-induced damping and bottom friction, while the second term denotes nonlinear transfer by resonant
four-wave interactions. The nonlinear transfer conserves total energy and momentum and is important in
shaping the wave spectrum and in the spectrum down-shift towards lower frequencies.
Let us first define the momentum and energy flux. The total wave momentum M ~ depends on the variance
spectrum F (ω, θ) and is defined as
Z 2π Z ∞ ~k
~ = ρw g
M dωdθ F (ω, θ), (24)
0 0 ω

where ρw is the water density and g the acceleration due to gravity. The momentum fluxes to and from the
wave field are given by the rate of change in time of wave momentum, and one may distinguish different
momentum fluxes depending on the different physical processes. For example, making use of the energy
balance equation (23) the wave-induced stress is given by
Z 2π Z ∞ ~k
~τin = ρw g dωdθ Sin (ω, θ), (25)
0 0 ω

while the dissipation stress is given by

Z 2π Z ∞ ~k
~τdiss = ρw g dωdθ Sdiss (ω, θ), (26)
0 0 ω

Similarly, the energy flux from wind to waves is defined by

Z 2π Z ∞
Φin = ρw g dωdθ Sin (ω, θ), (27)
0 0

and the energy flux from waves to ocean, Φdiss , is given by

Z 2π Z ∞
Φdiss = ρw g dωdθ Sdiss (ω, θ). (28)
0 0

It is important to note that while the momentum fluxes are mainly determined by the high-frequency part
of the wave spectrum, the energy flux is to some extent also determined by the low-frequency waves.
The prognostic frequency range is limited by practical considerations such as restrictions on computation
time, but also by the consideration that the high-frequency part of the dissipation source function is not
well-known. In the ECMWF wave model the high-frequency limit ωc = 2πfc is set as

fc = min{fmax , 2.5hf iwindsea } (29)

Thus, the high-frequency extent of the prognostic region is scaled by the mean frequency hf iwindsea of the
local windsea. A dynamic high-frequency cut-off, fc , rather than a fixed cut-off at fmax , corresponding to
the last discretised frequency, is necessary to avoid excessive disparities in the response time scales within
the spectrum.
In the diagnostic range, ω > ωc , the wave spectrum is given by Phillips’ ω −5 power law. In the diagnostic
range it is assumed that there is a balance between input and dissipation. In practice this means that all
energy and momentum going into the high-frequency range of the spectrum is dissipated, and is therefore
directly transferred to the ocean column.

Z 2π Z ∞ ~k
dωdθ (Sin + Sdiss + SN L ) = 0, (30)
0 ωc ω

and

20
 Z 2π Z ∞
dωdθ (Sin + Sdiss + SN L ) = 0, (31)
0 ωc

The momentum flux to the ocean column, denoted by ~τoc , is the sum of the flux transferred by turbulence
across the air-sea interface which was not used to generate waves ~τa −~τin and the momentum flux transferred
by the ocean waves due to wave breaking ~τdiss .
As a consequence, ~τoc = ~τa − ~τin − ~τdiss . Utilizing the assumed balance at the high-frequencies (30) and
the conservation of momentum for SN L when integrated over all frequencies and directions, one finds
Z 2π Z ωc ~k
~τoc = ~τa − ρw g dωdθ (Sin + Sdiss + SN L ) , (32)
0 0 ω

where ~τa is the atmospheric stress, whose magnitude is given by τa = ρair u2∗ , with u∗ the air side friction
velocity.
Ignoring the direct energy flux from air to ocean currents, because it is small, the energy flux to the
ocean, denoted by Φoc , is therefore given by −Φdiss . Utilizing the assumed high-frequency balance (31) and
the conservation of energy when SN L is integrated over all frequencies and directions, one therefore obtains

Z 2π Z ∞ Z 2π Z ωc
Φoc = ρw g dωdθ Sin − ρw g dωdθ (Sdiss + SN L ) , (33)
0 ωc 0 0

The high frequency (ω > ωc ) contribution to the energy flux

Z 2π Z ∞
Φochf = ρw g dωdθ Sin (34)
0 ωc

is parameterised following the same approach as for the kinematic wave induced stress (for more details
refer to the IFS documentation part VII)
2π ∞
(2π)4 fc5
Z Z
2 βm dω
Φochf = ρa u∗ 2 dθ F (fc , θ)[max (cos(θ − φ), 0)] µhf ln4 (µhf ), (35)
g 0 κ2 ωc ω2

In (35), the integral over directions can be evaluated using the prognostic part of the spectrum, whereas the
second integral is only function of u∗ and the Charnock parameter. It can therfore be tabulated beforehand.
Note that the integration is bounded because µhf <= 1

βm z0 1 dY
r Z  r
4 g z0
P HIOCHF = 2 2
µhf ln (µhf ), Yc = max ωc , x0 (36)
κ g Yc Y u∗ g
where for typical values of the Charnock parameter, x0 ∼ 0.05.
The archived energy fluxes are normalized by the product of the air density ρair and the cube of the
friction velocity in the air (u∗ ). Hence the normalized energy flux into waves (parameter 211) is obtained
from (27) divided by ρair u3∗ . Similarly the normalized energy flux into ocean (parameter 212) is obtained
from normalizing (33).
The normalized stress into ocean (parameter 214) is derived from (32) by dividing it with the atmospheric
stress τa = ρair u2∗ .
Figure 15 shows both dimensional energy and momentum fluxes into the ocean.

21
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea/heightAboveGround Normalized stress into ocean/10 metre wind direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Normalized energy flux into ocean
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Normalized energy flux into ocean
expver= 0001, Stand alone wave model,
Shading: Turbulent Energy flux into the ocean, Arrows: (intensity: Momentum Flux into the Ocean, direction: 10m Wind Direction)

Turbulent Energy flux into the ocean and Momentum Flux into the Ocean
1.6 N/m^2
1.6

1.2

0.8

60°N
0.4

0.0
11.0
10.0

8.0

6.0

4.0

2.0

0.0

10°W 0°E

(a)

Figure 15: Wave energy flux into the ocean (colour shading), and momentum flux into the ocean (arrows) on 23
March 2016, 6 UTC.

22
6 Freak wave parameters
The parameters that have been described so far all provide information on the average properties of the sea
state. In recent years, there has been a considerable effort to understand extreme events such as freak waves.
An individual wave is regarded as a freak wave when its height is larger than twice the significant wave
height. Clearly, in order to be able to describe such extreme events, knowledge on the statistical properties
of the sea surface is required. Recent work has presented a general framework that relates the shape of the
probability distribution function (pdf) of the surface elevation to the mean sea state as described by the two-
dimensional frequency spectrum. Under normal circumstances, the surface elevation pdf has approximately
a Gaussian shape, but in the exceptional circumstances that the waves are sufficiently nonlinear and that
the wave spectrum is narrow in both frequency and direction considerable deviations from Normality may
occur, signalling increased probability for freak waves.
The deviations from Normality are measured in terms of the kurtosis C4 (parameter 252) of the surface
elevation pdf. The determination of this parameter from the wave spectrum is described in Chapter 8 of
part VII of the IFS documentation, and it is shown that in the narrow-band approximation the kurtosis
depends on the Benjamin-Feir Index BF I and the directional width δω at the peak of the wave spectrum
with some correction for shallow water effects.

κ4
C4 = C4dyn + . (37)
8
where

κ4 = κ40 + κ04 + 2κ22 .

The κ’s refer to certain fourth-order cumulants of the joint pdf of the surface elevation and its Hilbert
transform (?).
κ40
κ40 = 18; κ04 = 0.; κ22 = = 3
6
√
with  the integral steepness parameter,  = k0 m0 , k0 the peak wave number, m0 the zero moment of the
spectrum.
and,
  2 
1 π vg gXnl
C4dyn =s 2
2 × 3√3 −BF I × c0 , (38)
 kω0 ω000
7 δθ
1+ 2 δω

where the relevant symbols are defined in capter 8 of the IFS documentation and the Benjamin-Feir Index
BF I is given by
√
 2
BF I = . (39)
δω
Note that the archived parameter (253) is the square of BF I.
The Benjamin-Feir Index is the ratio of the integral wave steepness  and δω the relative width of the
frequency. Initially the relative width of the frequency spectrum was solely estimated by using Goda’s
peakedness factor Qp (parameter 254) defined as
Z
2
Qp = 2 dω ωE 2 (ω) (40)
m0

with E(ω) the angular frequency spectrum and the integration domain D consists of all frequencies for
which E(ω) > 0.4 E(ωp ), with ωp the peak angular frequency.
The advantage of this integral measure is that, because of its dependence on the square of the frequency
spectrum, peaks in the spectrum are emphasized. Howver, from CY33R1 onwards, a sharper estimate of the
width in frequency and direction is obtained from a two-dimensional parabolic fit around the peak of the
spectrum. This procedure then also gives a more accurate estimation of the peak period (parameter 231).
In the operational model, the kurtosis C4 is restricted to the range −0.33 < C4 < 1.
Since CY40R3, the skewness of the pdf of the surface elevation C3 was introduced (parameter 207). This is
in particular relevant for the contribution of bound waves to the deviations of statistics from Normality, as
bound waves give rise to a considerable skewness. On the other hand, it should be noted that the skewness

23
of free waves is very small and therefore for really extreme events the skewness correction to the wave height
pdf is not so important. However, on average, the bound waves will determine the statistics of waves and,
therefore, in order to have an accurate desciption of the average conditions as well the skewness effect needs
to be included.

r
κ23
C3 = , (41)
72
where

κ23 = 5(k30
2
+ κ203 ) + 9(κ221 + κ212 ) + 6(κ30 κ12 + κ03 κ21 )

The κ’s refer to certain third-order cumulants of the joint pdf of the surface elevation and its Hilbert
transform (?).
κ30
κ30 = 3; κ03 = 0.; κ12 = 0.; κ21 = =
3
Finally, these deviations form Normality can be used to come up with an expression for the expectation
value of the maximum wave height Hmax (parameter 218).

hHmax i = hziHs , (42)

with Hs the significant wave height (4) where

p 1 n h io
hzi = ẑ0 + √ γ + log 1 + α̂C4 + β̂C32 , (43)
4 ẑ0
1
with ẑ0 = 2 log N and γ = 0.5772 is Euler’s constant
where
1
α̂ = 2ẑ0 (ẑ0 − 1) + (1 − 2ẑ0 )G1 + G2
2
and
1
β̂ = ẑ0 (4ẑ02 − 12ẑ0 + 6) − (6ẑ02 − 12ẑ0 + 3)G1 + 3(ẑ0 − 1)G2 − G3 ,
2
2
It may be shown that G1 = Γ0 (1) = −γ, G2 = Γ00 (1) = γ 2 + π6 , and G3 = Γ000 (1) = −2ζ(3) − γ 3 − γπ 2 /2).
Here, ζ(3) = 1.20206 is the Riemann zeta function with argument 3.
Finally the number of independant wave groups N in the time series of length TL has to be determined

2hc
N = √ ν ω̄TL . (44)
2π

with ω̄ = m1 /m0 the mean angular frequency,

√ while ν = (m0 m2 /m21 − 1)1/2 is the width of the frequency
spectrum and the reference level hc = 2.
It is clear that for operational applications a choice for the length of the timeseries (TL ) needs to be
made. A duration of 20 minutes (TL =1200 sec.) was selected to roughly correspond to buoy acquisition
time.
Figure 16 shows the ratio of Hmax to Hs corresponding to the synoptic situation as shown in Figure 5.

7 Miscellaneous
7.1 Radar Altimeter Data
These parameters are only for diagnostics carried out at ECMWF
Even though altimeter data are processed observations and thus not as such wave model results, their
processing has required some information from the model.
Following a quality control procedure which discards all spurious data, the raw altimeter wave height
data, which are available in a ±3 hours time window, is collocated with the closest model grid point. The

24
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Significant height of combined wind waves and swell/Mean wave direction
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Maximum individual wave height
Sunday 27 March 2016 06 UTC ecmf t+0 VT:Sunday 27 March 2016 06 UTC meanSea Maximum individual wave height
expver= 0001, Stand alone wave model,
Shading: Ratio of Hmax to Hs , Arrows: (intensity: Significant Wave Height (Hs), direction: Mean Wave Direction)

9 m
9.0

and Significant Wave Height (Hs)

7.0

5.0

60°N
3.0

1.0

2.3

2.1

Ratio of Hmax to Hs
1.9

1.7

1.5

1.4

10°W 0°E

(a)

Figure 16: Ratio of Hmax to Hs (colour shading), Significant Wave Height (Hs ) (arrow length and colour) on 23
March 2016, 6 UTC.

25
average value is computed for all grid points with at least two individual observations. The averaged data
are then archived on the same grid as all wave model fields as parameter 246.
Before these gridded altimeter wave heights are presented to the wave model assimilation scheme, cor-
rections are performed which depend on the type of Altimeter instrument. For example, however because
of a known underestimation of significant wave height by the ERS-2 satellite, which is due to the inherent
non gaussian distribution of the sea surface elevation and the method how wave height is obtained from the
waveform, a correction is derived from the model spectra which is applied to the altimeter data. Also data
from the Altimeters on board of Envisat and Jason-1 are bias corrected. The correction is obtained from a
comparison with buoy wave height data. The corrected data are used by the assimilation scheme and are
archived as parameter 247.
The altimeter range observation is also affected by the non gaussianity of the sea surface elevation. The
correction is a fraction of the observed wave height, where the fraction depends on the nonlinearity of the
sea surface. This number is also collocated with the wave model grid and archived as parameter 248.

References
J. Hanson and O. Phillips, 2001. Automated Analysis of Ocean Surface Directional Wave Spectra. J. Atmos.
Oceanic. Technol.,18, 277–293.

26
 Table 1: Archived parameters of the ECMWF wave forecasting system.

Code Mars Field Units

Figure Abbrev.
140120 SH10 Significant wave height of all waves with period larger than 10s m
140121 SWH1 Significant wave height of first swell partition m
◦
140122 MWD1 Mean wave direction of first swell partition
140123 MWP1 Mean wave period of first swell partition s
140124 SWH2 Significant wave height of second swell partition m
◦
140125 MWD2 Mean wave direction of second swell partition
140126 MWP2 Mean wave period of second swell partition s
140127 SWH3 Significant wave height of swell third partition m
◦
140128 MWD3 Mean wave direction of swell third partition
140129 MWP3 Mean wave period of swell third partition s
140207 WSS Wave Spectral Skewness -
140208 WSTAR Free convective velocity scale over the oceans m/s
140209 RHOAO Air density over the oceans kg/m3
140211 PHIAW Normalized energy flux into waves -
140212 PHIOC Normalized energy flux into ocean -
140214 TAUOC Normalized stress into ocean -
140215 UST u-component of Stokes drift m/s
140216 VST v-component of Stokes drift m/s
140217 TPM Expected Period for Hmax s
140218 HMAX Expected Hmax over 3 hour period m
140219 DPTH Bathymetry as used by operational Wave Model m
140220 MP1 Mean wave period from 1st moment s
140221 MP2 Mean wave period from 2nd moment s
140222 WDW wave spectral directional width -
140223 P1WW Mean wave period from 1st moment of wind waves s
140224 P2WW Mean wave period from 2nd moment of wind waves s
140225 DWWW Wave spectral directional width of wind waves -
140226 P1PS Mean wave period from 1st moment of total swell s
140227 P2PS Mean wave period from 2nd moment of total swell s
140228 DWPS Wave spectral directional width of total swell -
140229 SWH Significant wave height m
◦
140230 MWD Mean wave direction
140231 PP1D Peak period of 1-D spectra s
140232 MWP Mean wave period s
140233 CDWW Coefficient of drag with waves -
140234 SHWW Significant height of wind waves (windsea) m
◦
140235 MDWW Mean direction of wind waves (windsea)
140236 MPWW Mean period of wind waves (windsea) s
140237 SHPS Significant height of total swell m
◦
140238 MDPS Mean direction of total swell
140239 MPPS Mean period of total swell s
140244 MSQS Mean square slope -
140245 WIND 10 m neutral wind modified by wave model m/s
140246 AWH Gridded altimeter wave height m
140247 ACWH Gridded corrected altimeter wave height m
140248 ARRC Gridded altimeter range relative correction m
◦
140249 DWI 10 m wind direction
140251 2DFD 2-D wave spectra m2 s/rad
140252 WSK Kurtosis -
140253 BFI The square of the Benjamin-Feir Index -
140254 WSP Goda’s Peakedness parameter -

27