Ocean waves in geosciences

parts 1 and 2: general wave topics from deep to shallow water

10 c ros
s -tra
35 20 ck (
30
km) wave parameters vs latitude
30
km)

40

25
(a) (b) (c)
ck (

20
g-tra

upwave
15 jet
alon

10 local
current jet

5
N
Vsa SOFS mooring
0 t

0.10 Loo
k
0.05
surf 0.00
ace
elev
atio 0.05
n (m 0.10
)

Fabrice Ardhuin,

Laboratoire d’Océanographie Physique et Spatiale, Brest, France

doi: 10.13140/RG.2.2.16019.78888/12 , https://github.com/ardhuin/waves_in_geosciences

August 19, 2024

2

On the cover: Left: swell waves from extratropical storm Rosemary resolved by the Surface Water
Ocean Topography satellite, at the location of the Southern Ocean Flux Station mooring (Hay et al.,
2023), South of Australia, along track 19, on 9 June 2023. Center: from the resolved surface elevation
map, one can estimate swell wave parameters, here for waves with periods longer than 18 s, and investigate
their evolution along the satellite track. Such long waves waves are generated by the winds of the most
powerful storm, and propagate across ocean basins where they are are also influenced by currents and sea
ice. The right panel gives more context for these wave observations using the Ocean Virtual Laboratory
developed by Oceandatalab: SWOT-derived currents confirm the presence of a current jet at 51◦ S, part
of a southward meander of the Antarctic Circumpolar Current (ACC). Other satellite altimeters, here
only SARAL is shown, also confirm the wave height reductions associated to refraction of the westerly
waves in the ACC (https://odl.bzh/WjEKPhZa). More details on storm Rosemary and the capability
of the SWOT satellite mission can be found in Ardhuin et al. (2024).
Contents

I Waves in deep water 3

1 Introduction 5
1.1 Waves in geosciences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Wave motion: some observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Wave-by-wave analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.1 Time series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2 Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2 Main properties of linear waves 13

2.1 Waves: a question of gravity, pressure, mass and vorticity . . . . . . . . . . . . . . . . . . 13
2.2 Wave mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3 Eulerian equations for wave motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Small slope waves over a flat bottom: the Airy solution . . . . . . . . . . . . . . . . . . . 16
2.4.1 Solution: Laplace equation and vertical profiles . . . . . . . . . . . . . . . . . . . . 16
2.4.2 Solution: Momementum balance and dispersion relation . . . . . . . . . . . . . . . 17
2.4.3 Solution: polarization relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.4.4 Physical interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.4.5 Kinematics: influence of the non-dimensional depth kD . . . . . . . . . . . . . . . 19
2.4.6 And in the air? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.7 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.4.8 Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4.9 Energy and power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.4.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.4.11 Extending Airy wave theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3 Wave heights and spectra: theory and measurement 31

3.1 Frequency spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.1 Wavenumber or frequency? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Using spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2.1 Transfer function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2.2 Spectral and integral parameters: Hs , Tp ... . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Random waves in situ observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.1 Wave gauges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3.2 Wave buoys . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3.3 Pressure gauges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.4 P-U-V sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.5 Sensor arrays and ADCPs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.4 Optical measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4.1 From stereo-photography to stereo-video . . . . . . . . . . . . . . . . . . . . . . . . 41
3.4.2 Using polarization and/or light intensity . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.3 Grazing angle radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.4.4 Surface wave radar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

iii
iv CONTENTS

4 Practical estimation of the wave spectrum 45

4.1 General properties of discrete Fourier transforms . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.1 Spectral resolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.2 Nyquist frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.1.3 Co-spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2 Spectra from time series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2.1 Filtering of data and aliasing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2.2 Windows, Gibbs phenomenon and averaging . . . . . . . . . . . . . . . . . . . . . . 47
4.2.3 Interpretations and further developments . . . . . . . . . . . . . . . . . . . . . . . 48
4.3 Spectral analysis of directional buoy data . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.3.1 Case of 3-axis displacements or accelerations . . . . . . . . . . . . . . . . . . . . . 50
4.3.2 Case of other systems with 3 variables or more . . . . . . . . . . . . . . . . . . . . 51
4.4 Some links between spectral and wave-by-wave analysis . . . . . . . . . . . . . . . . . . . 52

5 Wave groups and fluctuations of wave parameters 53

5.1 Wave envelope, local amplitudes and their statistics . . . . . . . . . . . . . . . . . . . . . 53
5.1.1 Time series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.1.2 Spatial maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.1.3 Along-track variability in altimeter data . . . . . . . . . . . . . . . . . . . . . . . . 56

6 Wave data from satellites: Skylab (1975) to SWOT (2025) 61

6.1 Radar remote sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.1 Conventional or ‘delay’ altimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.2 delay-Doppler altimetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.1.3 Synthetic Aperture Radars (SARs) . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.1.4 SAR Across-track Interferometry: SWOT and near-nadir range bunching effects . 66
6.1.5 The wave spectrometer and the matching wavefront technique . . . . . . . . . . . 67

7 Measured wave evolution: main parameters and wave spectra 69

7.1 Wind-sea growth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.1.1 Full development and wave age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.1.2 Fetch limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.1.3 Time limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.1.4 Double limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.2 Swell . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.3 Frequency spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.3.1 The early days . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
7.3.2 The modern era . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
7.4 Directional spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.4.1 Early parameterizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
7.4.2 Bimodality of the directional spectrum . . . . . . . . . . . . . . . . . . . . . . . . . 78
7.4.3 Swell spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.5.1 Important parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
7.5.2 Spectral shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

8 Physics of spectral wave evolution: deep water 81

8.1 Generation of waves by the wind . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
8.1.1 Measuring and parameterizing wind-wave generation . . . . . . . . . . . . . . . . . 83
8.1.2 Parameterizations based on observations . . . . . . . . . . . . . . . . . . . . . . . . 84
8.1.3 Short waves and multiple-scale interactions . . . . . . . . . . . . . . . . . . . . . . 84
8.1.4 In summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
8.2 Weakly non-linear evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
8.2.1 Wave-wave interation theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.2.2 Conservation properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.2.3 Other properties of wave-wave interactions . . . . . . . . . . . . . . . . . . . . . . 88
8.2.4 Practical calculation of wave-wave interactions . . . . . . . . . . . . . . . . . . . . 89
8.3 Dissipation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
8.3.1 A classification of breaking waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
CONTENTS v

8.3.2 Parameterizations of dissipation due to breaking in shallow water . . . . . . . . . . 92

8.3.3 Breaking wave statistics and dissipation rates in deep water . . . . . . . . . . . . . 93
8.3.4 A spectral approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
8.4 Spectral energy balance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95

9 Waves and momentum 99

9.1 Stokes drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
9.2 The ‘shear’ of the Stokes drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
9.3 Random waves and practical estimation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
9.4 Radiation stresses and the flux of wave momtum . . . . . . . . . . . . . . . . . . . . . . . 102

10 Wave-current interactions 105

10.1 Current effects on waves: propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
10.1.1 A uniform current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
10.1.2 Effect of current changes along the propagation direction . . . . . . . . . . . . . . 105
10.1.3 Effects of current gradients in the transverse direction:
rays and refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
10.1.4 Waves over vertically sheared currents . . . . . . . . . . . . . . . . . . . . . . . . . 108
10.1.5 Practical importance of currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
10.2 Wave effects on currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
10.2.1 Mean flow equations integrated over the vertical . . . . . . . . . . . . . . . . . . . 110
10.2.2 Total energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
10.3 Energy exchange and wave action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

11 Interactions of waves and sea ice 115

11.1 From water to pancakes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
11.1.1 Suspensions of frazil cristals and effective viscosity . . . . . . . . . . . . . . . . . . 116
11.1.2 Pancake ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
11.2 Dispersion relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
11.3 Dissipation for solid floes: basal friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
11.4 Dissipation for solid floes: flexure and the importance of ice break-up . . . . . . . . . . . . 119
11.5 Scattering of waves by sea ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
11.6 Waves and icebergs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

12 Numerical modeling in deep water 123

12.1 Forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
12.1.1 Bathymetry and islands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
12.1.2 Wind fields: analyses, forecasts, re-analyses . . . . . . . . . . . . . . . . . . . . . . 124
12.1.3 Coastal winds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
12.1.4 Observed winds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
12.1.5 Currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.1.6 Sea ice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.2 Numerics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.2.1 Wave models are big . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.2.2 Spectral discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
12.2.3 Spatial discretization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
12.2.4 Integration methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
12.2.5 Source term integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
12.2.6 Spatial propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
12.3 Parameterizations of physical processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
12.3.1 A model development and validation strategy . . . . . . . . . . . . . . . . . . . . . 132
12.3.2 Overall performance for common parameters . . . . . . . . . . . . . . . . . . . . . 133
12.3.3 Parameterization of the dissipation Sds . . . . . . . . . . . . . . . . . . . . . . . . 133
12.3.4 Beyond Hs , Tp ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135

13 Extreme waves and historical storms 137

13.1 Extreme significant wave heights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
13.2 Physical processes at high wind speeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
13.3 Variability and trends of sea state parameters . . . . . . . . . . . . . . . . . . . . . . . . . 139
vi CONTENTS

14 Air-sea interactions:
wind stress and mixing 141
14.1 Sea state influence on air-sea fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
14.1.1 Wind stress and drag coefficient . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
14.1.2 Swell and stress direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
14.1.3 Other effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
14.2 Drift and mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
14.2.1 Momentum flux for the Eulerian mean current . . . . . . . . . . . . . . . . . . . . 143
14.2.2 Quasi-Eulerian currents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
14.2.3 Stokes drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
14.3 Langmuir circulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

15 Waves and ocean remote sensing 149

15.1 Back-scatter from a rough surface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
15.1.1 near-nadir incidences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
15.1.2 higher incidence angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
15.2 Various applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
15.2.1 Roughness and surface current gradients . . . . . . . . . . . . . . . . . . . . . . . . 152

II Waves in coastal and nearshore environments 155

16 Linear shoaling, refraction and reflection 157
16.1 Wave shoaling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
16.2 Refraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
16.3 Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
16.4 Reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
16.4.1 Reflection at the shoreline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
16.4.2 Reflection by underwater topography . . . . . . . . . . . . . . . . . . . . . . . . . . 165
16.4.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166

17 Nonlinear wave shoaling 167

17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
17.2 Boussinesq and KdV equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
17.3 Wave evolution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
17.3.1 Waves over a flat bottom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
17.3.2 Sloping bottom and the Iribarren number . . . . . . . . . . . . . . . . . . . . . . . 170
17.3.3 Wave shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
17.3.4 Wave spectra and bi-spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
17.4 Infragravity waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
17.4.1 Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
17.4.2 Theories for IG waves generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
17.4.3 Free wave radiation from the coast . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

18 Bottom boundary layer: processes and parameterizations 177

18.1 Viscous solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
18.2 Streaming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
18.3 Turbulent boundary layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
18.4 Bottom roughness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
18.4.1 Sand and ripples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
18.4.2 Rocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
18.5 Joint wave and current boundary layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185

19 Wave-driven nearshore flows: water levels and currents 187

19.1 Mean flow equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
19.2 Mean sea level: set-down and set-up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
19.2.1 Wave set-up in practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
19.3 The longshore current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
19.4 Cross-shore flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
CONTENTS vii

20 Numerical wave modelling at regional to beach scales 193

20.1 Forcing fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
20.1.1 Winds, currents and water depth . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
20.2 Numerics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
20.3 Parameterizations of physical processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
20.3.1 Bottom friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
20.3.2 Triad wave-wave interactions: sum interactions . . . . . . . . . . . . . . . . . . . . 196
20.3.3 Triad wave-wave interactions: difference interactions and infragravity wave gener-
ation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
20.4 Examples of wave model results and validation . . . . . . . . . . . . . . . . . . . . . . . . 196
viii CONTENTS
Foreword

There are countless scholarly articles and books about ocean waves, with many different points of view,
going from mathematical treatises to naval architecture. Among these we can single out the excellent
textbooks by Kinsman (1965), Dean and Dalrymple (1991), and Holthuijsen (2007), the engineering
manual from the U. S. Army Corps of Engineers (2002), and many excellent scientific monographs by
Phillips (1977), Dingemans (1997), Young (1999), Lavrenov (2003), Janssen (2004), Lannes (2013) ...
So why another one?
First of all, scientific developments never stop, making these previous works not obsolete but less up
to date and complete. This will happen with the present book, even if I am trying to update it on a
regular basis. The parts 1 and 2 and the associated teaching material (jupyter notebooks ...) is designed
to be at the level of Master students in oceanography.
Second, and more important, all these books, except possibly the jewel by Phillips (1977) have a
rather narrow scope, and do not cover aspects for which no recent monograph exist. I particularly think
about microseisms, infragravity waves or measurement techniques including satellite remote sensing.
Some of these topics are only treated in part 3.
Working with coastal engineers, geomorphologists and seismologists, has motivated me to bring to
the forefront those results that are often obscure or very hard to follow. My point of view is that ocean
waves play a very particular role in the Earth System, both as an important element of air-sea or land-
ocean exchanges, and also as a deforming mirror that modifies our measurements of ocean properties
using remote sensing and even in situ techniques. As a result, much insight and cross-fertilization can
come from the integration of many geoscientifc fields, from microseisms to remote sensing, as well as
applied disciplines such as marine meteorology or coastal and ocean engineering. At the very least, these
different disciplines are providing new data, tools, and different points of view that complement each
other in constraining our physical understanding of wave processes, and the parameterizations used in
numerical models or remote sensing algorithms.
On these topics, I have tried to be clear without compromising the accuracy of the results, but this is
a very difficult balance. If you find it unclear, do not hesitate to contact me and I will try again to clarify
in the next revision. I shall finish with a final warning: my selection of topics is clearly biased to my own
tastes and interests, which are clearly favouring geosciences versus engineering. It does not mean that
the topics ignored here are not important. For example, a good discussion of extreme waves and sea state
analysis would be much more useful for all engineers than our development on three-dimensional wave-
current interactions. For this you may go to section 4.3 of Holthuijsen (2007) or, with more details, to
Boccotti (2000). I hope that the present book will be a good combination of useful and interesting topics.

The document is organized in two parts, one relevant to waves in deep water, another providing
additional information on coastal and shallow water aspects. Both are complemented by a separate
book, which contains a third part that goes into some details that are probably not relevant for most
readers, in particular Master students.
The present book is designed to make it easier to read in electronic form, including hypertext links
within the document and towards outside sources, such as the cited references. It is designed as a teaching
material for the wave-related Master courses at University of Brest and ENSTA-Paris Tech. Because the
present document is trying to follow the latest advances in research – and my imperfect understanding
of these. The permanent evolution unfortunately leads to the presence of errors, more so in part III.
I thank Nicolas Rascle, Nadine Paugam, Clément Gandon, Nobuhiro Suzuki, Sophia Brumer, Marine
De Carlo and Oyvind Brevik for many corrections, Jean-François Filipot for contributions and help in
translating chapter 3, and Philippe Bonneton for discussions and help on the structure and contents of
chapter 17. I thank you in advance for finding any dubious or strange contents, or broken links.

1
2 CONTENTS
 Part I

Waves in deep water

3
Chapter 1

Introduction

1.1 Waves in geosciences

From the ripples on a water puddle to large breaking waves on the beach, we have all seen waves. They
can be familiar or threatening, possibly deadly for seasoned fishermen or yachtsmen (e. g. Pierson,
1972; Greenslade, 2001). Waves can exert huge forces: just try to stand up in the surf, in front of a
two-meter tall wave that breaks. And two meters is a far cry from the maximum recorded wave height
at sea, towering 32 m above the following crest (Liu et al., 2008). The most severe sea state, estimated
from satellite radar altimetry, had a significant wave height of 18.5 m (Hanafin et al., 2012; De Carlo and
Ardhuin, 2024), which, as we shall see in this chapter, means that some wave heights probably exceeded
35 m. Surfing contests have also focused on some specific coastal areas where waves are strongly amplified
and produce waves of 30 m height or more, such as in Nazare, Portugal.

Wave period
0.1 sec 1 sec 30 sec 5 min 12 hr 24 hr

CAPILLARY ULTRA ORDINARY INFRA LONG PERIOD ORDINARY TRANS-

WAVES GRAVITY GRAVITY GRAVITY WAVES TIDE TIDAL
WAVES WAVES WAVES WAVES WAVES

WIND WIND & STORMS & SUN STORMS

ORDINARY EARTHQUAKES & SUN & MOON
GRAVITY MOON
WAVES

2 3 4 5
0.1 1 10 10 10 10 10 Wave period in seconds
-2 -3 -4 -5
10 1 0.1 10 10 10 10 Frequency in Hertz

Figure 1.1: Classification of ocean surface waves with usual names in red, as a function of wave periods
(x-axis), with dominant forcing mechanisms in green. Adapted from Munk (1950).

To be more precise, we need now to introduce some classification of the different wave motions. A
simple classification, as proposed by Munk (1950) and shown in figure 1.1, is based on the typical time
scales of between the passing of two crests. We shall call this time scale the wave period and denote
it with the symbol T , even though the motion does not repeat itself exactly and is not mathematically
speaking periodic.
Figure 1.1 has boundaries between capillary waves, ultra gravity waves, ordinary gravity waves,
infragravity waves and longer period waves (including tides) at periods of 0.1, 1, 30, and 300 s. Most
of this book will focus on ”ordinary gravity waves” with periods 1 to 30 s. As a physicist by training,
I’ve never liked classifications with dimensional quantities that are often used in natural sciences: these
distinction are practically useful but they can be misleading. Indeed, in some circumstances I would like

5
6 CHAPTER 1. INTRODUCTION

to call ”infragravity waves” some waves of periods around 10 s because they are generated by the same
process as the usual infragravity waves with a period of 100 s.
This period classification is closely related to a physical classification that distinguishes between the
different ”restoring forces” which pulls back the surface towards a flat state, and the different generation
processes for these waves. The two main restoring forces that we will consider are surface tension, and
gravity. Surface tension is mostly relevant for wave periods under 0.1 s, and will not be discussed much
in the present book.
Once the restoring force is known, the next important things for any wave phenomenon are their
ˆ generation,
ˆ propagation
ˆ and dissipation.
The goal of the present book is to describe and make understandable these three aspects, both quali-
tatively and quantitatively. This quantitative understanding allows an accurate forecast of local wave
statistical properties, which we will call the sea state, as well as fluxes of energy and momentum between
the atmosphere, ocean, sea ice, and solid Earth.
In this book, we will restrict ourselves to waves which are more or less directly generated by the wind,
leaving out tsunamis or ship wakes. But before leaving them out, let’s say a few words about tsunamis.
Tsunami propagation and dissipation properties are the same as the wind-waves described here, but whats
sets them apart from wind-waves is their very large wavelength, which cannot be generated by wind in the
same way as the usual wind-waves. Hence their generation mechanisms are specific, namely earthquakes,
landslides, and meteorite impacts, which are all important but rare events, leading to very different
statistical properties compared to the waves continuously generated by the wind. Another slightly more
common source of meteo-tsunamis are abrupt changes in wind properties. All these generation events
are transient, and typically cause a depression or bump of the sea surface that appears very fast but on a
very large scale. This depression or bump then radiates a train of waves, with a period of the order of 10
minutes that is given by the size of the initial surface perturbation. These waves are strongly amplified
in shallow water. The first perturbation to arrive on land can be a trough. If you see the sea retreating
rapidly, this is it ... do not rush out to pick up crabs, but instead run to high ground, as a big crest will
likely follow and flood what was the dry land.
Instead of these transient wave trains, wind-generated waves are incessant and irregular. The time
between the passing of two crests, which we define as the period T , is typically less than 30 s. This limit
is related to wind speeds, as explained in chapter 8. These wind-waves also give rise to infra-gravity
waves of periods 10 s to 10 minutes. For all these motions the average distance between two crests, which
we shall call the mean wavelength Lm , increases with the mean period Tm . This wavelength goes from
a few centimeters to about a kilometer. For wave shorter than a few centimeters, the effect of surface
tension must be taken into account, and these short waves are called gravity-capillary waves.
Indeed, the propagation properties of waves is related to the balance of forces near the air-sea
interface. If gravity or surface tension is the dominant force then the propagation is different. Gravity
fights against surface slopes, setting up a pressure gradient that tends to reduce the sea level slope and
is the main force for large wavelengths. Surface tension, instead, fights against surface curvature: it
arises from the difference in thermodynamic properties of the interface between the two fluids, that are
air and water. This difference gives an energy to their interface, which is proportional to the area of the
interface: the more curved the surface the larger the energy, which is the equivalent of the gravitational
potential energy. This force explains why water droplets are spherical, it is simply the geometry that
minimizes the area (hence the energy) for a given volume. This surface tension also explains why short
breaking waves are not energetic enough to generate bubbles and foam at the sea surface. The presence
of a continuous layer of ice can also act like an elastic layer with an effect similar to surface tension.
In this case even long waves can be influenced by the elasticity of the ice layer, this influence depends
on the ice thickness (e.g. Squire et al., 1995). Because curvature is the second derivative of the surface
elevation, surface tension effects is much stronger than gravity for small scales.
Whether gravity or surface tension is the main restoring force, the work of these forces produces a
motion with an associated kinetic energy. The oscillations of the air-sea interface are thus maintained by
an exchange between potential and kinetic energy, until this energy is dissipated. Our waves are thus
surface gravity waves, gravity-capillary or capillary waves for the shortest.
In the family of gravity waves, at the other extreme towards the large scales, the slow oscillations
on time scales of several hours to several days are also influenced by the Coriolis force, caused by the
1.1. WAVES IN GEOSCIENCES 7

21

20.5
pressure (m)

20

19.5

19

18.5
0 100 200 300 400 500 600 700 800 900 1000
time (s)

Figure 1.2: Example of the evolution of the bottom pressure in about 20 m of water in Bertheaume bay,
France, on January 31, 2004.
The pressure in Pascals was converted here to an equivalent water height in meters by dividing the
absolute pressure recorded by a Nortek Vector instrument, by the product ρw g of water density ρw ≃
1026kg/m3 , and gravity g ≃ 9.81m/s2 , after subtracting the atmospheric pressure recorded nearby,
pa ≃ 105 Pa. The fast oscillation caused by a swell of significant wave height Hs = 2.85 m is superimposed
on the tide that is gently falling, about 20 cm in 20 minutes, as shown by the red line.

rotation of the Earth, and the waves become inertial-gravity waves, also known as Kelvin and Poincaré
waves. The main generation forces for these are the wind and the difference in the gravitational pull
exerted by the Moon and Sun on the center and the surface of the Earth. Kelvin waves share many
properties of gravity waves.
In practice, all these waves co-exist. Fortunately, it is often easy to sort them out and study them
separately. Wind-waves and tides have very different periods and wavelengths, as illustrated by figure 1.2.
There is no such clear separation between capillary and gravity waves, except at low wind speeds when
there is clear gap in the wave spectrum around the wavelength of 1.7 cm. For waves longer than 20 cm,
we will ignore the effect of the surface tension which will greatly simplify our calculations. However,
surface tension should not be ignored in general, in particular when considering wave dissipation by
breaking and the effects of small scale surface roughness, both important for air-sea fluxes and remote
sensing of the ocean surface.
The propagation of waves are generally well known thanks to the works of Laplace, Poisson, Stokes,
Airy, Rayleigh and Boussinesq in the 18th and 19th centuries, with many later refinements. For a
historical perspective, you may read the works of Darrigol (2003) and Craik (2004), and specific problems
and questions are still open. The questions of generation and dissipation are very active research
topics, with a fundamental problem posed by the multi-scale nature of real ocean waves: how short
waves influence long waves and vice versa is very difficult to measure and analyze. Because of the
strong demand for results, successful forecasting methods have been developed on more or less empirical
grounds. Modern wave forecasting started with swell forecasts for Morocco, in the 1920s (Gain, 1918;
Montagne, 1922). This approach was generalized by Sverdrup and Munk (1947) who considered the full
life cycle of waves, from generation by the wind to dissipation in the middle of the ocean and on beaches.
This latter work was motivated by the planning of the allied amphibious operation Torch in Morocco in
1942, which led to a method later applied to Normandy and many Pacific islands (Bates, 1949). Their
British colleagues, forming the W group at the Admiralty, included Deacon, Darbyshire, Barber, Ursell
and Longuet-Higgins who developed similar methods and introduced the spectral analysis of waves in
1945 (Ursell, 1999), paving the way for today’s numerical wave models. The first numerical spectral
wave model was developed by Gelci et al. (1957), the group that continued the Morocco wave forecasting
effort in Casablanca.
Knowing and predicting the properties of waves is necessary for sea-going operations, the design of
any marine structure such as a jetty, an offshore platform or a ship. Because waves are generated by the
8 CHAPTER 1. INTRODUCTION

Beaufort descriptive wind speed specification for observation (open sea)

number term (m/s)
0 Calm 0-1 Sea like a mirror
1 Light air 1.5-2.5 Ripples with the appearance of scales are formed
but without foam crests
2 Light breeze 3 –4 Small wavelets still short but more pronounced;
crests have a glassy appearance and do not break
3 Gentle breeze 4.5 –6 Large wavelets; crests begin to break; foam of
glassy appearance, perhaps scattered white horses
4 Moderate breeze 6.5–8 Small waves becoming longer, fairly frequent white horses
5 Fresh breeze 8.5-10.5 Moderate waves, taking a more pronounced long form;
many white horses are formed (chance of some spray)
6 Strong breeze 11–13.5 Large waves begin to form; the white foam crests
are more extensive everywhere (probably some spray)
7 Near gale 24–16.5 Sea heaps up and white foam from breaking waves begins
to be blown in streaks along the direction of the wind
8 Gale 17–20 Moderately high waves of greater length; edges of crests
begin to break into the spindrift; the foam is blown in
well-marked streaks along the direction of the wind
9 Strong gale 20.5–23.5 High waves; dense streaks of foam along the direction of
the wind; crests of waves begin to topple,
tumble and roll over’; spray may affect visibility
10 Storm 24-27.5 Very high waves with long overhanging crests;
the resulting foam, in great patches, is blown in dense
white streaks along the direction of the wind; on the whole
the surface of the sea takes on a white appearance;
the tumbling of the sea becomes heavy and shock-like;
visibility affected
11 Violent storm 28–31.5 Exceptionally high waves (small and medium-sized ships
might be for a time lost to view behind the waves);
the sea is completely covered with long white patches of
foam lying along the direction of the wind; everywhere
the edges of the wave crests are blown into froth;
visibility affected
12 Hurricane 32 and over The air is filled with foam and spray; sea completely
white with driving spray; visibility very seriously affected

Table 1.1: The Beaufort scale for wind speed reproduced from Alcock and Morgan (1978).

wind, there is an only tradition of telling the wind speed from the geometry of the waves, this is widely
known as the Beaufort scale, reproduced in Table 1.1 from Alcock and Morgan (1978).
Using the geometry of the waves and/or the properties of foam on the surface is still very much how
we measure winds from satellite data, using scatterometers that measure the power of the radar echoes
from the sea surface, or radiometers that measure the brightness temperature of the ocean surface. The
relation between wind speed and those measurements is not unique because the waves may be at different
stages of development. There is also a Beaufort scale for wave heights given in table 1.2, and clearly
there is no one to one relationship between winds and waves.
Waves modify the fluxes of momentum between the ocean and atmosphere and thus influence more or
less directly the oceanic and atmospheric circulation. Waves are also an important agent in the pick-up
and transport of sediments, and the main source of background seismic motions.
Today, we can forecast with good confidence the main properties of the sea state and its consequences,
including forces on a structure at sea, ship motions, working range of a radar... although the details of
the generation and dissipation processes are not well known. This is a tribute to the flair of those who
invented rules and equations to represent the complex and poorly known reality. However, given this
empirical part, it is not surprising that the same models may not be accurate for secondary properties
of the sea state, such as the distribution of the energy radiated in different directions or the statistics of
short breaking waves.
1.1. WAVES IN GEOSCIENCES 9

Wind sea
S code descriptive term wind sea wave height in meters (well developed)
0 calm (glassy) 0
1 calm (rippled) 0–0.1
2 smooth (wavelets) 0.1–0.5
3 Slight 0.5–1.25
4 Moderate 1.25 –2.5
5 Rough 2.5–4
6 Very Rough 4–6
7 High 6–9
8 Very high 9–14
9 Phenomenal Over 14

Table 1.2: The Beaufort scale for sea states: source World Meteorological Organization, table 3700

Going against the long-term specialization and separation of geosciencies in many sub-disciplines,
there has been a strong interest since the late 1990s in the interaction of waves with the atmosphere,
ocean currents, turbulence, sediment motion, from the scale of the global ocean to the small scale of any
particular beach. This is motivated by integrated approaches for climate projections or the understanding
of sediment transport from sources to sinks. These efforts should be continued to properly understand
the interactions of waves and turbulence, and the multi-scale properties of the ocean surface. We hope
that after reading the present book, that gives a broad view of what is known and tries to highlight
what is still unclear, the reader will gather the courage to continue the exploration of ocean waves after
Stokes, Boussinesq, Munk, Longuet-Higgins, Hasselmann, Zakharov, Katsaros, Phillips and many other
less famous scientists who made today’s knowledge possible.
Courage, though, may not be enough, and some tools will be needed to start for this journey, including
a working knowledge of calculus and fluid mechanics.
The following books should be consulted for complements on different topics,
ˆ Kinsman (1965) on general principles and wave measurements, in particular with arrays of sensors.
Although a bit old, this book is very well written and easy to get into.
ˆ Phillips (1977) for a comprehensive view, although not up to date, of upper ocean processes (waves,
internal waves, turbulence ...)
ˆ Mei (1989) for wave propagation, mass transport and wave-structure interactions
ˆ Dean and Dalrymple (1991) gave a real textbook oriented towards engineering applications
ˆ Komen et al. (1994) gives the fundamental – but not basic – concepts of numerical wave prediction
in the open ocean, this is not an easy read
ˆ Komar (1998) wrote a nice textbook for coastal geomorphology, an excellent starting point for
those who do not have a strong physics background
ˆ Young (1999) combined deep and shallow water waves, including also global wave climatology.
ˆ the Coastal Engineering Manual (U. S. Army Corps of Engineers, 2002), replaced the Shore Pro-
tection Manual. This book is edited by the U.S. Army Corps of Engineers, the body in charge of
shoreline defenses and management of ports and waterways. This combines general principles with
empirical formulas for coastal engineering. This is freely available on the web.
ˆ Janssen (2004) gives his perspective on wind and wave forecasting, with many theoretical details
on wind-wave generation and nonlinear wave evolution.
ˆ Holthuijsen (2007) A very well illustrated textbook centered on numerical wave modelling, specifi-
cally for coastal environments, although a bit weak on physical processes, such as bottom friction.
A lot of interesting material and teaching aids can be found on the web, from Tony Dalrymple’s Java
applets, to the UCAR Meted program targeted at meteorologists. A list of useful links will be proposed
separately for each chapter.
Let us now beg the tides and currents to stop their flow, so that we may study waves quietly. We
will see later, in chapters 10 and 19 how waves interact with other oceanic motions.
10 CHAPTER 1. INTRODUCTION

Vertical displacement (m)

4

-2

-4
0 50 100 150 200 250
time (s)

Figure 1.3: Principle of wave by wave analysis: a series of elevation records is broken in individual
waves of duration T . The separation from one wave to the next is the zero down-crossing of the vertical
displacement. This example was obtained from a Datawell Waverider buoy deployed offshore of Crozon,
France, in May 2004.

1.2 Wave motion: some observations

The random nature of ocean waves has long puzzled observers and made difficult all scientific investi-
gations. Figure 1.2 gives a good example of a random sequence of high and low waves. Initially the
forecasting of waves was formulated in terms of the highest wave. The notion of wave height distribution
was only introduced after 1945, thanks to the development of wave recording devices, and the availability
of computing power. Two types of methods have been developed to represent the random nature of the
wave field. One of these is the spectral analysis, which will be heavily used in the following chapters. The
other, is the wave-by-wave analysis, which we briefly describe here. More recent time-frequency analyses
are a kind of hybrid of these two methods.
Both methods are very useful and have their own limitations. Spectral analyses are well suited for
the wave forecasting, in particular at large scales, because it explicitly represents the dispersion of waves
that have different periods. Spectral analysis decomposes the sea surface in elementary sinusoidal waves,
it is thus very important to know the properties of these sinusoidal waves. This is the topic of chapter
2. However, some sort of wave-by-wave analysis must be use to investigate localized events associated to
the finite amplitude of the waves, such as breaking, as discussed in chapter 8.

1.3 Wave-by-wave analysis

1.3.1 Time series
Time series are the most common type of measurement these days, let us see what we can learn about
waves from time series. We shall work with series of sea surface elevation, but we could have used any
other physical quantity such as the velocity or pressure in the water or in the air.
We first define a single wave in time series by the time interval between two consecutive crossings of
the mean sea level as the surface goes down, as illustrated in Fig. 1.3. The choice of ’down’ instead of
’up’ is fairly arbitrary but it keeps the forward face of the wave, which is usually more interesting, within
a single wave, whereas the rear face is split between two consecutive waves. For each wave we define a
period T , which is the length of the time interval, and a height H, which is the difference between the
maximum elevation (the crest) and the minimum elevation (the trough) during the perid.
From a sequence of heights, we can define a probability density function (PDF) dP as the limit,
when dH goes to zero, of the probability P that a wave height is between H and H + dH, divided by
dH. For a statistically stationary sea state, the surface elevation is well approximated by the sum of a
large number of sine waves which are independent from one another. Applying the central limit theorem
1.3. WAVE-BY-WAVE ANALYSIS 11

0.2

0.15

dp(Hv) H s=H 1/3= 6 m

0.1

0.05

0
0 2 4 6 8 10 12 14
H v (m)

Figure 1.4: Rayleigh distribution of wave heights

dp × dH(Hv ) is a probability for a single wave to have a height between Hv and Hv + dH. In red: the 1/3
of the highest waves in the distribution. The average height in this red part is H1/3 , which is one way to
define the significant wave height Hs . In the following chapters we will rather define Hs as Hm0 , equal
to four times the standard deviation of the sea surface elevation. That other definition is recommended
by the World Meteorological Organization and it gives a number very close to H1/3 .

to this approximate model, we find that the surface elevation is a Gaussian process, with negative and
positive anomalies around the mean sea level with statistics defined uniquely by the standard deviation
of the sea surface elevation. As a result, and this was proven in the narrow frequency band limit by Rice
(1944) and Cartwright and Longuet-Higgins (1956), the heights follow a Rayleigh distribution, as shown
on figure 1.4,
2H 2 2
dP (H) = 2 exp−H /Hrms . (1.1)
Hrms
R∞
This PDF is normalized to give 0 dP dH = 1. The Rayleigh distribution is generally a good approxima-
tion for 98% of the distribution, sometimes even more. For extreme values, a better approximation was
given by Tayfun (1980), taking into account the correlations among wave components due to second-order
nonlinearities.
One useful result is that the probability that the height exceeds a given threshold H
b is given by
2
b = e−(H/H
P (H > H)
b rms )
. (1.2)

This expression can be used to compute the height threshold H b associated to a fixed fraction of the wave
population. For example H1/3 is the height beyond which there are the highest 1/3 of the waves, and it
b
b 1/3 = (ln(3))1/2 Hrms which is nearly 1.05×Hrms .
is H
A more commonly used scale for the wave heights is given by integrating (1.2) to get the average
height of the 1/3 highest waves. This is one definition of the significant wave height, denoted H1/3 or Hs .
This scale roughly corresponds to the visual impression of wave heights, which was the most common
source of measurements until the 1940s. More generally but still for a Rayleigh distribution, the average
height of the 1/x fraction of the highest waves is,
h
1/2 √ h
1/2
i i
(− ln(1/x)) + π × erfc (ln(x)) /2 Hrms (1.3)

where erfc is the complementary error function. For x = 1/3, this gives H1/3 = 1.4157 × Hrms .
From the definition, the full Rayleigh PDF p(H) is determined by Hrms . In practice the average H1/3
is more commonly used, and there is also a lot of interest in the maximum height Hmax , but that one
depends on the length of the record. When recording more waves, Hmax is likely to be higher. Waves
that have a height H larger than 2.1H1/3 are called freak waves or rogue waves. If we follow the Rayleigh
statistics, this correspond to 1 in 5700 waves. In practice, they are a bit more frequent for conditions
12 CHAPTER 1. INTRODUCTION

with large average steepnesses, as predicted by Tayfun (1980). Also, for real waves the spectrum is not
√ √
always narrow and on average H1/3 ≃ 3.8 m0 instead of H1/3 ≈ 4.004 m0 , where m0 is the variance
of the surface elevation (Goda, 1985).
In the context of the design of coastal or oceanic structures, there is a great interest in defining the
maximum wave conditions that will occur over the expected lifetime of a structure, typically 50 to 100
years, or with a very low probability of occurrence to ensure maximum safety. For example, some sections
of the Dutch dyke system are required by law to resist waves that occur only once in 10,000 years. The
material and size of the structure is then designed to withstand these extreme conditions. If the extreme
wave height and period is overestimated, the cost of construction is higher than it could have been. If the
conditions are underestimated, the structure is likely to fail in a time shorter than the expected lifetime.
This early failure happened for the first oil platforms built in the North Sea, in an age when there were
no routine wave measurements.
For these extremes, the Rayleigh distribution does not hold, because the Hs is itself a random variable
on the scale of days to centuries. The extreme wave statistics on these long time scales are determined by
the distribution of extreme meteorological event, or, in shallow water, the joint distribution of water levels
(including the astronomical tide) with weather events. These long term statistics are clearly different
from the short term statistics. For short term, the sea state was a superposition of many independent
wave trains, and we could use the central limit theorem. For long terms, we first need to determine
the distribution of the significant wave heights Hs , and the probability that the height of a single wave
exceeds H0 is then the conditional probability p(H > H0 |Hs = Hs0 ). The distribution of Hs follows a
generalized Pareto distribution, and the probability p(H > H0 ) is given by integrating the conditional
probability over Hs0 .
Obviously, this requires stationary statistics. In some regions these statistics fluctuate with climatic
patterns like the North Atlantic Oscillation, which particularly impacts the wave heights on the European
Atlantic coast, or the El Nino-Southern Oscillation (ENSO) which has a strong impact on waves in the
central Eastern Pacific or on the U.S. West coast (Bromirski et al., 2005). As an example, Edward
Thornton, a renown specialist of nearshore dynamics, was asked in 1996 to provide some consulting
advice for the construction of a wastewater pipeline in Monterey Bay, Central California. Construction
started in 1997 in the middle of a very strong El Nino, the hundred-year wave height, which had been
properly evaluated by Ed Thornton, was exceeded and the half-ton rocks protecting the pipe were too
small to stay in place and were dispersed by the waves. For such a strong El Nino event, the storm
waves that caused the damage on the U.S. West Coast were normal. One should thus use statistics
with caution. Defining extreme wave heights is also very challenging in tropical areas where they are
associated with hurricanes that have random tracks: a 20-year record of hurricane tracks generally does
not contain all possible tracks, in particular the one of the next hurricane that will destroy this or that
facility. Finally, there are also long-term trends associated with global change (e.g. Wang and Swail,
2004; Charles et al., 2012), especially in the Arctic where the trend in sea ice extent is leading to higher
wave heights and periods (Stopa et al., 2016b).

1.3.2 Maps
Wave statistics from time series cannot represent all wave properties. Some of these, such as the length
of crests, are defined from the spatial patterns in the wave field. This parameter, although secondary
to the wave height, gives information on the spatial coherence of wave-induced motions and is thus very
important when navigating a seaway or designing a structure that may be wider than the crest length.
Just like we have defined heights and periods, heights and lengths L can be defined in the case of
waves propagating all in the same direction, say x. In that case, the crests are infinitely long in the other
direction y. One important parameter is then the wave slope, defined from the ratio H/L. For sinusoidal
waves, the maximum slope is πH/L.
Things get more complex when considering real waves that propagate in all directions. One can use
the theory of random fields, developed by Adler (1981). In that theory, the crests are defined as subsets
of the horizontal plane that are simply connected and that are above the mean sea level. Using this
definition requires a bit of topology. In this context it becomes more difficult to associate a crest with
a trough. There is a strong development of spatial statistics for ocean engineering and oceanography,
thanks to the development of video measurement techniques (Fedele et al., 2009).
Chapter 2

Main properties of linear waves

This chapter explains how and why water moves in waves, once the waves have been generated. A more
logical sequence might have been to study the generation process of the waves first, but it is much more
complex and can only be understood once one knows how the waves move. The properties exposed in
this chapter are independent of the generation mechanism, and thus common to all surface gravity waves,
including tsunamis, ship wakes, and wind-generated waves.
In order to make things simple, we consider a flat, non-deformable bottom located at z = −h, and
periodic waves in both space and time. This may sound very restrictive, but waves at sea generally
behave as if the bottom were locally flat, and the periodicity allows us to study elementary waves that
are later superimposed thanks to the quasi-linear wave motion, with some possible non-linear corrections
to better fit the equations of motion. We will thus start with the linear wave theory of unidirectional
and monochromatic waves that was first developed by Airy (1841), and which is a good approximation
for waves of small height in not-too-shallow water. These are the typical conditions found in swells for
depths larger than 50 m or so. Swells are the waves generated by winds in remote storms. It is also
instructive for the oceanographer to play the game of differences, and find the common traits and main
differences between these swells, and the tidal waves in shallow water that take the form of Kelvin waves.
In fact, most of the wave properties derived here also apply to Kelvin waves. The main difference is
the geostrophic balance along the crest of Kelvin waves that does not exist for shorter gravity waves for
which the Coriolis force can be neglected for most properties.
We will try here not to get carried away by mathematics, which are necessary to arrive at quantitative
predictions. Instead we will use the equations only as tools to help us reveal the wave motion in terms of
forces, pressure and flow, which should help us navigate the many simplifying assumptions ot understand
the role term.

2.1 Waves: a question of gravity, pressure, mass and vorticity

Before jumping into equations, we should make a few mechanical remarks. There is a motion in and
above waves because the crests and troughs of the sea surface, located at z = ζ(x, y, t), correspond to
different weights of water, which creates a pressure difference. We can already note that these pressure
variations are of the order of ρw gζ, with ρw the water density and g the vertical apparent1 acceleration
of gravity. With a 1 m difference in height from crest to trough, we get a pressure difference of 10 kPa.
What is important for the motion is the pressure gradient. With a wavelength of 200 m, this gives a
crest-to-trough pressure gradient of the order of 100 Pa/m at the same level z, which drives a horizontal
acceleration of 0.1 m/s2 .
The water set in motion flows in an orderly way, following several principles. First, the mass of water
is conserved, and for the slow velocities considered in this chapter, the flow is incompressible, hence
non-divergent. We will also consider that the bottom does not deform under the waves. As a result,
the horizontal convergence between a crest and the next trough must give rise to a vertical divergence.
Besides, because the waves are generated by pressure forces, and propagated by pressure forces, the
motion is, in a first approximation, irrotational. The vertical and horizontal velocities are thus strongly
constrained by these two properties: incompressibility and zero vorticity.
1 Apparent means that this is not just Newton’s general gravitation but it also includes the centrifuge acceleration due

to the Earth rotation, giving g ≃ 9.81m s−2 .

13
14 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

z=0

z=-h

Figure 2.1: Definition of vertical levels: ζ is the sea level, −h is the bottom, both defined relative to an
arbitrary vertical datum z = 0. As a result the mean water depth is D = h + ζ.

In summary, the free surface position ζ and gravity determine the near-surface pressure p, which gives
the horizontal velocity u. The vertical velocity w is determined from the horizontal velocity using the
zero divergence and vorticity conditions, and w also modifies the pressure field. We thus can express the
problem of wave mechanics as a set of four equations for the four unknowns that are ζ, p, u and w.

2.2 Wave mathematics

We shall see that two important quantities, the wavelength L and period T are closely related for waves
of small amplitude a. Here we define ”small” to mean that both the sea surface slope ka and non-
dimensional depth a/D are small, where k = 2π/L is the wavenumber, and D = h + ζ, is the mean water
depth. We remind the reader that z = −h is the vertical position of the bottom, and ζ is the mean sea
surface elevation, both relative to an arbitrary datum z = 0.
These small parameters ka and a/D will appear repeatedly in this book. Their ratio gives a third
parameter, which is independent of the wave amplitude, and that will be very important for the wave
kinematics, this is the non-dimensional depth kD.
We will now go into the details of the linear wave theory, first laid out by Airy in the 19th century. It
has the great advantages of being linear, hence any combination of the linear solution is also a solution
of the equations of motion. More importantly, this linear model explains many of the waves properties,
and is very accurate for swells in not-too-shallow water, and is not too far off for most wave properties,
even for waves breaking in the surf zone.
Let us make two final remarks before we get into the equations. First, the linear waves exist only on
paper, as all monochromatic waves are unstable, with a development of this instability that is faster for
higher waves. This aspect is discussed in more details in chapters 8 and Part 3. Monochromatic waves are
an acceptable solution for short evolution times, at least a few periods, possibly much more. Second, the
choice of the Eulerian framework for the equations of motion is not a very good choice for the accuracy,
as the linear Lagrangian theory is much more accurate than the Eulerian theory of Airy, because of the
very simple balance between the pressure gradient and the fluid parcel acceleration. von Gerstner (1809)
did find an exact theory that exactly satisfies the condition p = pa at the free surface, but it has an
non-zero vorticity, which must be compensated by a sheared current. The use of Lagrangian equations
is unfortunately more complex, and this is the main reason why we do not use it here. It is interesting
to note, that mass conservation is linear in an Eulerian framework, while momentum conservation is
non-linear, but the opposite is true for a Lagrangian framework.

2.3 Eulerian equations for wave motion

Our starting point is the conservation of mass and momentum applied to the ocean, with the former
reduced to a zero divergence as we consider a constant density, which is true for the ocean within a few
parts per thousand, and incompressible flow. The horizontal position is defined by the two-component
vector x = (x, y) and the vertical is z. The corresponding velocities are u = (u, v) and w. Considering
sea water as a perfect fluid, we apply the Navier-Stokes equations,

∂2u
 
∂u ∂u 1
+ u · ∇u + w = − ∇p + ν ∇2 u + 2 , (2.1)
∂t ∂z ρw ∂z

∂2w
 
∂w ∂w 1 ∂p 2
+ u · ∇w + w = −g − +ν ∇ w+ , (2.2)
∂t ∂z ρw ∂z ∂z 2
2.3. EULERIAN EQUATIONS FOR WAVE MOTION 15

∂w
∇·u+ = 0, (2.3)
∂z
where ∇ is the horizontal gradient operator. We thus have four scalar equations since 2.1 has two
horizontal components for the four unknowns that are u, v, w and p. Our problem will be well posed
as soon as we define the boundary conditions, from the continuity of velocity and or stresses (pressure
and shear stresses) and initial conditions. At the bottom, we only impose a free slip
w = −u · ∇h at z = −h (x) (2.4)
which simplifies to w = 0 because we chose h to be constant
At the surface, we make the further hypothesis that for any horizontal position x there is only one
value z = ζ of the free surface position (this excludes an overturning of the surface, as found in plunging
breakers). The free surface is a material surface, which means that water parcels on the surface must
stay on the surface. This can be expressed by the condition
d ∂ζ
(z − ζ) = w − u · ∇ζ − =0 at z = ζ. (2.5)
dt ∂t
An interpretation of this surface kinematic boundary condition is that the vertical motion ∂ζ/∂t is the
combination of the vertical velocity w and the horizontal advection of the water parcels sliding along the
surface u · ∇ζ.
To that kinematic boundary condition we add the dynamic boundary condition that express the
continuity of stresses at the air-sea interface. Neglecting the wind stress and surface tension, these
reduce to a continuity of the pressure. In this chapter we will assume that the atmospheric pressure is
takes the constant value pa ,
p = pa at z = ζ. (2.6)
We now assume irrotational motion, so that the velocity field is given by the gradient of a velocity
potential ϕ. Namely, u = ∇ϕ and w = ∂ϕ/∂z. In this and all our notations, the classical operators
(gradient ∇, Laplacian ∆ ...) are restricted to the horizontal plane, in order to simplify notations. This
assumption of irrotational wave motion is generally consistent with observations of real waves. Still, the
vorticity can be very strong locally, for example in in the bottom and surface boundary layers or just after
a wave has broken. Also, in the presence of sheared currents, the wave motion always has some vorticity,
and a weak vorticity also arises from the Earth rotation. We finally start off by neglecting the viscous
terms in the Navier-Stokes equations. This is justified by the fact that any significant wave-induced
motion has scales of velocity U and length L that are large enough to make the Reynolds number U L/ν
of the order of 104 or more.
All the assumptions made above remove some real features in the wave motion. Real waves always
have some vorticity, which is also linked to the viscosity effects in the boundary layers. A rigorous
treatment of these effects is possible and will be discussed in other chapters. We are here looking for the
most simple solution that will capture most of the important properties of real waves.
Replacing velocities by gradients of ϕ in (2.1)–(2.2) one arrives at
"  2 ! #
∂ϕ 1 2 ∂ϕ p
∇ + |∇ϕ| + + + gz = 0, (2.7)
∂t 2 ∂z ρw

and "  2 ! #
∂ ∂ϕ 1 2 ∂ϕ p
+ |∇ϕ| + + + gz = 0. (2.8)
∂z ∂t 2 ∂z ρw
These two equations establish that the term in brackets is not a function of position and can only be a
function of time γ(t),
 2 !
∂ϕ 1 2 ∂ϕ p
+ |∇ϕ| + + + gz = γ(t). (2.9)
∂t 2 ∂z ρw
The other important equation given by the conservation of mass is linear, The mass conservation
equation was already linear,
∂w
∇·u+ =0 (2.10)
∂z
and is equivalent to Laplace’s equation for ϕ
∂2ϕ
∇2 ϕ + = 0, at − h ≤ z ≤ ζ. (2.11)
∂z 2
16 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

2.4 Small slope waves over a flat bottom: the Airy solution
So far, we had only assumed an incompressible flow (assumption A1), zero viscosity (A2), irrotational
motion (A3), a flat bottom (A4).
We shall now linearize the momentum equation (2.9), assuming that the wave amplitude is small
enough to neglect the non-linear terms. This is verified in chapter Part 3, provided that the wave
amplitude a is much less than the wavelength L and a is much less than the mean water depth D.
This gives the linearized Bernoulli equation
∂ϕ p
≃− − gz + γ(t). (2.12)
∂t ρw

Assuming that waves are propagating 2 at a speed C, the motion is periodic as a function of x − Ct and
thus γ(t) is a true constant, that can be determined from the dynamic boundary condition p = pa at
z = ζ, This gives the pressure as a function of the velocity potential, for z < ζ,
∂ϕ
p = pa − ρw g(z − ζ) − ρw . (2.13)
∂t
The Bernoulli equation states that the pressure is the hydrostatic pressure plus some correction due to
the non-stationary motion. The static pressure term has been removed by the linearization.
The linearized kinematic boundary condition, which expresses the continuity of velocities, reads
∂ϕ ∂ζ
w= ≃ at z = ζ, (2.14)
∂z ∂t
which corresponds to a flat sea surface approximation. We can further approximate that the actual sea
level is not too far from the mean sea level z = ζ.
Finally, the bottom kinematic boundary condition becomes,
∂ϕ
w= =0 sur z = −h. (2.15)
∂z
Taking ∂(2.12 at z=ζ)/∂t +g×(2.14) we eliminate the unknown ζ, and obtain our wave equation

∂2ϕ ∂ϕ
+g = 0, at z = ζ. (2.16)
∂t2 ∂z
Equations (2.15)–(2.16) are the bottom and top boundary conditions for Laplace’s equation. The set
of equation (2.11)–(2.16) is usually called the Euler equation. Its full non-linear form is given in Part 3.

2.4.1 Solution: Laplace equation and vertical profiles

Since we have a linear wave equation, it is natural to solve it using the Fourier transform that gives us
a full basis of solutions. Without loss of generality, we thus look for solutions of the form,
 
ϕ = R ϕe (z) eik·x−σt , (2.17)

where R(x) is the real part of x. Replacing in Laplace’s equation gives the Helmolz equation

∂ 2 ϕe
−k 2 ϕe + = 0. (2.18)
∂z 2
Any solution is thus the sum of two exponentials, which can be recombined in the form

ϕ(z)
e = A cosh (kz + kh) + B sinh (kz + kh) . (2.19)

The bottom boundary condition imposes that B = 0, and we thus only keep the first term,
cosh (kz + kh)
ϕe (z) = Φ0 . (2.20)
cosh (kD)
2 In the presence of standing waves, γ(t) is an oscillating function of the order of ga2 /L, as derived by Miche (1944a).

Longuet-Higgins (1950) further showed that this pressure field is actually an acoustic wave when the water compressibility
is considered, which explains the generation of seismic and acoustic waves by ocean wind-generated waves.
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 17

Considering that the amplitude of the vertical velocity w = ∂ ϕ/∂z

e at the surface, where z + h = D,
is the radian frequency σ times the elevation amplitude a, we have

|Φ0 |k sinh(kD)/ cosh(kD) = σa. (2.21)

Hence, from Laplace’s equation alone we get the orbital velocities. Wave motion has a maximum hori-
zontal speed at the crest with a magnitude σa/ tanh(kD), and in the direction of propagation. Because
of the Laplace equation, when kD >> 1 the motion decays exponentially away from the surface over a
typical distance that is 1/k = L/2π.

2.4.2 Solution: Momementum balance and dispersion relation

Replacing ϕ by eq. (2.17) using (2.20) in the wave equation (2.16) gives

−σ 2 Φ0 + gk tanh (kD) Φ0 = 0. (2.22)

This equation expresses the balance of forces between the horizontal pressure gradient which, just below
the surface is is hydrostatic, with an amplitude ρw gak, and the horizontal acceleration, which has an
amplitude σ 2 a/ tanh(kD). This is the dispersion relation for linear waves over a flat bottom, first given
by de Laplace (1776),
σ 2 = gk tanh (kD) . (2.23)
As a result, where D is reduced but σ is constant, waves have a lower phase speed C = σ/k for example
near the shoreline. Contrary to a widespread misconception, this reduction of the phase speed has
nothing to do with bottom friction. in fact the wave orbital velocity increases, and in the Airy theory
there is no dissipation of energy.

2.4.3 Solution: polarization relations

Having solved for the velocity potential, we can now link it back to the surface elevation using eq. (2.14).
Defining the phase of the free surface elevation,

Θ = k · x − σt + Θ0 , (2.24)

with Θ0 a constant between 0 et 2π, eq. (2.14) links the surface elevation amplitude a and the amplitude
Φ0 of the velocity potential at the surface
σ
a = i Φ0 , (2.25)
g
the elevation, velocities and pressure of the Airy (1841)3 We thus have the following polarization relations
that link the amplitudes and phase of all oscillating variables

ζ − ζ = a cos Θ, (2.26)

and the velocity potential

a cosh (kz + kh)
ϕ= σ sin Θ. (2.27)
k sinh (kD)
gives the velocity field
k cosh (kz + kh)
u=a σ cos Θ, (2.28)
k sinh (kD)
sinh (kz + kh)
w = aσ sin Θ. (2.29)
sinh (kD)
The pressure field is obtained from the linearized Bernoulli equation with the constant given by the
dynamic surface boundary condition, as in eq. (2.13),

cosh (kz + kh)

p = pH + ρw ga cos Θ, (2.30)
cosh (kD)
3 Although Laplace, Cauchy and Poisson gave all elements of this theory many years before Airy, the latter was the

first to address the problem of a single propagating monochromatic wave train. Poisson solved several problems of greater
complexity, including stationary waves and circular waves (Craik, 2004).
18 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

x (m)
0 1 2 3 4 5 6
t=0
0

z (m) -1

-2

-3

0 5 10 15 20 25
p-pa (kPa)

Figure 2.2: Pressure and velocity fields for a monochromatic wave of period T = 2 s in a mean water
depth of D = 3 m.

where the hydrostatic pressure is pH = −ρw g(z − ζ) + pa with pa the atmospheric pressure. We also have
We obtain the displacements of water particles by integrating the velocity field in time. To a first
order of approximation (at first order in ε = ka), we have the horizontal displacement

k cosh (kz + kh)

ξeh = −a sin Θ, (2.31)
k sinh (kD)

and the vertical displacement

sinh (kz + kh)
ξe3 = a cos Θ. (2.32)
sinh (kD)
Because of the wave propagation, the water parcels spend more time under the crest, where the horizontal
velocity is larger, than under the trough, where the velocity is weaker. As a result there is a net, order
(ka)2 drift in the direction of propagation, even for linear waves. This is a Stokes drift - it is also called
the wave (pseudo)-momentum. Like the energy and the wave action, the Stokes drift is an intrinsic
quadratic property of the wave field. This aspect is discussed in more details in chapter 9.
Eq. (2.23)–(2.27) are all approximate solutions corresponding to the Airy waves. Stokes (1849)
extended this Airy solution to take into account the non-linear terms of the full equation (see Part 3),
which improves the agreement with observations. In the deep ocean, all measurements confirm that
waves are very nearly irrotational and well described by the theories of Airy and Stokes (see for example
Thornton and Kraphol, 1974; Herbers et al., 1992). The general solution can be expressed as a series
in powers of ε1 = ka, which was shown to converge by Levi-Civita (1925). Many methods have been
developed to obtain various numerical approximations of the exact solution.

2.4.4 Physical interpretation

The wave motion is maintained by the restoring force of gravity, which acts in quadrature to the velocity
field, and thus always overshoots the equilibrium and creates crests where there were troughs. The
acceleration induced by the pressure field feeds back into the pressure, and the motion is not hydrostatic.
Near the surface, the pressure is nearly hydrostatic (figure 2.2). If the depth is large enough, at some
level the vertical acceleration eventually cancels the pressure oscillations so that there is no significant
motion at a depth much larger than the wavelength. This is typical of a surface wave, the motion decays
exponentially from the surface, it is ’evanescent’ or ’inhomogeneous’.
The flow is better understood by looking at the pressure p corrected for a hydrostatic pressure
pH = pa − ρw g(z − ζ). This reveals a striking property, which is only true for kD ≫ 1, the isobars of
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 19

p − pH are also the streamlines. The streamfunction ψ is such that u = ∂ψ/∂z, which gives

a sinh (kz + kD)

ψ= σ cos Θ, (2.33)
k sinh (kD)

which is, for kD → ∞ the pressure times ρw gk. This flow is in cyclostrophic equilibrium: the pressure
gradient balances the centrifugal force of a water parcel turning around its circular orbit.

2.4.5 Kinematics: influence of the non-dimensional depth kD

The changes in kinematics and dispersion from deep to shallow water are related to the hyperbolic
functions cosh, sinh and tanh, which are plotted in figure 2.5.

Airy waves in deep water

A good number to remember is that, at a depth equal to half the wavelength, the motion amplitude is
reduced by a factor exp(π) ≃ 25 compared to the value at the surface. As a result, waves such that
D > L/2 which corresponds to kD > π, are generally considered to be ’deep water waves’.
The dispersion relation becomes,
σ 2 ≃ gk, (2.34)
the orbital velocities and pressures become
σ
u = ak ekz cos Θ (2.35)
k

w = aσekz sin Θ. (2.36)

H kz
p = p + ρw gae cos Θ. (2.37)
and displacements (2.31)-(2.32) are now

k
ξeh = −a ekz sin Θ, (2.38)
k

ξe3 = aekz cos Θ, (2.39)

which is the parametric equation of a circle of radius aekz . In first approximation, the water parcels
follow circular orbits with diameters that vanish with deph.

Airy waves in intermediate water depth

For smaller water depths, the orbital velocity is significant near the bottom and the trajectories are
now ellipses with horizontal major axis measuring 2acosh (kz + kh)/sinh (kD), and a vertical minor axis
2asinh (kz + kh)/sinh (kD) that shrinks from 2a at the surface, to zero at the bottom where the motion
is back and forth along the bottom.
In that range there is no asymptotic expression for the dispersion relation and one may use Table 2.1.

Airy waves in shallow water

For very shallow water (say kD < 0.1), the dispersion relation becomes

σ 2 = gDk 2 , (2.40)

and the velocities and pressure are

σ
u = ak cos Θ (2.41)
k sinh(kD)
(kz + kh)
w = aσ sin Θ. (2.42)
(kD)
p = pH + ρw ga cos Θ = pH . (2.43)
This last expression states that the pressure is hydrostatic, just like in tidal waves. Tsunamis also nearly
follow this shallow water limit.
20 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

t=0 t=4 s
0 t=0
0
-1

-2

-3 -0.5
0 1 2 3 4 5 6
x (m)
t=0.25 s
0

-1
-1
-2

z (m)
z (m)

-3
t=0.50 s
0 -1.5

-1

-2
-2
-3
t=0.75 s
0

-1
-2.5

-2

-3

-3
0 0.2 0.4
-1 -0.5 0 0.5 1 x (m)
p - pa - rgz (kPa)

Figure 2.3: Left, in colors: pressure anomaly p − pH and velocities at different phases of the propagation
of 2 s period wave in 3 m water depth, going from left to right. Right: trajectories of water particles.
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 21

0
−1
−2
−3
0 10 20 30 40 50
x (m)
0
−1
−2
−3
0
−1
−2
−3

−1.5 −1 −0.5 0 0.5 1 1.5

u(m/s)
(e)
(d)
0

−0.5

−1
z (m)

−1.5

−2

−2.5

−3
−0.5 0 0.5 −0.4 0 0.4
x (m)
Figure 2.4: Velocity field in water for a wave train of period 10 s, and amplitude 18 cm in 3 m of
water. Snapshots are shown for (a) T = 0, (b) 1.25 and (c) 2.5 s. The trajectories of water parcels are
integrated over two Eulerian periods (20 s) for (d) linear waves, and (e) nonlinear waves with the same
period and height, using streamfunction theory (see part 3). For the height chosen here, H/D = 0.12
and H/L = 0.0067, so that linear theory gives a fairly good approximation.
22 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

2
cosh(kz)

kz
1
sinh(kz)
tanh(kz)
exp(kz)/2

0
0

4
value

Figure 2.5: The main hyperbolic functions, used again and again in ocean wave theory.

Y = X tanh(X) X Y = X tanh(X) X Y = X tanh(X) X

0.05 0.2255 1.30 1.4511 2.55 2.5795
0.10 0.3216 1.35 1.4934 2.60 2.6273
0.15 0.3973 1.40 1.5360 2.65 2.6753
0.20 0.4627 1.45 1.5788 2.70 2.7234
0.25 0.5218 1.50 1.6218 2.75 2.7716
0.30 0.5767 1.55 1.6651 2.80 2.8200
0.35 0.6284 1.60 1.7085 2.85 2.8684
0.40 0.6778 1.65 1.7523 2.90 2.9170
0.45 0.7255 1.70 1.7962 2.95 2.9657
0.50 0.7717 1.75 1.8405 3.00 3.0145
0.55 0.8168 1.80 1.8850 3.05 3.0634
0.60 0.8611 1.85 1.9297 3.10 3.1123
0.65 0.9046 1.90 1.9747 3.15 3.1613
0.70 0.9476 1.95 2.0199 3.20 3.2104
0.75 0.9902 2.00 2.0653 3.25 3.2596
0.80 1.0324 2.05 2.1110 3.30 3.3088
0.85 1.0744 2.10 2.1570 3.35 3.3581
0.90 1.1163 2.15 2.2031 3.40 3.4075
0.95 1.1580 2.20 2.2495 3.45 3.4569
1.00 1.1997 2.25 2.2961 3.50 3.5063
1.05 1.2414 2.30 2.3428
1.10 1.2831 2.35 2.3898
1.15 1.3249 2.40 2.4370
1.20 1.3668 2.45 2.4843
1.25 1.4088 2.50 2.5318

Table 2.1: Table of the inverse function of X tanh X. Defining Y = σ 2 D/g it gives k = X/D, which √
allows to invert the dispersion relation in the absence of current. For Y < 0.05 one should use X = Y ,
and for Y > 3.5 one should use X = Y .
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 23

1.5 0.12
(a) (b)

1 0.1

maginute of velocity (m/s)

0.5 0.08

z (m)
0 0.06

-0.5 0.04

-1 0.02

-1.5 0
-0.1 -0.05 0 0.05 0.1 0 1 2 3
u (m/s) x (m)

Figure 2.6: (a) Profile of the horizontal and vertical velocities on both sides of the sea surface at a wave
crest (x = 0), and (b) velocity field, with the white line indicating the position of the free surface. These
velocities correspond to waves of amplitude a = 3 cm, period T = 1.5 s, in D = 3 m water depth, which
correspond to a wavelength L = 3.5 m, and a non-dimensional water depth kD = 3.4.

The orbital displacement (2.31)-(2.32) are now

k
ξeh = −a sin Θ, (2.44)
k sinh(kD)

(kz + kh)
ξe3 = a cos Θ, (2.45)
(kD)

2.4.6 And in the air?

So far we have solved for the water motion. The same hypotheses of irrotational and incompressible
flow will, in the air, produce the same equations and solutions. The only difference is that the bottom
boundary condition is replaced by u → 0 as z → ∞. The air motion over waves is thus similar to the water
motion in deep water waves. The air pressure oscillates, with an amplitude that decays exponentially
with elevation.
These results were confirmed by the measurements of Elliot (1972), who found a vertical decay that
is slightly faster than e−kz , due to the effect of the mean shear in the wind speed. When this shear is
taken into account, the Laplace equation is replaced by the Orr-Sommerfeld equation, as detailed in Part
3. Besides, the velocity jump at the air-sea interface, gives rise to a boundary layer that is laminar for
low wave heights and wind speeds (Dore, 1978), but becomes turbulent otherwise (Perignon et al., 2014),
probably leading to an important dissipation of long waves traveling across oceans that is discussed in
Part 3.

2.4.7 Dispersion
The velocity at which the wave crests or trough propagate is called the phase speed and is given by
C = σ/k which is also equal to the ratio L/T . Using the dispersion relation (2.23), we can eliminate σ

σ hg i1/2
C= = tanh (kD) . (2.46)
k k
The phase speed is clearly a function of the wavelength, and thus waves are dispersive, meaning that
wave packets that contain different components will spread over a larger space as they propagate, with
the long waves traveling faster than the shorter waves. As a result, the waves arriving from a distant
24 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

700
D =10m
600 D =20m
D =50m
D >300m
500
Wavelength L (m)

400

300

200

100

0
0 5 10 15 20
Period T (s)

Figure 2.7: Wavelength as a function of the wave period and the water depth D, for linear waves in the
absence of currents.

storm will always have a long period at the beginning and the average period will become shorter over
time.
1/2
This dispersion property disappears in shallow water (kD << 1) where C goes to (gD) , inde-
pendently of k. On figure 2.7, this gives a constant slope, for example for D = 10 m and T > 5 s).

We also note that, for a fixed period, the phase speed decreases with the water depth. This property
is also true for a slowly varying water, in which case one can consider that the water depth is ’locally
constant’. This variation of C is the cause of refraction (see chapters 10 and 16).

2.4.8 Energy

For any water particle, there is an oscillation of the kinetic and potential energies that are exchanged.
Once integrated over the water depth and averaged over a wave period, that average is represented here
by the overbar, the potential energy per unit horizontal surface is the vertical integral of the potential
energy per unit volume ρw gz. In practice, we ignore the constant energy between the bottom and the
mean sea level ζ, and we set our reference level such that ζ = 0,

Z ζ(x,t)
1 2
Ep = ρw gzdz = ρw g (ζ)
0 2
1
= ρw ga2 cos2 (k · x − σt)
2
1
= ρw ga2 . (2.47)
4
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 25
 
2
For the kinetic energy we integrate the kinetic energy per unit volume, |u| + w2 to obtain a kinetic
energy per unit horizontal surface,
Z ζ(x,t)
1  2 
Ec = ρw |u| + w2 dz
−h 2
 2 " Z ζ Z ζ #
1 agk
≈ ρw cos2 Θ cosh2 (kz + kh) dz + sin2 Θ sinh2 (kz + kh) dz
2 σ cosh (kD) −h −h
 2 Z ζ
1 agk
≈ ρw cosh (2kz + 2kh) dz
4 σ cosh (kD) −h
 2
1 agk sinh 2kD
≈ ρw
4 σ cosh (kD) 2k
1 2
≈ ρw ga , (2.48)
4
1
ρw ga2 = ρw gE,
Et = Ec + Ep = (2.49)
2
where E is the variance of the sea surface elevation, here E = a2 /2. We have thus found that, to a first
order of approximation, the average kinetic and potential energy Ec and Ep are equal.
Wave propagation is associated to a flux of energy. This flux of energy is transmitted by pressure
forces from one water column to the next. This flux is the work of pressure forces given by eq. (2.30)
with a velocity given by eq. (2.28). When integrated over the depth and averaged over a wave period,
this gives the flux per unit crest length (i.e. per unit horizontal distance in the direction perpendicular
to the propagation direction),
Z ζ
W = pudz
−h
ζ
cosh2 (kz + kh)
Z
2
= ρw ga σcos2 (k · x − σt) dz
−h sinh kD cosh kD
Z ζ
2σ 1
= Et [cosh (2kz + 2kh) + 1] dz
sinh(2kD) −h 2
 
2σ sinh 2kD D
= Et +
sinh(2kD) 4k 2
= Cg Et (2.50)

where    
σ 1 kD 1 kD
Cg = + =C + . (2.51)
k 2 sinh 2kD 2 sinh 2kD
W is a flux of energy per unit distance and Cg the average speed at which the energy density Et is
radiated, which defines the group speed. 

The full expression for the energy flux should also include the advective flux u ρw gz + 0.5 u2 + w2 ,
but that part is negligible in the absence of currents.
This speed is called ’group speed’ because it is indeed the speed at which a group of waves travels,
because Cg = ∂σ/∂k. Indeed, the superposition of two monochromatic waves of equal amplitude and
similar frequencies gives a surface elevation

ζ = a cos [(k − ∆k) x − (σ − ∆σ) t] + a cos [(k + ∆k) x − (σ + ∆σ) t] , (2.52)

which writes
ζ = 2a cos (∆kx − ∆σt) cos (kx − σt) . (2.53)
The first factor is the envelope of the group, with a length 2π/∆k and period 2π/∆σ, that propagates
at the speed c′ = ∆σ/∆k. This speed is, in the limit ∆k → 0, equal to Cg = ∂σ/∂k. Two examples are
shown in figure 2.8.
We note that for kD → ∞, equation (2.51) gives Cg = σ/(2k) = C/2. Thus, in deep water the groups
of waves travel at a speed that is half of the phase speed. Things are very different in shallow water
26 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

0
sea surface elevation (m)
-2
0 50 100 150 200 250 300 350 400
time (s)
2

-2
0 50 100 150 200 250 300 350 400

Figure 2.8: Wave groups produced by the superposition of two monochromatic waves. The top panel
corresponds to ∆σ/σ = 0.2, and the bottom panel ∆σ/σ = 0.1. The narrower the spectrum, the larger
the number of waves in the group.

12

11

10

9
Group speed (m/s)

4
T=12 s
3 T=9 s
T=6 s
2
0 20 40 60 80 100
Depth (m)

Figure 2.9: Group speed variation for different periods, as a function of the water depth.

(kD ≪ 1), where Cg = C. In that shallow water limit, waves of all frequencies travel at the same speed
(they are not dispersive) hence the groups also travel at that same speed. This is only true for linear
waves. In part 3you may see that phase and group speeds are also a function of the wave amplitude.

2.4.9 Energy and power

Eq. (2.50) gives the mean energy flux per unit crest length. For example, in the case of monochromatic
waves with a height of 2 m, a period of 12 s and a water depth of 15 m, this flux is W ≈ 50 kW m−1 .
This means that if we take a surface facing the waves, 1 meter along the crest and the full water depth,
there is 50 kW of mechanical power that goes through this surface. This is 5 MW for 100 m along the
crests, which is the peak power of two 150 m high windmills. This number, for rather modest wave
heights, shows the strong concentration of power in water waves. Unfortunately, this power is difficult
to tap to produce electricity, in particular because it is very intermittent in most places.
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 27

Water depth: general case kD ≫ 1 kD ≪ 1

dispersion relation σ 2 = gk tanh(kD) σ 2 = gk σ 2 = gDk 2
1/2 1/2 √ 1/2
phase speed C = σ/k = [g tanh(kD)/k] C = (g/k) = g/σ C = gD
kD
group speed Cg = C 0.5 + sinh(2kD) Cg = C/2 Cg = C
Linear properties (z < ζ)
horizontal velocity u = a kk σ cosh(kz+kh)
sinh(kD) cos Θ u= akσ kz
k e cos Θ u= akσ
k2 D cos Θ
vertical velocity w= aσ sinh(kz+kh)
sinh(kD) sin Θ w = aσe kz
sin Θ w= z+h
h aσ sin Θ
Quadratic properties
Mean energy per unit surface Et = ρw gE = ρw ga2 /2 Et = ρw ga2 /2 Et = ρw ga2 /2
Stokes drift Us = σkE cosh(2kz+2kh)
sinh2 (kD)
Us = 2σkEe2kz Us = σkE/(kD)2

Table 2.2: Main results of Airy wave theory with deep and shallow water limites. We remind that the
phase is Θ = k · x − σt + Θ0 .

Assumption justification consequences

A1. incompressible u ≪ αw ∇ · u + ∂w/∂z = 0
A2. rigid bottom bottom motion ≪ water motion w = −u · ∇h at z = −h
A3. inviscid high Reynolds number viscous stresses are zero
A4. irrotational motion driven by pressure forces + (A3) u = ∇ϕ, w = ∂ϕ/∂z
A5. flat bottom bottom slopes usually ≪ 1 with A2, w = 0 at z = −h
A6. sine wave basis function of the Fourier transform ϕ(z) ∝ cosh(kz + kh)
A7. pa constant on z = ζ ρ a ≪ ρw no wave generation by the wind
A8. no surface tension γ ≪ g/k 2 for L ≫ 0.1 m (in clean water) no capillary waves
A9. ε1 = ka ≪ 1 ε1 < 0.44 for periodic waves with A10, gives linear equations
A10. ε2 = a/D ≪ 1 because it makes equations simpler ... but beware for kD < 1!
A11. no mean current most often U ≪ C for dominant waves simple dispersion σ 2 = gk tanh(kD)
A12. constant density ρ′ /ρw < 0.03 for sea water no internal waves
A13. no Earth rotation f3 ≪ σ

Table 2.3: Assumptions needed to derive Airy’s theory. αw is the sound speed in water, of the order of
1500 m s−1 (in the absence of air bubbles), and U is the mean current velocity. Finally ρ′ is a scale for
density perturbations relative to the mean ρw , and γ is the surface tension, such that γρw (R1 + R2 ) is
the additional pressure under a surface with radii of curvature R1 et R2 , counted positive for a surface
that is convex on the air-side, e.g. a crest.

2.4.10 Summary
We have obtained the main properties of regular linear waves, summarized in table (2.2). Because these
Airy waves are solutions of the linearized equations of motion, they can be combined to obtain the general
solution. Hence the surface elevation, velocities and pressure are given by the sum of monochromatic
waves, each proportional to their amplitude a. We will see in the next chapter that it is also possible
to add up the quadratic properties that are the energy and Stokes drift, which are proportional to the
surface elevation variance E = a2 /2. These properties come from a series of assumptions, listed in table
2.3 and that will be discussed or removed in the following chapters that extend Airy theory, giving access
to all sorts of corrections and allowing to determine the evolution of Airy waves caused by different forcing
and dissipation processes. Indeed, we have determined here the eigenvectors of the linearized equations
of motion: these are free waves that can exist without forcing. In practice, the forcing is necessary to
generate these waves in the first place, and this forcing is balanced by dissipation when long time scales
and large spatial scales are considered. That dissipation requires to include vorticity and viscous effects.

2.4.11 Extending Airy wave theory

What happens when we do not make one of the assumptions A1 to A13? In which conditions should we
do this?
ˆ A1. Except when considering acoustic and seismic noise generated by ocean waves, as in part 3,
28 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES

we can safely ignore compressibility effects.

ˆ A2. Bottom deformations are relevant when considering acoustic and seismic noise generation. In
fact, the bottom deformation should be considered for all acoustic wave propagation in the ocean.

ˆ A3. In the boundary layers at the sea surface, and more importantly at the sea bottom, we will need
molecular viscosity and turbulence effects (which can often be represented by an eddy viscosity).
However, these layers are very thin, with a thickness of the order of δ = (ν/σ)1/2 , which is typically
less than a millimeter for the kinematic viscosity of water ν ≃ 4 × 10−6 m2 s−1 , and wave periods
T > 1 s, which is consistent with measurements of the water-side surface boundary layer by Banner
and Peirson (1998). At the bottom, turbulence effects are important but the wave boundary layer
is only a few centimeters thick.

ˆ A4. For a viscous flow, the vorticity from the top and bottom boundary layers diffuses in the water
column and in the air (Longuet-Higgins, 1953; Weber and Førland, 1990). Besides, there is also a
weak vorticity caused by the Earth rotation, see A13 below.

ˆ A5. The bottom boundary condition becomes −∇ϕ · ∇h = ∂ϕ/∂z (inviscid case), and is only
satisfied exactly in the presence of at least two wave trains, one incident and one reflected. The
incident wave train is also modified by diffraction effects. For small slopes, diffraction and reflection
are generally weak, and the wave motion is well approximated by a ”WKBJ” approximation:
replacing the phase Θ by a function S(x, t) with k = ∇S and σ = −∂S/∂t (see chapter 16).
In this context small means that refraction or diffraction effects do not produce of significant
variation of the wave amplitude at the scale of one wavelength. A more accurate approximation
of the dispersion relation over a sloping bottom was given by Ehrenmark (2005), in the form
σ 2 = gk tanh(khβ/ tan β), where β is the angle between the bottom and the horizontal. This
correction is weak, only 4% for a slope β = 10◦ . For steep slopes, reflection becomes important
and the separation of the variables x and z becomes meaningless. The velocity potential can be
obtained numerically (e.g. Athanassoulis and Belibassakis, 1999; Belibassakis et al., 2001).

ˆ A6. For a flat bottom this is not an assumption: we have the right to decompose the waves into
sine waves because these are a complete basis. For small bottom slopes, the wave train is only
locally equivalent to a sine wave. For steep slopes, the wave train can suffer strong distortions at
the scale of one wavelength (e.g. Magne et al., 2007).

ˆ A7. An atmospheric pressure oscillation on the scale of the wavelength can produce an amplification
or attenuation of the waves, depending on its phase relative to the waves. This aspect is discussed
in detail in chapters 8 and Part 3.

ˆ A8. Due to surface tension, the pressure under the surface is increased by −γ ∂ 2 ζ/∂x 2



+ ∂ 2 ζ/∂y 2 .
This added pressure modifies the dispersion relation to give σ 2 = gk + γk 3 tanh(kD). This
modification is negligible for wavelengths larger than a few centimeters. The presence of a layer
of ice at the sea surface has a similar effect on the waves, with added terms due to bending and
inertia, and is significant for periods of 10 s and less (Liu and Mollo-Christensen, 1988) in the case
of an ice layer thickness of one meter or more, and thicker layers have an influence on longer waves.
ˆ A9. Non-linear effects associated to ε1 ̸= 0 are fairly complex and will be discussed in chapters 8
and part 3. One particular consequence is that a monochromatic wave train is generally unstable
(Benjamin and Feir, 1967). For waves in one dimension, as produced in the laboratory, this can cre-
ate very high (freak) waves. Another consequence is that different wave trains interact, exchanging
energy an momentum.

ˆ A10. Non-linear effects associated to ε2 ̸= 0 are important for kD < 1 even if the wave height is
small. This is particularly true near the shoreline, and the shape of waves can be strongly modified,
as discussed in chapters 17 and part 3.
2.4. SMALL SLOPE WAVES OVER A FLAT BOTTOM: THE AIRY SOLUTION 29

ˆ A11. Since the laws of physics are unchanged by a change of Galilean reference frame, a uniform
current U in the absence of bottom friction only introduces a Doppler frequency shift, and all
results established here remain valid, replacing Θ by ΘD = Θ + k · U. One can define an absolute
frequency, as measured in the reference frame attached to the bottom,
1/2
ω = σ + k · U = k · U + [gk tanh (kD)] . (2.54)

For a current U(z) that varies only in the vertical, and in the limit ε1 ≪ 1 and ε2 ≪ 1, this
dispersion relation generalizes to the form,

ω ≡ σ + k · UA (2.55)

with the advection speed given by Kirby and Chen (1989),

Z ζ
2k cosh [2k(z + h)]
UA = k · U(z) dz. (2.56)
−h sinh(2kD)

Finally, when U also varies horizontally, the phase speed varies horizontally, so that refraction and
diffraction effects appear, just like they do on a sloping bottom. These questions are addressed in
chapter 10.

ˆ A12. When the ocean is stratified, due to a vertical variation of temperature and/or salinity, the
Airy waves are the ’external mode’ in a family of waves that also include internal waves. Because
the equations of motion are weakly nonlinear, these different modes are coupled, with an exchange
of energy between surface waves and internal waves (Thorpe, 1966). This aspect is still relatively
unexplored (Osborne and Burch, 1980; Kudryavtsev, 1994). This stratification can also be due
to the presence of air bubbles near the surface or sediments near the bottom, with important
consequences for the bottom boundary layer and wave dissipation. (e.g. Winterwerp et al., 2003;
Styles and Glenn, 2000).

ˆ A13. Taking into account the Earth rotation with a vertical Coriolis parameter f3 , a very weak
transversal velocity component v appears, of the order f3 u/σ and in phase with w. This transversal
component is important for the surface drift current (Hasselmann, 1970; Xu and Bowen, 1994;
Ardhuin et al., 2004; Rascle and Ardhuin, 2009). In the case of wind-generated waves, f3 /σ is
typically of the order of 10−4 , so that the effect on the wave kinematics and dispersion is not
measurable. There is also a very weak deviation of the propagation direction of the waves (Backus,
1962). This is still true for tsunamis which have much larger periods than wind-waves, typically a
few minutes up to 30 minutes. Airy wave theory thus also applies to tsunamis, as long as non-linear
effects are weak. On the contrary, for motions with longer periods, such as tides, the Earth rotation
must be taken into account. This is why these waves are called inertia-gravity waves.
30 CHAPTER 2. MAIN PROPERTIES OF LINEAR WAVES
Chapter 3

Wave heights and spectra: theory

and measurement

A detailed knowledge of the wave motion in all locations is generally not required, one may rather be
interested in the evolution of some wave properties on distances much larger than the wavelength. A
statistical approach is therefore preferred. Note also, that even in a wave tank, waves are never strictly
monochromatic.
The most common method to represent the random nature of waves is the spectral analysis. It
owes its success to the dispersive nature of waves (different components travel at different speeds), and
it was made practical by the development of computer sciences in the 1960s and to the elegant Fast
Fourier Transform (FFT) algorithm. Those curious to know how a spectrum can be calculated without
a computer can read the amazing account of the rotating system with variable speed invented by Barber
et al. (1946) to read off the energies of the spectral components from a wave time series printed on paper.
With the Fourier method, a record of wave elevation ζ(t) is decomposed into a superposition of
sine waves, with regularly spaced frequencies. If the record has the three dimensions, ζ(x, y, t) this
decomposition can be done in frequency, wavelength and propagation direction. The phases of theses
sine waves are generally random, namely, the phase of one component in one record and the next record
are not correlated at all. The only slowly varying quantity is the spectral density, defined as the amount
of elevation variance per unit spectral bandwidth. The procedure can be applied to other variables, not
just the elevation: pressure, velocity ... This slow variation of the spectrum allows a numerical prediction
of ocean wave spectra.

3.1 Frequency spectrum

The spectral analysis that is applied to wave measurements is fairly different from the harmonic analysis
which is used for studying tides. Tides are described as a sum of discrete components whose frequencies
are very well known as they are associated with astronomical motions. For tides we thus have a finite
set of frequencies at which the amplitude and phase can be determined with great accuracy. Waves are
described as a the sum of a many components with energy at all frequencies. There is no gap in the
wave spectrum, and the phases are completely random, uniformly distributed between 0 and 2π while
the amplitudes also have random fluctuations. The result of the spectral analysis is a wave spectrum
that describes the wave energy distribution as a function of frequency. The practical method used to
compute wave spectra from time series of surface elevation or pressure is detailed in Chapter 4.
Figure 3.1 shows how a tide elevation signal is already quite well reproduced by the superposition of
only two sine waves (these two waves are called S2 and M2). On the contrary for the high frequency
signal, dominated by waves, one, two or three sines waves (black, red and green) are far from sufficient
for representing the initial signal (blue). Indeed, to reconstruct the wave signal a great number of sine
waves with relatively close frequencies are required. For simplicity, let us start with one realization m of
an elevation time series, that can be expressed in terms of a Fourier series,

N
X
ζm (t) = am,i cos(2πfi t + Θ0,m,i ) (3.1)
i=1

31
 32 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

8
(a) (c)
average over 10 minutes

2 M2
Water level (m)

amplitude (m)
1.5

4
1
S2
2 Predicted tide
0.5
only M2+S2
Measured
0 0
0 5 10 15 20 0.076 0.078 0.08 0.082 0.084 0.086
Frequency (cycles per hour)
Time (days after 2009/02/01)
10 8
(b) (d)

: E(f) (m2/Hz)
9
Water level (m)

8 6

7
4

Spectral density
6

5 2
4

3 0
0 50 100 150 200 250 300 350 400 0 0.05 0.1 0.15 0.2 0.25
time (seconds after 2009/02/10 - 4:00 UTC) Frequency (Hz)

Figure 3.1: Example of surface elevation time series (a and b) deduced from pressure measurements
collected as the foot of Western cliff of Banneg Island, Molene Archipelago. The corresponding spectra
for the high and low frequency part illustrate the difference between a tide spectrum, presented as a
function of amplitude and a wave spectrum. Below: a sample of the initial elevation signal and one, then
two, then three sine waves are combined to approximate the initial signal. In the case of tides, the fit
is already good with 2 components. In the case of waves, the detailed shape of the elevation cannot be
reproduced without many components.

Where am,i , fi and Θ0,m,i are the Fourier amplitudes, frequencies and phases of the Fourier mode i,
found for the realization m of the sea state. As explained above, N must be a large number. In practice
the phases are nearly random and uniformly distributed over [0, 2π] 1 . The ensemble mean of the Fourier
amplitudes, expressed as function of the frequencies, A(fi ) = ⟨am,i ⟩ is called the amplitude spectrum.
For waves, identical sea state realizations can only be obtained in controlled laboratory experiments.
Instead, the sea state is assumed stationary and the ergodicity theorem is evoked to replace the ensemble
mean by a temporal mean. In practice, M samples of a given (stationary) wave record simulate M
realizations of the sea state and provide an equivalent ensemble mean.
For such random signals, the ’power’ spectrum is preferred to the amplitude spectrum. As demon-
strated in the previous chapter, the mechanical wave energy per unit surface of ocean, for an sine wave
of amplitude a is ρw ga2 /2. As a consequence, the energy spectrum is
  M
1 2 1 X 1 2
a = a , (3.2)
2 m,i M m=1 2 m,i

With this definition, the values taken by the spectrum decrease proportionally to the spectral resolu-
tion ∆f that is the inverse of the length of time over which the spectral analysis is performed. In order
to avoid this dependency on the record length, it is customary to work with a power spectral density
(PSD for short),  
1 1 2
E(fi ) = a . (3.3)
∆f 2 m,i
1 The phases are not exactly random for an actual sea state, waves are slightly asymmetric, the front face being steeper

than the back, and the crests sharper than troughs (e.g. Agnon et al. (2005)). For most applications, these effects can be
neglected.
3.1. FREQUENCY SPECTRUM 33

In the limit of large record lengths, the frequency interval ∆f tends towards zero, and we obtain the
continuous wave energy frequency spectrum,
 
1 1 2
E(f ) = lim a . (3.4)
∆f →0 ∆f 2 i
While wave are irregular, the spectrum is relatively smooth, evolving slowly in space and time, with
a typical time scale of a few hours. This regularity that contrasts with the apparent irregular motion
of the sea surface, allows for a predictive numerical modeling. Note further, that for convenience we
continue to call (abusively) ’energy’ the elevation variance E which has units of length squared. The
true energy, in Joules per unit surface, is in fact ρw gE.
This approach can be generalized to waves travelling in all directions. The Fourier representation of
the sea surface elevation becomes,
N X
X M
ζm (x, y, t) = am,i,j cos(2πfi t − ki cos(θj )x − ki sin(θj )y + Θ0,m,i,j ), (3.5)
i=1 j=1

as illustrated by figure 3.2

Figure 3.2: Reconstruction of a given sea state from the superposition of a great number of plane waves
each with a particular direction and wavelength. Illustration of equation 3.5. After Pierson et al. (1955).

In this expression ki and fi are related by the linear dispersion relation and θj is the direction of
propagation of the Fourier mode (i,j). In the same fashion as for the frequency, the continuous frequency-
direction wave energy density spectrum,
 
1 1
E(f, θ) = lim lim ρw ga2i,j . (3.6)
∆f →0 ∆θ→0 ∆f ∆θ 2
Obviously, E(f, θ) is just a redistribution of E(f ) on the different wave directions and we recover the
heave spectrum by summing on these directions,
Z 2π
E(f ) = E(f, θ)dθ. (3.7)
0

Keeping only the wave energy and its distribution reduces the representation of the waves properties
to a manageable amount of information, but some information is lost in the process. Indeed, it is
34 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

not possible to reconstruct the sea surface from the spectrum, especially because the phases are not
conserved. In practice, the phase couplings are often negligible, and any reconstructed sea surface with
random phases is statistically similar to the initial wave field. In this sense, for a Gaussian sea surface
elevation, the spectrum contains the full statistics of the sea surface.

3.1.1 Wavenumber or frequency?

Depending on the measurement method, the numerical model or the application, one may want to
perform the spectral analysis in the wavenumber or frequency space. The following relations between
the frequency and wavenumber spaces, assume that waves follow the linear dispersion relation. The rule
is simple, the variance of a given quantity does not depend on the coordinates in spectral space. The
variance is the spectral density times the spectral width, hence,

E(f, θ)dθdf = E(k, θ)dθdk (3.8)

which yields
∂k 2π
E(f, θ) = E(k, θ) = E(k, θ) (3.9)
∂f Cg
In the same manner,
2π 2π
E(f, θ) = E(k, θ) = kE(kx , ky ) (3.10)
Cg Cg
The relationship must be used with caution for the high frequency part of the spectrum, because
of nonlinear harmonics. This is significant at frequencies higher than three times the wind sea peak
frequency (Leckler et al., 2015; Peureux et al., 2018). Finally, and we shall see why in chapter 10, when
the effects of currents on waves are included, numerical models usually work with the action spectrum
instead of the energy spectrum. This action spectrum is usually defined as
1 1
A(k, θ) = E(k, θ) = E(kx , ky ) (3.11)
σ σ
For instance, the numerical model WAVEWATCH III (Tolman and Booij, 1998) calculates the evolution
of A(k, θ) discretized over N frequencies and M directions, through the variable ASPEC(I,J) with 1 ≤
I ≤ N and 1 ≤ J ≤ M . However, the model output is transformed back to E(f, θ).
The spectrum is the primary variable of wave forecasting model, and, as such, it is important to
be familiar with its physical meaning. Figure 3.3 illustrates the relation between the sea surface and
spectrum shapes. The reader is invited to try to recompose a sea surface in the physical space from the
spectrum produced by a numerical model,
M X
X N q
ζ(x, y, t) = 2E(fm , θn )∆f ∆θ cos [km cos θn x + km sin θn y − σm + Θ0 (m, n)] , (3.12)
m n

where ∆f and ∆θ are the spectral resolutions, and M and N are the number of frequencies and directions
used to discretize the spectrum. Rigorously, M and N should be taken infinite, but we may start with
the typical resolution of a spectral wave model, of the order of 30 frequencies and 24 directions, with a
significant level of energy in maybe only 20 components. If you want to get a more realistic surface it is
best to interpolate the spectrumpon a fine spectral grid, before taking the inverse Fourier transform.
The component amplitude 2E(fm , θn )∆f ∆θ is consistent with the fact that the total variance
elevation is the sum of the amplitudes squared divided by two, or the spectrum integral over the entire
spectral domain. The phase Θ0 (m, n) can be taken randomly distributed over [0,2π]. The resulting
surface will look smoother than an actual sea state, and the observed crest-trough asymmetry will not be
reproduced. This is due to the random phase assumption that disregards the phases coupling between
spectral components. You may add some nonlinear effects with the ”choppy” method, shifting your
elevation map using an estimate of horizontal displacement, giving more realistic wave crest shapes
(Nouguier et al., 2009).
3.2. USING SPECTRA 35

L
Physical space
y y y

x x x
regular waves wave group Short-crested waves

ky ky ky

kx kx kx

Spectral space k=√(kx2 + ky2) = 2π / L

Figure 3.3: Relation between spectral and physical spaces. Schematic spectra of monochromatic waves
(left panel) and modulated in terms of wavenumber or direction. For the last two cases, the surface is
composed of three components.

3.2 Using spectra

Skipping here the technical details necessary for a practical estimation (see chapter 4), we now have a
spectrum. Very nice, but what can we do with it?

3.2.1 Transfer function

Depending on the application, one can transform the elevation variance spectrum. For a variable A (for
instance the bottom pressure), related to the surface elevation through a linear relation:

A = M ζ, (3.13)

where M is a complex number that takes values in the spectral space, the variance spectral density of
A, EA is
EA (kx , ky ) = |M |2 E(kx , ky ) or EA (f, θ) = |M |2 E(f, θ) (3.14)
or any other equivalent expression in other spectral coordinates. For instance, the bottom velocity
variance will be given from the polarization relation (2.26)–(2.32), that yield M = σ cos(θ)/ sinh(kD),
with θ the angle between the x-axis and the wave direction of propagation. We hence get the spectrum
of the x-component of the bottom velocity.
Two of the most commonly used transfer functions are M = σ for getting the surface orbital velocity
magnitude in deep water, hence M 2 = 4π 2 f 2 , and M 2 = ρw g/Cg = ρw g/(4πf ) for getting the energy
flux in deep water. Using these transfer functions you get the surface orbital velocity variance
Z ∞
< u2 + v 2 >z=ζ = 4π 2 f 2 E(f )df = 4π 2 m2 (3.15)
0

where m2 is the second moment of the spectrum, and in general the p-th moment of the spectrum is
defined as,
Z ∞ Z 2π
mp = f p E(f, θ)dθdf (3.16)
0 0
36 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

Now, you may wonder why somebody would like to know the variance of the orbital velocity, well, it
comes up, for example in Morison’s equation that is the most widely used equation in offshore engineering
to estimate forces on a structure (Boccotti, 2000), and other variances will matter for other applications.

3.2.2 Spectral and integral parameters: Hs , Tp ...

It can be inconvenient to describe a sea state by a two-variable function. Even when it is discretized into a
numerical model, typically over 30 frequencies and 24 directions meaning 720 spectral components, which
is a lot of numbers to describe a sea state. This information can be summarized into a few meaningful
parameters.
As the spectrum is a decomposition of sea surface variance, the most important parameter is certainly√
the variance E, often abusively called energy. From the elevation variance E, we get√ a length scale E.
Going back to sine waves, the variance is a2 /2 with a the wave amplitude. Hence 2E is an equivalent
amplitude for random waves. More precisely, it is the root mean square amplitude.
√ The root mean
square wave height for a random wave is thus twice this amplitude Hrms = 2 2E.
For practical applications, the most widely used height scale is the significant wave height Hs that
corresponds to the visual feeling given by the sea. From the wave height distribution we can define H1/3
(see chapter 1). From the spectrum we define,
sZ
√ ∞
Hs ≡ Hm0 ≡ 4 E = 4 E(f )df (3.17)
0

In practice Hm0 ≃ H1/3 , this property can be demonstrated in the limit of a narrow spectrum (Longuet-
Higgins, 1952). Following the recommendations of the World Meteorological Organization, we shall
consider Hm0 to be the definition of the significant wave height Hs . The ”m0” index indicates that it is
based on the zeroth moment of the spectrum.
In many cases, the directional information is not available, and we only have the frequency spectrum
E(f ). This distribution of the energy as a function of frequency contains information on the typical time
scales of the signal. E(f ) generally exhibits a sharp maximum around the frequency fp , E(fp ) = Emax .
fp is the peak frequency, corresponding to the peak period Tp = 1/fp . This peak period can be noisy in
the presence of several peaks, as in Fig. 3.4. The frequency distribution can also be characterized from
the spectral moments, mp ,
" Z ! !#−1/p
fmax Z fmax
−1/p p
Tm0,p = [mp /m0 ] ≃ f E(f )df / E(f )df . (3.18)
0 0

The most widely used periods are Tm0,−1 , Tm0,1 , Tm0,2 . Each of these has more or less weight on low
frequency part of the spectrum. If one is interested in an effect proportional to f 2 , as is the case of the
wave forces exerted on a structure, it is reasonable to use Tm0,2 . Besides, Tm0,2 is very close to the mean
period Tz given by wave-by-wave analysis. Note however that in practical calculations, Tm0,2 depends
on the choice of fmax . The values of the different periods for a typical spectrum are shown in figure 3.4.
Finally, we define,
Z 2π Z 2π
a1 (f ) = cos θE(f, θ)dθ/ E(f, θ)dθ, (3.19)
0 0
Z 2π Z 2π
b1 (f ) = sin θE(f, θ)dθ/ E(f, θ)dθ, (3.20)
0 0

that can be estimated from the elevation spectrum E(f ) and the elevation-horizontal displacements
co-spectra, Exz and Eyz , as detailed in chapter 4, eqs. (4.8)–(4.9).
The mean wave direction at frequency f is,

θm (f ) = arctan(a1 (f )/b1 (f )), (3.21)

and the directional spreading, as defined by Kuik et al. (1988) is the standard deviation (in radians) of
the spectral width in the limit of a narrow spectrum,
h i1/2
σθ (f ) = 2(1 − (a21 (f ) + b21 (f ))1/2 ) . (3.22)
3.3. RANDOM WAVES IN SITU OBSERVATIONS 37

15

Buoy 51001 (Hawaii)

11/01/2007 18:00 UTC
E(f, ) (m2/Hz/rad)
Hs =3.1 m
0.0 5. 10. 15 20 25
Tp=17.8 s
Tm0,2 =8.0 s
Tm0,-1 =11.6 s
1
10
U10 =9.5 m/s
2
0.30 0.20 0.10 0.0 3
f (Hz)
E(f) (m2/Hz)

f0-1 =0.087 Hz
1: swell 1
Hs1 =2 m
Tp1 =11 s
dp1=55°

5 2: swell 2
Hs2 =1.7 m
f02 =0.125 Hz Tp2 =17.8 s
dp2=310°

3: wind-sea
Hsws =1.6 m
Tpws =7.5 s
dp3 =90°
fp =0.056 Hz
0
0.1 0.2 0.3 0.4
f (Hz)

Figure 3.4: Typical example of a oceanic spectra in tropical area, measured by buoy 51001, moored
350 km North West of Kauai island, on January 11, 2007. The different peak and mean periods are
indicated along with the parameters produced by a decomposition of the spectrum into a primary swell,
secondary swell and wind sea. This analysis is generally not possible without directional information.
Note that the wind sea only appears as a soft ”shoulder” to the right of the secondary swell, while it
comes from a different direction. This shows the possible difficulty of separating swell and wind sea
from a frequency spectrum. Also note that the significant wave height Hs is less than the sum of Hs1 ,
Hs2 , H
psws , of the three systems: the energy can be summed but the wave heights cannot. Namely
Hs = Hs1 2 + H2 + H2 .
s2 sws

√
For equal energies in opposite directions, σθ is maximum at 2 radians, which is 81 ◦ .
The directional wave properties of the dominant waves, can also be characterized with the mean
direction and directional spreading of the spectral peak: θm (fp ) and σθ (fp ). θm (fp ) is often referred to
as ”main direction”, while the mean direction would rather be an average over the entire spectrum,
Z ∞ Z ∞ 
θM (f ) = arctan b1 (f )E(f )df / a1 (f )E(f )df (3.23)
0 0

With all these directions, one must be careful with the directional convention. The directions are usually
counted from North (direction 0), progressing clockwise (90 east, 180 south, 270 West). However, de-
pending on the authors, the direction convention is either meteorological (direction from where the waves
or wind are coming, this is the convention used in the present manuscript) or oceanographic convention
(direction toward which the waves or currents are moving). Be careful!

3.3 Random waves in situ observations

The most common usage of wave time series is the determination of the frequency spectrum E(f ) and,
when more than one variable are measured, the directional wave spectrum E(f, θ) may be estimated.
Details of this processing given in chapter 4. In practice, the most common in situ instruments for wave
measurements are surface-following buoys or bottom-mounted pressure gauges or ADCPs.

3.3.1 Wave gauges

These are reference sensors that directly measures the free surface elevation ζ(x, y, t) at a fixed hori-
zontal position (x, y). The measure is done through the electrical resistance or capacity of one (or two)
38 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

6 x (m)
y (m)
z (m)
4

-2

-4
300 350 400 450 500 550
temps (s)

Figure 3.5: Left panel: Deployment of a Datawel Waverider buoy, 0.9 m of diameter, equipped with a
(long) HF radio antenna and with a (short) Orbcomm satellite antenna for data transfer. Right panel:
examples of displacements measured by this buoy offshore Crozon in May 2004. (same time series as in
figure 1.2). Of course, the 10 m high waves were not measured by the times of the buoy deployment.

conducting wires forming a loop closed by the sea surface. This type of sensor is widely used in the
laboratory, but it is not so common at sea because this requires a fixed platform and the wires are
susceptible to damage by small floating debris. Also, the large wave heights encountered in the field
require long enough gauges. These wave gauges can also be mounted on a buoy that filters out the long
waves through its motion (Graber et al., 2000). Wave gauges are most often associated in an array (see
below) of several gauges synchronized and mounted on a single platform so as to provide a wave direction
estimation (Cavaleri et al., 1981).
The direct measurement of surface elevation can also be performed by radar and LIDAR systems,
which determine the distance to the sea surface from the travel time of an electromagnetic or sound
wave, and that can likewise be arranged in an fixed array or integrated in a scanning system.

3.3.2 Wave buoys

Buoys measure either the successive positions and velocities, determined by a global navigation satellite
system (such as the Global Positioning System - GPS) or vertical acceleration recorded by a buoy floating
at the free surface that give after a double integration in time, a surface elevation signal ζ(x, y, t).
Depending on the type of instrument and on the presence of currents, the horizontal position (x, y) is
not fixed but nearly follows the wave orbital motion. This later property may be annoying for purists
of the wave profile, as it partly filters out the free surface nonlinearity (the linear Lagrangian motion
involves a part of the nonlinear Eulerian motion). In addition to the heave measurement, that was for
long the most common, the wave direction can be determined by measuring the horizontal accelerations
that yields, through double integration, the horizontal displacement x and y (figure 3.3.2). The use of
precise satellite positioning now allows a direct measure of the 3-component position and velocity, the
latter being generally more accurate (Herbers et al., 2012). Many companies have commercialized such
buoys.
The largest buoys generally use a measurement of the components ∂ζ/∂x and ∂ζ/∂y and of the local
free surface slope: pitch and roll as the first prototypes of Longuet-Higgins et al. (1963) and Cartwright
and Smith (1964). This is the case of the US National Data Buoy Center (NDBC) 3-m diameter buoys.
Details of buoy processing are given in chapter 4. Both methods, acceleration and pitch-roll allow,
thanks to the three signal covariance, the determination of the first four Fourier coefficients of the
angular distribution, also known as angular moments,
ˆ a1 and b1 , defined by (3.19)–(3.20) and calculated from the co-spectra Exz and Eyz (equations
4.8–4.9)
ˆ a2 and b2 , defined as a1 andf b1 but with cos θ and sin θ replaced by cos 2θ and sin 2θ, respectively.
A more complete measure of the directional spectra from floating buoys has been developed but has
not been as successful as expected: the cloverleaf buoy that consists in three pitch-roll buoy linked to
3.3. RANDOM WAVES IN SITU OBSERVATIONS 39

each other (Mitsuyasu et al., 1975). In principle this layout provides a measure of the surface curvature
and the Fourier coefficient up to a8 and b8 . From a conventional buoy, one has to infer the function
S(f, θ) from only the four independent number a1 , b1 , a2 and b2 . This is not so much a measurement
problem, but rather one of choosing a statistical estimator. There are many method for this. Among
these, the Maximum Entropy Method (MEM Lygre and Krogstad, 1986), has the advantage of conserving
the angular moment a1 , b1 , a2 and b2 . The MEM method tends however to to give a bimodal shape
(two-peak spectra), which is often (Ewans, 1998) but not always realistic (Benoit et al., 1997).
Other recent analysis techniques aim at increasing the the directional resolution of this kind of
measurements. For instance, Donelan et al. (1996), have proposed an interesting method based on a
wavelet transform. Unfortunately their method assumes that for any given frequency the wave field is
dominated by waves coming for one single direction, which is not the case. This analysis yields a very
high directional resolution, and can be interesting to detect the presence of waves from a given direction,
but it cannot be interpreted as a directional spectrum as it leads to a low bias in the directional spreading.
Other methods can be biased and should be avoided. This is the case of the maximum likelihood method
(MLM) which yields output spectra that are systematically too broad and have moments a1 , b1 , a2 and
b2 that are different from the input parameters.

3.3.3 Pressure gauges

When surface elevation measurements are too expensive or not possible, which may be due to strong
currents incompatible with mooring lines, breaking waves in the surf zone, a usually good alternative
is the measurement from bottom-mounted sensors. The most common and robust are pressure gauges
that can be used to measure tidal elevations at the same time. To recover the surface elevation from
the pressure signal, we can invert the linear transfer function given by eq. (2.30), namely, for a sensor
at a height hd over the bottom, M = ρw g cosh(khd )/ cosh(kD), with k estimate from the frequency f
of each spectral component of the time series, using the Airy wave dispersion relation. In shallow water
and for waves of large amplitudes, this method often underestimates the wave height because the linear
dispersion is not very accurate (Filipot et al., 2010a; Bonneton and Lannes, 2017). The water depth D
is also given by the measurement of the mean pressure once it is corrected for the atmospheric pressure.
Because the elevation to pressure transfer function M decreases when k increases, it is usually impossible
to recover wave elevations for frequencies above a cut-off value fc . That value fc is a function of water
depth, instrument noise, background noise (usually due to currents), but also of the directional wave
spectrum. Indeed, figure 3.6 shows an example of data recorded in 100 m depth, in which the second
order pressure is larger than the linear pressure for frequencies above 0.13 Hz. As discussed in part 3,
this second order spectrum is a function of the frequency spectrum E(f ) but also of the directional wave
distribution. In that case, it is not possible to recover E(f ) for frequencies above 0.12 Hz. For example,
on day 4, the yellow-orange sloping line at frequencies 0.05 to 0.1 Hz is a due to swell waves arriving from
a distance of about 4000 km (see explanations in chapter 7, eq. (7.12)). At the same time, there is a
fainter blueish line at twice these frequencies which is caused by the second order effect. The vertical blue
stripes above 0.13 Hz are caused by the tidal current effects on the directional wave spectrum (Ardhuin
et al., 2013).

3.3.4 P-U-V sensors

As indicated by its name, this sensor measures the pressure p and the two horizontal components of
the velocity, u and v. It is indeed the combination of two instruments: a current meter (acoustic,
electromagnetic because a high sampling rate is required) and a pressure sensor, often piezo-electric.
This instrument is of particular interest as it is designed to be installed on the sea bottom. It is the
simplest mooring you may imagine, if one is not too worried about fishermen, for instance. Of course, it
is difficult ut not impossible to recover real time data (cables, acoustical modems with surface buoys...).
We have seen in Chapter 2 that the fluid pressure and velocities exponentially decay from the surface
to the bottom with a typical scale which correspond to the wave length 2π/k. The ”P-U-V” sensor is
thus perfect if one wish to measure the agitation at the bottom. To measure the wave heights one may
use the theory that provide transfer functions between pressure, velocities, elevation, etc (equation 3.14).
In this situation, the closer to the surface, the more reliable will be the measurement (e.g. an instrument
mounted on a floating of fixed platform).
40 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

0.5

frequency (Hz)
0
0.4
-20
0.3
0.2 -40

0.1 -60

0.0 -80
0 2 4 6 8 10
time (days)
Figure 3.6: Bottom pressure spectra
Time-frequency diagram of the pressure recorded over 11 days in 100 m depth off the French Atlantic
coast in October 2015. The colors show the pressure level in dB relative to 1 m2 /Hz times ρ2w g 2 . These
measurements were performed by a very sensitive Paroscientific piezo-electric sensor, included in a RBR-
duo system, with a noise level below -80 dB. At our depth of 100 m, the usual linear pressure signal,
with a level between -40 to 10 dB, can be used to recover the surface elevation spectrum for frequencies
below 0.12 Hz. At higher frequencies, the pressure is dominated by a second order effect due to waves in
opposite directions (e.g. Miche, 1944b; Ardhuin et al., 2013). That effect has very important consequences
for seismology, as discussed in part 3.

3.3.5 Sensor arrays and ADCPs

A set of wave gauge can be combined to record more covariances between the measurements. This kind
of measure allowed to get the first accurate spectra (Donelan et al., 1985) and is particularly used for the
air-sea interaction studies, for which the short waves spectrum is crucial (Graber et al., 2000; Pettersson
et al., 2003). It is essential to synchronize the sensors with an accuracy that is small compared to
the measured wave period. Similar techniques are employed in RADAR and SONAR technologies to
determine the sources of echoes. The original array processing algorithms used for waves were actually
developed for seismology.
These techniques have been largely applied to pressure sensors arrays with a number of statistical
methods for the spectrum estimation (e.g. Davis and Regier, 1977; Long and Hasselmann, 1979; Pawka,
1983; Herbers and Guza, 1990). An example of a spectrum is given in Figure 3.7 determined from
a coherent array of pressure sensors deployed in 8m depth on the site of the US Army Corps Field
Research at Duck, North Carolina. For large arrays, the underwater instruments positioning much be
very accurate. An acoustical positioning is generally used. The directional resolving power of the array
generally increases with the number of sensors in the array (see Kinsman, 1965). Arrays of pressure
sensors are excellent reference instruments for measuring the spectra of dominant waves, but they are
relatively expensive to deploy and maintain. O’Reilly et al. (1996) used such an array for the validation
of directional properties of two different types of buoys.
A recent and convenient alternative is the use of current profilers (ADCP). The combination of the
velocities measured by different acoustic beams, allows, in principle, for an interesting measure of the
directional spectra. However, the typical noise of up-looking ADCPs does not allow a higher angular
resolution than that of a simple P-U-V (Herbers and Lentz, 2010). The main benefit of ADCPs, however,
is the use of measurements close to the sea surface, where the wave motion is less attenuated than at the
bottom.
3.4. OPTICAL MEASUREMENTS 41

Hs: 1.64m
5
fp:0.0659 Hz
θpθ: 83.27 deg
(a) (b)
σθ p,: 14.87 deg
4

3
0.20 0.15 0.10 0.05 0.00

E ( f) (m2 s)
f (Hz or s-1)
2

0
0 19 38 57 76 95 0.05 fp 0.10 2fp 0.15 0.2 0
Spectral density E(f,θ) /f (m2 s2) Frequency (Hz)

Figure 3.7: (a) Example of frequency-direction wave spectra, divided by frequency, computed from 8m
depth pressure measurements in Duck, NC, October 19, 1994, 7:00 (EST). (b) Corresponding Frequency
spectra whose secondary peak matches the first harmonic of the spectral peak (f = 2fp ), and which is
likely due to the effect of nonlinear interactions that are of great importance in shallow water.

3.4 Optical measurements

Waves are usually the first thing that you see when looking at the sea. But turning beautiful pictures
into numbers for scientific analysis is not so easy. We will not discuss here the techniques that are mostly
used in the laboratory (e.g. light refraction across the air-sea interface), but instead we present the main
methods in use for application to real ocean waves.

3.4.1 From stereo-photography to stereo-video

The first methods used to measure wave shapes were based on stereo-photography: pairs of photographs
taken at the exact same time can be analyzed to produce a 3D map of the sea surface (Schumacher,
1939). The basic principle is to determine the (x, y, z = ζ(x, y)) coordinates of points that have been
identified in both images. This identification can be done using automatic correlation analysis. The
pair of pictures can be obtained from a single platform to cover a modest area and reveal interesting
details about the wave shapes (Banner et al., 1989), typically less than 30 by 30 m, or from a pair of
aircraft to get a view of a much broader region (Cote et al., 1960; Holthuijsen, 1983) and produce a
full directional wave spectrum. Now that everyone is carrying a digital camera around, and that stereo
processing is much more common, there are many opportunities to measure the full evolution of the
sea surface in space and time ζ(x, y, t). Recent efforts by Benetazzo (2006) and Gallego et al. (2011)
have demonstrated the capabilities of stereo processing, leading to new applications (Fedele et al., 2013;
Leckler et al., 2015). Latest developments include auto-calibration and the proper motion corrections
needed for ship deployments. A general issue that remains is that not all light conditions are favorable.
Alternatively, the use of more expensive infrared cameras or polarization cameras is a very interesting
extension for overcoming the variability of lighting conditions and the lack of texture at small scale for
a correlation analysis (Sutherland and Melville, 2013; Laxague et al., 2015).
The great advantage of having the full surface ζ(x, y, t), is that we can now measure a 3D spectrum,
without the need to use linear wave theory. This is most important for the short wave components, for
which nonlinear contributions are important. Figure 3.8 shows slices of the 3D spectrum at a constant
frequency. These are obtained from a stereo-video system installed 11 m above the water in a platform
in the Black Sea. The image processing uses the epipolar method: the position of the sea surface is
obtained only by a knowledge of the geometry of the camera system. This record from October 4, 2011,
was acquired when the wind speed was 14 /s and the wave peak frequency is fp = 0.33 Hz (Leckler et al.,
2015). The Non-linear contributions to the frequency spectrum dominate for f > 4fp . For example
at f = 1.2 Hz, there is more energy in the peak located near (kx = 0, ky = −3) than along the linear
dispersion relation shown wit the white dashed line. Such non-linear effects are also important for the
statistics of extreme wave heights, as shown by Fedele et al. (2013) and Benetazzo et al. (2017).
42 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

4 8
(a) 0.7 Hz, -30 dB (c) 1.0 Hz, -40 dB 10 (e) 1.4 Hz, -58 dB
6
2 4
5
2 0
0 0 0
−2
-5
ky (rad/m)

−5
−2 −4
−6
−10 -10
−4 −8
−4 −2 0 2 4 −5 0 5 −10 −5 0 5 10
5
(b) 0.8 Hz, -34 dB
8 15 -15
6 10
4
5
-20
2
0
0 0 -25
−2 −5
−4
−10 -30
−6
(d) 1.2 Hz, -50 dB (f) 1.6 Hz, -60 dB
−5 −8 −15
−5 0 5 −5 0 5 −10 0 10

kx (rad/m)

Figure 3.8: Slices of the double-sided spectrum for positive apparent frequencies 0.7, 0.8, 1.0, 1.2, 1.4
and 1.6 Hz. The energy appears in the direction from where it is coming. For each panel the color
scale spans 30 dB with the dark red corresponding to the power indicated on the figure (e.g. -30 dB)
relative to 1 m4 /Hz. Note that 1.4 and 1.6 Hz are twice 0.7 and 0.8 Hz, so that the first harmonic of
the components in (a) and (b) appear at approximately twice the wavenumbers in panels (e) and (f).
In each panel, the linear dispersion relation without current is plotted in black, and the white dashed
line gives the linear dispersion with a uniform current U = 0.15 m/s oriented towards the trigonometric
angle 99 degrees. The white dotted line marks approximately the separation between the linear part of
the spectrum and the faster non-linear components (Adapted from Leckler et al. 2015).

3.4.2 Using polarization and/or light intensity

Such a stereo system has difficulties in measuring waves with frequencies higher than 1.4 Hz that have
small heights. Other measurement techniques that are directly sensitive to slopes are better suited
for these shorter components: these include polarimetry (Zappa et al., 2008) or a measurement of the
radiance that can also be combined with the epipolar method (Gallego et al., 2011; Yurovskaya et al.,
2013).
Such a technique can also be applied to high resolution airborne or satellite imagery (with pixel sizes
less than 30 m in order to resolve waves). Kudryavtsev et al. (2017) have particularly taken advantage
of the O(1 s) time lag in the acquisition of the different color channels of the Multi Spectral Instrument
on board Sentinel-2. Clouds or haze obviously limit the application of optical methods, which is why
radar is often preferred for wave remote sensing.

3.4.3 Grazing angle radar

Rotating radars are usually deployed at the coast, for ship traffic monitoring, or on ships, for navigation
safety, providing a detection of ships, land, and sea ice. Most of these radar work in X-band, with a
wavelength around 3 cm, but S-band can also be used. These radars also record many echos from the
sea surface due to waves. This ‘sea clutter’ is another case of the saying, somebody’s noise is somebody
else’s signal. The clutter greatly limits the detection of ships at sea, but wave and current information
can be extracted from it (Young et al., 1985). Because of the low elevation afforded from a ship, the
incidence angle of radar waves on the sea surface is grazing. Hence, the sea surface exhibits shadows
behind wave crests and the relationship between the backscatter image and the sea surface is fairly
nonlinear. Still, the wave spectrum may be related to the radar image spectrum using a transfer function
(figure 3.9). The analyses procedure uses a sequence of several images, in order to decrease the noise.
The 3D (frequency-wavenumber-direction) spectrum is estimated to filter the data into ’wave patterns’
that follow the linear gravity wave dispersion relation, and other echoes and noise. This kind of system
generally gives a good spectral energy distribution, but the gain (proportionality factor between the
radar image spectrum and the surface elevation spectrum) is not very accurate, leading to typical errors
3.4. OPTICAL MEASUREMENTS 43

Figure 3.9: Examples of radar images recorded with a grazing radar (WAMOS system, Oceanwaves
GmbH), onboard the Hydrographic vessel Laplace. The measurements were undertaken in March 2003,
with a very long and high swell from the West (Hs = 6 m, Tp = 26 s). The spectrum shown here was
obtained in the shadow of Ouessant. The swell go around the islands. At the ship location part of the
energy has been refracted and arrives from the south.

of 20% or more on the estimate of Hs . An additional measurement form an altimeter, or a motion sensor
giving heave or pitch/roll from a ship, can provide an independent measure of this gain. In systems that
allow a measurement of the Doppler signal (e.g. Farquharson et al., 2005), the orbital velocities may be
measured, but the radar really measures the velocities of scatterers which – at this incidence – contains
also the phase velocities of steep waves. A side-product of the dispersion relation analysis is are maps
of surface currents, with sub-kilometer resolution, and water depth if the water is shallow enough. This
is probably the best instrument available today for the measurement of sub-mesoscale currents over a
footprint of about 10 km in diameter. A full exploitation of the dispersion relation can also provide an
estimate of the vertical current shear near the surface (Lund et al., 2015).

3.4.4 Surface wave radar

An extreme case of grazing angle occurs when the radar waves propagate along the surface. This kind
of propagation is possible in the HF-VHF range (from 2 to 50 MHz). This surface wave allows indeed to
get information beyond the horizon.
The radar reflection is well explained by Bragg theory, with the maximum backscatter power at the
Doppler frequency fD occuring for electromagnetic incident and reflected wave numbers vectors ki and
kr that match with the wavenumber kw = ±ki ∓ kr and and the wave frequency fD = f , where f and
kw are related by the linear dispersion relation. In the monostatic configuration where the receiving and
transmitting antenna are almost at the same place, we have ki = ke , kr = −ke , and kw = 2k2 .
The observed Doppler spectrum can be interpreted as the superposition of simple reflections (or first
order reflections) and multiple reflections. The extraction of sea state spectra is possible from the second
order that is a convolution of the spectrum (see for instance Wyatt, 2000). These second order echoes are
weaker than first order echoes at Bragg frequencies. The main echoes are used for currents measurement.
Their frequency provides a measurement of the Bragg wave phase velocity. The deviation of this phase
velocity from the expected linear wave phase speed can be interpreted as a surface drift current (Ardhuin
et al., 2009a). This current includes non-linear corrections to the phase velocity, which is like a filtered
Stokes drift (Stewart and Joy, 1974; Broche et al., 1983; Ardhuin et al., 2009b). The secondary peaks
are caused by either the multiple reflection of radar waves, or√the single reflection off nonlinear wave
components. In some cases a peak can be seen at the frequency 2fB which corresponds to√ the reflection
from the first harmonic of waves with wavenumber k = ±ke , which have a frequency, 2 gke in deep
water. Because different wave components have different sensitivities to the current profile, it is possible
44 CHAPTER 3. WAVE HEIGHTS AND SPECTRA: THEORY AND MEASUREMENT

Figure 3.10: Example of Doppler spectrum from a HF radar.

This measured spectrum comes from the Porspoder radar (Finistere, France). The radar frequency
is 12.4 MHz, corresponding to a wavelength Le = 2π/ke = 24 m. The main echoes is due to Bragg
scattering, which selects the ocean waves with a wavelength Le /2 = 12 m that propagate towards
(f > 0) or away from the radar (f < 0). Multiple scattering by waves with wavenumbers k1 and k2 ,
such that 2ke = k1 + k2 gives the ’second order echoes’.
√ At first order, i.e. for linear waves, the waves
of wavenumber kw = 2ke have a frequency fB = ± g2ke = ±0.36 Hz in deep water. The anomaly df
of the two highest peaks in the Doppler spectrum indicate that the phase speed 2πf /k is different from
the linear wave theory without current. the main reason for that difference is usually the presence of a
current, with a velocity Ur = 2πdf /(2ke ) in the direction of the radar.

to use that peak to measure the vertical shear of the current (Ivonin et al., 2004).
Chapter 4

Practical estimation of the wave

spectrum

4.1 General properties of discrete Fourier transforms

Time series can be obtained from many sensors, for example; a pressure gauge in the water at at
fixed depth, a range measurement from a laser or radar mounted on a platform or ship, velocity from
underwater acoustic or electromagnetic systems, accelerometer in a floating system. In the laboratory,
it is common to measure the surface elevation with resistive or capacitive wave gauges. These times
series thus consist of a signal ζ(t) or p(t) sampled at a fixed frequency fs , typically 1 < fs < 4 Hz in the
ocean, and fs ∼ 10 Hz or more in the laboratory because laboratory waves are generally shorter and
also, in the laboratory, the internal memory and power consumption of instruments is less of an issue:
either they are directly cabled to the acquisition system or the experiment is short enough. In general,
it is very important that fs is at least 4 times the expected frequency of the signal of interest. Indeed,
representing a cosine wave with only 4 points is already relatively coarse. As a result, some of the signal
(the very short waves) is not resolved.
Assuming we have such a series of discrete surface elevations ζ(n) with 1 ≤ n ≤ M . The duration of
the recording is (M − 1)/fs . All data processing softwares have a discrete Fourier transformation routine
that will provide
M
1 X
Zm = ζ(n)e−2iπ(m−1)(n−1)/M . (4.1)
M n=1

4.1.1 Spectral resolution

For any frequency index m, the complex amplitude Zm is large when the signal ζ actually contains
fluctuations at the frequency fm = (m − 1)/M fs . The complex amplitudes contain both amplitude and
phase information, i.e. Zm = |Zm | exp[i arg(Zm )]. The Fourier transform is simply a projection of the
signal onto the elementary functions that are the complex exponentials, or if you prefer sines and cosines,
with all frequencies fm = (m − 1)/M fs where m goes from 1 to M. We note that the difference between
frequencies fm and fm+1 is the spectral resolution df = fs /M . Because fs is the inverse of the sampling
interval dt, we have df = 1/(M dt), which is the inverse of the record duration. Namely, the frequency
resolution is the same as the lowest frequency which is at m = 2, and it such that the lowest period
equals the record duration. We recall that m = 1 gives the average value of the signal.

4.1.2 Nyquist frequency

The decomposition of a discretized real signal with M pieces of information into M frequencies cannot
produce M amplitudes and frequencies that are independent, that would be 2M pieces of information.
Indeed, the frequencies from M/2fs to M fs do contain the same information as those between 0 (M/2 −
1)fs , as shown in figure 4.1.
With the definition given by eq. (4.1), Zm = ZM +2−m , where the overbar is the complex conjugate,
so that the modululs of the spectrum is symmetric around m = M/2. This index M/2 corresponds
to the Nyquist frequency fN = fs /2 = M/2df . The symmetry means that above fN there is no new

45
46 CHAPTER 4. PRACTICAL ESTIMATION OF THE WAVE SPECTRUM

0.6 0.04 1
0.5
(a) 0.035
(b) (c)
0.4 0.03 0.5
surface elevation (m)

0.3 0.025

|Z|/M (m)

phase / π
0.2 0.02 0
0.1 0.015
0 0.01 −0.5
−0.1 0.005
−0.2 0 −1
0 100 200 300 400 500 0 100 200 300 400 500 0 100 200 300 400 500
time index frequency index frequency index

Figure 4.1: (a) Example of surface elevation time series sampled at 12 Hz, (b) Discrete Fourier transform
amplitude of the signal, and (c) phase of the same signal.

information. In other words, fN is the highest resolvable frequency. At this Nyquist frequency, a cosine
wave is only represented by 2 points over a period.
The continuous spectral density F (f ) is obtained in the limit when the spectral resolution df goes to
zero of the following expression

Zm Zm
F (f = (m − 1)fs /M ) = 2 . (4.2)
df

The factor 2 comes from the combination of positive and negative frequencies in the interval [0, fN ].
That definition is called the single-sided spectrum. For some applications it may be convenient to keep
the double-sided spectrum defined without this factor 2, for f in the range [0, 2fN ] or [−fN , fN ]. In
practice, the zero spectral resolution is never achieved because it corresponds to an infinitely long record.
We are therefore stuck to finite spectral resolutions df , and thus a discrete spectrum sampled at df ,
corresponding to record lengths over which the random variable of interest is stationary.

4.1.3 Co-spectra
In the same manner, we can define the co-spectrum of two variables a and b having Fourier transforms
dA and dB,
Am Bm
Cab (f = (m − 1)fs /M ) = = P (f ) + iQ(f ), (4.3)
df
where P and Q are real numbers, the co-spectra in phase and quadrature.
We note that the product of two Fourier transforms is the Fourier transform of the convolution of
the two functions. When these two functions are the same, this result tells us that the spectrum is the
Fourier transform of the auto-correlation function. When the two functions are different, the co-spectrum
is the Fourier transform of the correlation function.

4.2 Spectra from time series

4.2.1 Filtering of data and aliasing
For any physical parameter, the complex amplitude at a frequency above fN is the complex conjugate
of the amplitude at a frequency below fN : dZ(N − m) = dZ(m). In other words, the energy above the
Nyquist frequency is aliased at lower frequencies. This is one reason for applying a low-pass filter before
the Fourier transform, in order to remove signals that would otherwise be aliased.
For example, a signal of period 1 s sampled every second (fs = 1 Hz) is constant. In general, any
signal with frequency higher than the Nyquist frequency gives an apparent frequency in the range [0fs /2],
with a value obtained by folding back the spectrum along a vertical line at f = fs /2 or f = 0 as many
times as necessary, until falling in [0fs /2]. Figure 4.2 shows an example of a true signal (in red) with
frequency at 9/10 which gives 1-9/10=1/10 by aliasing. This is well known problem in oceanography in
the case of the measurement of tides from satellite data. In the case of ocean waves, it is really necessary
to measure waves with a sampling frequency that is at least 4 times the dominant frequency.
In practice the high frequency roll-off of the wave spectrum, for f > fp in particular for pressure
or velocity at the ocean floor, means that this filtering may not be necessary for the surface elevation.
 4.2. SPECTRA FROM TIME SERIES 47

0 1 2 3 4 5 6 7 8 9 10

Figure 4.2: Example of spectral aliasing. A signal with period 10/9, in red, and sampled with a step of
1 (black dots) gives an apparent period of 10, in blue. (Moxfyre, wikimedia commons).

Figure 4.3 shows an example of the filtering of a surface elevation record and its impact on the spectrum.
Here the filter helps in reducing the noise associated with the measurement method. The worst filter is

−1
0.6 0.6 10
95% confidence
−2 interval
0.4 0.4 10
surface elevation (m)

surface elevation (m)

−3
10

E(f) (m2/Hz)
0.2 0.2
−4
10
0 0 raw data
−5
10 3−point filter
−0.2 −0.2
−6 3−point boxcar
10
−0.4 −0.4
−2 −1 0
1420 1425 1430 1435 1440 0 500 1000 1500 10 10 10
time (s) time (s) frequency (Hz)

Figure 4.3: Example of surface elevation measured by stereo-video from the Katsiveli platform, in 20 m
depth, on Octoebr 4, 2011 (see Leckler et al., 2015; Aubourg et al., 2017, for further detailed analysis
of this dataset). (a) small piece of record around t = 1430 s with (red) and without (black) a 3-point
smoothing filter (b) full record (1780 s) (c) spectra of the time series with different filters applied.

clearly the boxcar (or moving average) filter with equal weight given to the consecutive data values, this
does not suppress very well the shortest components.

4.2.2 Windows, Gibbs phenomenon and averaging

The discrete Fourier transform of signal ζ that takes values at locations 1 to N, corresponds to the
Fourier transform of a periodic signal that would repeat itself with ζ(n + mM ) = ζ(n) for any n and m.
If ζ(1) ̸= ζ(M ), then this periodic function has a sharp jump from ζ(M ) to ζ(M + 1). Such a jump can
give a strong spectral signature, all across the spectrum.
This artifact is known as the Gibbs phenomenon, and it is generally removed by multiplying the
signal ζ(n) by a window function W (n) which goes to zero or very small values for both n = 1 and
n = N . Obviously the spectrum of ζ(n) × W (n) is different from the spectrum of ζ(n). The first obvious
effect is that the variance of the signal has been reduced by a factor that is the average of W 2 . That is
easily corrected for.
Let use the time series shown in figure 4.3.b. We first start with a small piece of only 1000 points. The
original time series, in black in figure 4.4.a gives the power spectral density in figure fig:anaspec:spectre1.b.
If the time series is brought to zero at both ends (in red) by multiplying with a Hann window, then the
spectrum is transformed. However, because of the random fluctuations of the spectral estimates this is
not obvious. Indeed waves are random and the surface elevation is nearly-Gaussian. As a result, the
complex Fourier amplitudes are also random and Gaussian, so that they modulus, which is the power
spectral density has the shape of a χ2n distribution with n = 2 degrees of freedom. Because the χ2n
distribution has a mean of n, the distribution of estimates E(f b ) of the spectrum, once normalized by the
2
mean spectrum E(f ) follows the rescaled χn disitribution, with a mean equal to 1.
Using figure 4.5, we find that the expected ratio of the lower and upper bound of a 95% confidence
interval is 146. This number is the ratio of χ22,0.975 and χ22,0.025 , that give the probabilities that χ2n > χ2n,α
is equal to the acceptance threshold α.
 48 CHAPTER 4. PRACTICAL ESTIMATION OF THE WAVE SPECTRUM

0.6
(a) (b) (c)
−2
0.4 10
surface elevation (m)

0.2

E(f) (m /Hz)
−4

2
10 95% confidence
0 interval

−0.2 95% confidence

−6
10 interval

−0.4
−2 −1 0 −2 −1 0
0 20 40 60 80 10 10 10 10 10 10
time (s) frequency (Hz) frequency (Hz)

Figure 4.4: (a) First 1000 points of time series shown in (4.3).b, with (red) and without (black) multi-
plication by a Hann window. (b) Resulting spectra (c) Average spectra using Welch’s method with 21
independent windows and thus 42 degrees of freedom.

For an expected value E, b the Gaussian statistics theory predicts a 95% probability that the estimate
b × 0.0506/2 and E2 = E
of E(f ) is in the range [E1 , E2 ] with E1 = E b × 7.38/2, where 0.0506 and 7.38 are
the values for which the χ2 cumulative probability function for 2 degrees of freedom is α = 0.975 and
α = 0.025, respectively as given 4.5. In other words, the random fluctuations of the spectrum estimated
by a single Fourier transform spans more than two orders of magnitude.
That is fairly annoying. There are two ways to reduce this uncertainty: the first, which is most easily
done in the laboratory is to run more experiments, repeating the same conditions, with random wave
phases, and average the results. For measurements in the field, this is the same as processing a longer
time series, but it only makes sense if the conditions are stationary: same wind, same wave age, same
current, etc. In practice, this stationarity constraints limits the length of records from half an hour to a
few hours. As we average N spectra together, the number of degrees of freedom increases √ to 2N . Here
we have used 21 independent spectra, and we get an uncertainty that narrows like 1/ N for large N .
For 20 spectra, the ratio E2 /E1 is 59.34/34.43 ≃ 1.7. As N increases, the spectral resolution becomes
coarser.
Because the Hann window practically removes part of the data, Welch (1967) has defined a method
in which the windows are shifted by half their length, as shown in figure 4.6.a. The lower panel shows
the 21+20 spectra estimated, and the average result.
Another method that is almost equivalent to Welch’s is the smoothing of the spectrum, also known
as band-averaging. Because the Fourier transform of a shorter window has a coarser resolution, both
methods effectively trade off the spectrum accuracy against the spectral resolution. An extension of such
methods is the use of wavelet transforms (e.g. Liu and Babanin, 2004) which aims at localizing events in
both time and frequency.
Whatever the choice of method, without any prior knowledge on the signal, the product of the spectra
uncertainty and the square root of the frequency uncertainty remains constant. Thus the optimal choice
of df is up to the user. For wind seas, the typical relative width of the spectrum is 0.1, and for a typical
peak period of 0.1 Hz, resolving the peak requires df < 0.01 Hz. For swells one may like to have an even
narrower frequency resolution. In the example above, df = 0.012 Hz is enough for the relatively short
wind sea found in the Black Sea.

4.2.3 Interpretations and further developments

The general idea of spectral analysis is to decompose a signal into its basic constituents. If waves
were indeed linear, the Fourier components would be truly independent and the spectrum would give
the energy of different wave components. In practice, there is a significant level of nonlinearity, which
actually dominates the frequency spectrum E(f ) at frequencies above 3 to 4 times the wind sea peak (e.g.
Leckler et al., 2015). As a result, the interpretation of these high frequencies (f > 1 Hz in the example
above) as the energy of linear waves is wrong. Several methods have been developed to try to separate
linear and non-linear components, including higher order analysis (Hasselmann et al., 1963) which is
illustrated in chapter 17. This question has inspired other nonlinear methods, such as the Empirical
Mode Decomposition method by Huang et al. (1998).
4.2. SPECTRA FROM TIME SERIES 49

Figure 4.5: Table of χ2 distributions. This table gives the confidence intervals for the estimation of
spectra with n = 2N degrees of freedom, where N is the number of independent spectra used.
50 CHAPTER 4. PRACTICAL ESTIMATION OF THE WAVE SPECTRUM

1.5 4121
1 22 2 23 3 ...
1
0.5
0
0 200 400 600 800 1000 1200 1400 1600
time (s)

−1
10
−2
10
E(f) (m2/Hz)

2 d.o.f.
−3
10
−4
10 42 d.o.f.

−5
10
−6
10 −2 −1 0
10 10 10
frequency (Hz)
Figure 4.6: (a) Succession of the Hann windows applied to the data, with 21 independent windows and
20 overlapping windows, from number 22 to 41. The solid line is the mean spectrum and the dotted and
dashed line show the expected 95% confidence intervals for 2 and 42 degrees of freedom. à 95%.

4.3 Spectral analysis of directional buoy data

4.3.1 Case of 3-axis displacements or accelerations

We have seen in section 3.2.1 that the spectrum of x-component velocity at the ocean bottom is given
from the surface elevation spectrum multiplied by a transfer function,

σ 2 cos2 θ
EU x (f, θ) = E (f, θ) . (4.4)
sinh2 (kD)

The same method applies to spectra of displacements, velocity and slopes at the sea surface. For a
water particle at the surface, the spectra of displacement in the three directions are given by (2.32),

Z 2π
1
Ex (f ) = 2 E (f, θ) cos2 θdθ (4.5)
tanh (kD) 0
Z 2π
1
Ey (f ) = E (f, θ) sin2 θdθ (4.6)
tanh2 (kD) 0
Z 2π
Ez (f ) = E (f, θ) dθ. (4.7)
0

We note that this last spectrum is the usual elevation spectrum E(f ), also called heave spectrum, with a
minor modification due to the fact that it is not obtained at a fixed position (x, y) but at a positions that
moves with x and y. As a result, the shape of the waves and the shape of the spectrum are modified,
with a strong reduction in the contribution of nonlinear harmonics: a surface buoy signal looks much
more linear than a wave staff or stereo video record.
4.3. SPECTRAL ANALYSIS OF DIRECTIONAL BUOY DATA 51

The co-spectra of horizontal and vertical displacements are

Z 2π
i
Cxz (f ) = E (f, θ) cos θdθ, (4.8)
tanh(kD) 0
Z 2π
i
Cyz (f ) = E (f, θ) sin θdθ. (4.9)
tanh(kD) 0
Z 2π
1
Cxy (f ) = E (f, θ) sin θ cos θdθ. (4.10)
tanh2 (kD) 0

These co-spectra are thus related to the mean direction and directional spread, through the directional
moments introduced in section 3.2.2
Z 2π
a1 (f ) = E (f, θ) cos θdθ, (4.11)
0
Z 2π
b1 (f ) = E (f, θ) sin θdθ, (4.12)
0
Z 2π
a2 (f ) = E (f, θ) cos(2θ)dθ, (4.13)
0
Z 2π
b2 (f ) = E (f, θ) sin(2θ)dθ. (4.14)
0

To summarize, starting from the displacement time series x(t), y(t)) and z(t) one obtains the spectra
and co-spectra Cxx (f ), Cyy (f ), Czz (f ), Cxz (f ), Cyz (f ) , Cxy (f ). Using cos(2θ) = cos2 (θ) − sin2 (θ) and
sin(2θ) = 2 sin θ cos θ, they give the following directional moments, in which ℑ stands for the imaginary
part,

a1 (f ) = −ℑ(Cxz (f ))/ [Czz (f ) (Cxx (f ) + Cyy (f ))] (4.15)

b1 (f ) = −ℑ(Cxy (f ))/ [Czz (f ) (Cxx (f ) + Cyy (f ))] (4.16)
a2 (f ) = (Cxx (f ) − Cyy (f ))/(Cxx (f ) + Cyy (f )) (4.17)
b2 (f ) = 2Cxy (f )/(Cxx (f ) + Cyy (f )) (4.18)
(4.19)

from which we get directional parameters, with mod the modulo operator,

θ1 (f ) = mod (270. − atan2(b1, a1) × 180/π, 360) (4.20)

  q 0.5
σ1 (f ) = 2 2
2 1 − a1 (f ) + b1 (f ) × 180/π (4.21)

θ2 (f ) = mod (270. − 0.5atan2(b2, a2) × 180/π, 360) (4.22)

  q 0.5
σ2 (f ) = 0.5 1 − a22 (f ) + b22 (f ) × 180/π (4.23)

These 4 parameters can be used in statistical estimators to obtain the directional spectrum E(f, θ).
A commonly used estimator is the Maximum Entropy Method of Lygre and Krogstad (1986). See also
the review by Benoit et al. (1997).

4.3.2 Case of other systems with 3 variables or more

in the above method, we can replace x and y by the slopes ∂ζ/∂x and ∂ζ/∂y as done in the pitch-and-roll
buoys developed by Cartwright and Smith (1964) and still widely used. For example, most of the 3-m
discuss buoys operated by the U.S. National Data Buoy Center are based on this measurement (Steele
et al., 1992). Also, the horizontal velocities at any given level, (u, v). These give access to the heave
spectrum E(f ) and the same directional moments a1 , b1 , a2 and b2 .
In order to go beyond these first 5 parameters, one can built an array of instruments. The cloverleaf
buoy of Mitsuyasu et al. (1975) was such an attempt, and arrays of pressure gages or lasers have been
routinely used. Today’s optical techniques (Benetazzo, 2006; Fedele et al., 2013; Laxague et al., 2015)
are other ways to get to more details about the sea surface.
52 CHAPTER 4. PRACTICAL ESTIMATION OF THE WAVE SPECTRUM

Figure 4.7: Definition of parameters M ,m, T1 et T2 used to estimate the amplitude a = (M − m)/2,
period T = T1 +T2 , vertical asymmetry α = (M +m)/2 and horizontal asymmetry aπ(T1 −T2 )/(T1 +T2 )/2
(from Elfouhaily et al. (2003), ©Elsevier). Here m < 0, because m is defined as the trough level between
two up-crossing zeros that define the start and end of the wave.

4.4 Some links between spectral and wave-by-wave analysis

The spectrum E(f ) gives the distribution of the surface elevation variance as a function of time scales.
As discussed in chapter 1, one can also chop the signal in individual waves and study the statistics of
their properties. A model for the sea surface elevation could be a signal of constant amplitude with a
random frequency modulation. In that case, Woodward (1952)’s theorem tells us that the spectrum of
the signal is the distribution of its frequencies Pf (f − fc ) where fc is the carrier frequency,

a2
E(f ) = Pf (f − fc ). (4.24)
2
Extension of that theorem by Blachman and McAlpine (1969) with an additional amplitude modula-
tion makes it applicable to ocean waves. As shown by Elfouhaily et al. (2003) the spectrum is then given
by the joint probability distribution of wave heights and periods P (a, f ) with a = H/2 and f = 1/T .
Separating linear (naked) and full (dressed) spectra, Elfouhaily et al. (2003) showed that the peak region
is dominated by linear components
Z
1
Ebare (f ) = a2 P (a, f )da. (4.25)
2
At high frequency the nonlinear contributions dominate the spectrum which can be interpreted as
fast modulations of the short components. Taking into account that effect leads to asymmetries α et β
(see figure 4.7), and the dressed spectrum
Z Z 
1
Edressed (f ) = Enu (f ) + α2 P (α, f /2)da + β 2 P (β, f /2)da ≃ E(f ) (4.26)
2
 Chapter 5

Wave groups and fluctuations of

wave parameters

Before we look at the evolution of the wave field over tens of kilometers and hours time scale, it is
worth looking at consequences of the shape of the wave spectrum on fluctuations in wave properties at
the scale of few minutes and kilometers. Indeed, the fact that waves are random introduces small scale
variations. Most early work was focused on defining the statistics of series of high waves (Arhan and
Ezraty, 1978; Masson and Chandler, 1993), which can be useful for example when catching waves on
a surfboard, avoid the high waves when navigating a landing craft through the surf zone, or landing a
helicopter on a ship. In this chapter we will start with another application which has become prominent
as we are starting to look at smaller and smaller scale details in the wave field: estimating the expected
fluctuations associated with groups (De Carlo et al., 2023), so that we may separate it from other effects,
including refraction induced by currents and water depth, wave breaking, etc, which will be discussed in
the following chapters. This investigation will also allow us to estimate lower bounds for uncertainties
of wave measurements that will be defined from the time and space footprint of the measurements. The
full uncertainty also contains instrument noise and measurement noise effects.

5.1 Wave envelope, local amplitudes and their statistics

Let ζc be the complex number such that ζ = Re(ζc ) is the free surface, ζc is usally called the analytic
signal. The envelope η of the signal is defined by η = |ζc |, with an example shown in Fig. 5.1, using
bottom pressure p(z = −h) instead of surface elevation ζ. This defines a local amplitude of the signal

1.5 2
10
(a) (b)
1
10 0
0.5
E(f) (m 2 /Hz)
p (m)

0 -2
10
spectrum of signal
-0.5
-4
Ψ(f)= spectrum of envelope
10 spectrum of abs(signal)×π/2
-1 detrended p A× spectrum of signal²
abs(detrended p) A×Ψ2(f)= spectrum of envelope²
envelope A=(8-2π)/(16 var(signal))
-1.5 -6
10
0 50 100 150 200 250 300 350 400 450 500 0.01 0.02 0.04 0.1 0.2
time (s) f (Hz)

Figure 5.1: (a) time series of a signal, here the detrended ocean bottom pressure in Berthaume on
January 31st 2004 (same data as in Figure 1.2), absolute value of the signal and signal envelope. (b)
Spectra of the signal, the envelope, the rectified signal, and the approximate spectrum Ψ2 obtained from
a spectral convolution.

that is continously defined everywhere. One interesting property of the envelope is that it does not
contain the small scale crest-to-trough (positive to negative) variations of the original surface, so that
its spectrum actually contains only longer components (larger periods in the case of time series). It is

53
54 CHAPTER 5. WAVE GROUPS AND FLUCTUATIONS OF WAVE PARAMETERS

interesting to note that a similar operation is obtained by rectifying the signal, i.e. taking its absolute
value. This was particularly studied for electric signals by Rice (1944) and the rectifier in that case is
a simple diode bridge. In the case of ocean waves, an important application is that of forces on moored
ships, and you can think of the mooring line as a kind of rectifier, which will thus endure low-frequency
forces.
In fact it is easy to show that the signal and the rectified signal have the same mean squared value
(i.e. the same integral of the spectrum including f = 0), but they have very different spectra: the mean
squared of the rectified signal is actually split half and half between sub-harmonics (very low frequencies)
and super harmonics (higher than the peak frequency). Because the mean envelope squared is twice the
mean rectified signal squared (if it is not obvious, think about it), then the envelope spectrum happens
to be 4 times the spectrum of the rectified signal at low frequencies.
The spectrum of the envelope turns out to be a pretty important quantity for many applications.
Indeed, any non-linear effect will typically give the same kind of sub-harmonic and super-harmonic. In
particular, any measurement device that is weakly non-linear will produce spurious large scale fluctua-
tions: this is the case of mooring lines for floating buoys, it is also the case of the KaRIN radar on board
the SWOT satellite (Peral et al., 2015). Separating the spurious signal from the real signal can involve
the spectrum of the envelope. So you may be pleased to know that we can compute the spectrum of the
envelope from the wave spectrum itself using the following approximation (Tayfun and Lo, 1989)

8 − 2π
Ψ≃ Ψ2 , (5.1)
Hs2

where Ψ2 is the spectrum of the envelope squared and is obtain by convolving the wave spectrum with
itself. This expression is valid for the surface elevation envelope. For any other quantity (orbital velocity,
Stokes drift ...), just replace Hs by 4 times the standard deviation of the quantity of interest. And before
you complain that this is only an approximation, you may note that we have an exact equation for the
spectrum of the envelope squared: if you enjoy the satifying beauty of exact equations please consider
studying the fluctuations of the variances instead of the fluctuations of the amplitudes, and things will
be very nice.
We will now use this envelope spectrum to compute the expected fluctuations of wave measurements
when estimated from a short record or a small area: here ”short” or ”small” is relative to the size
of a few wave groups in time and space. The concept is the same when working with time series or
two-dimensional maps, so we will treat both situations, starting with time series.

5.1.1 Time series

From a time series ζ(t), the envelope can also be defined using the Hilbert transform H(ζ) of the time
series,
η = |ζ + H(ζ)| . (5.2)
A practical calculation of the Hilbert transform is obtained using Fourier and inverse Fourier transforms.

Definition of a local wave height

From the envelope η we define the wave height Hr as an average over a time segment of ”radius” τ
r
2
Hτ (t) = 4 (η ⊗ gτ )(t) (5.3)
π
where ⊗ is the convolution operator and gτ is a filtering kernel of radius τ , more explicitly
r Z τ
2
Hτ (t) = 4 η(t + u)gτ (u)du (5.4)
π −τ

Under the Gaussian approximation for the distribution of sea surface p elevations this time average
actually converges to the usual significant wave height Hs : the factor 2/π is there simply to correct
for the fact that the envelope is always above the rectified signal.
Now, you may think of this filter gτ as your ”observation operator”. I know, for time series it sounds
a bit silly but there is always some kind of filter in the instrument, which can be the sensor itself or the
effect of the structure that it is mounted on. If we consider the most simple case, our filter will be a
5.1. WAVE ENVELOPE, LOCAL AMPLITUDES AND THEIR STATISTICS 55

box-car, taking the constant value 1/2τ between −τ and τ and zero otherwise. What if we estimated
the significant wave height as H b τ , with τ = 64 s? Some of these estimates would occur when a group
of large waves is present, giving a large value, and others would give a lower value. In general, we may
expect a variability, quantified by a variance var(Hτ ).
We can estimate this variance from the spectrum of the envelope Ψ(f ), by summing all the contri-
butions from scales longer than 4τ , i.e. frequencies lower than 1/(4τ ),
Z 1/(4τ )
32
var(Hτ ) = Ψ(f )df. (5.5)
π 0

We may also use the fact that the (single-sided) spectrum of the envelope squared Ψ2 (f ) is the convolution
of the spectrum of the single-sided surface elevation spectrum E(f ) by itself,
Z ∞
Ψ2 (f ) = 8 E(u)E(u + f )du. (5.6)
0

In practice people have rather studied the variations of Hs and not that of Hs2 , Although the details
of the theory are more complex, the important result is that, for low frequencies, the spectrum of the
envelope Ψ(f ) has the same shape as the spectrum of the envelope squared Ψ2 (f ) (Rice, 1944). More
specifically, Tayfun and Lo (1989) have showed that a good approximation for Ψ is given by eq. (5.1).
If τ is large enough, then the frequency 1/(2τ ) will fall in the region where Ψ(f ) ≃ Ψ(f = 0), and we
can estimate the variance of H b τ as
Z ∞
32 Ψ(f = 0) 16(16 − 4π) 4−π 2 2
var(Hτ ) ≃ ≃ 8 (E(f ))2 df = H Q (5.7)
π 4τ 4πτ Hs2 0 2πτ s f

with the frequency peakedness defined as the reciprocal of the frequency bandwidth (Saulnier et al.,
2011), R∞ 2
2 E (f )df
Qf = R0∞
2 . (5.8)
0
E(f )df
Qf is a parameter that characterizes the shape of the frequency spectrum, and that is generally well
correlated with the period Tm0,−1 .
We note that almost the same result for the uncertainty of estimates Hτ can be obtained by starting
from the χ2 distribution for spectral densities. Young (1986) noted that E is the sum of χ2 distributed
random variables and therefore is also χ2 distributed, with a number of degrees of freedom equal to

νf,H = 2/Q2f df (5.9)

where df is the frequency resolution of the spectral analysis, corresponding to df = 1/(2τ ). The wave
height estimate Hτ is thus a χ-distributed random variable with its variance and mean given by functions
of νf,H and a ratio s
std(Hτ ) Γ2 (νf,H /2)νf,H
= − 1. (5.10)
mean(Hτ ) 2Γ2 ((νf,H + 1)/2)
For most practical applications, the number of degrees of freedom is large enough and we can use
std(Hτ ) √
≃ 0.5Qf / 2τ , (5.11)
mean(Hτ )
where Γ is the Euler gamma function.
Figure 5.2.a shows that the predicted variability is not exactly like the estimated variability, p
but it
provides a useful order of magnitude, with the ”noise” on wave height estimate decreasing like 1/τ ,
which is why wave records are normally taken over 20 minutes. In the present case, using 2τ = 20
minutes
p averaging instead of averaging over 2τ = 3 minutes reduces statistical uncertainties by a factor
20/3 = 2.6, giving a mean value of std(Hτ )/Hs = 6.4%, which is of the same order of magnitude as
the 5% relative uncertainty for buoy data for wave heights around 2 m that is estimated from triple-
collocation techniques (Dodet et al., 2022). This suggests that sampling errors caused by wave groups
are a large part of the uncertainty in buoy data.
It is always possible to average over longer times but at some point one loses the time resolution
that may be needed to investigate how the wave height varies during the tidal cycle or due to other fast
 56 CHAPTER 5. WAVE GROUPS AND FLUCTUATIONS OF WAVE PARAMETERS

0.4 7 0.4
(a) (b) (c)
0.35 6 0.35

0.3 0.3
5

std(Uτ)/Ub,rms
std(Hτ)/Hs

0.25 0.25
4
0.2 0.2

Qf
3
0.15 0.15
2
0.1 0.1

0.05 1 0.05

0 0 0
0 0.1 0.2 0.3 0.4 5 10 15 20 0 0.1 0.2 0.3 0.4
Qf √(4-π)/2πτ Tm0,-1 Qf,u√(4-π)/2πτ

Figure 5.2: Variability of measured parameters over 21 days of measurements in Bertheaume bay, January
2004 using τ = 180 s and each symbol corresponds to 1 hour of data: (a) Standard deviation of Hτ
normalized by Hs as a function of the predicted variability using Qf (b) correlation of Qf and the
mean period Tm0,−1 , (c) variability of bottom orbital velocity amplitude Ub,rms against the relevant Qf,u
peakedness parameter.

evolving phenomena. We also note that quantities that involve higher frequency moments of order n such
as the orbital velocity with n = 2 are less ”groupy”, their spectrum shape given by E(f )f n are broader
than E(f ), and we can compute a similar Qf for these parameters that will be lower, as shown in 5.2.c,
using the bottom orbital velocity data from the Nortek Vector instrument. I used pressure and velocity
at the ocean bottom in Figure 5.2.a and 5.2.c, and these are already filtered by the water depth, so that
the difference between these Qf and Qf,u is not as large as it would be if one used surface elevation and
surface velocity.

5.1.2 Spatial maps

Following the same steps we took for time series, we may define the wave height meeasured by altimeters
HL as a spatial average of the envelope over a square of side L. The wave height estimate Hτ is again
χ-distributed random variable with its variance and mean given by functions of νf,H and their ratio given
by eq. 5.10. The only difference is that now νf,H is replaced by

νkk,H = 1/Q2kk,H dkx dky (5.12)

where dkx dky is the spectral resolution of the spectral analysis, corresponding to dkx dky = (2π)2 /(2L)2 ,
with R∞ R∞ 2
E (kx , ky )dkx dky
Qkk = R−∞ R−∞
2
2 , (5.13)
∞ ∞
−∞ −∞
E(k ,
x y k )dk x dk y

where E(kx , ky ) is the centrally symmetric double-sided spectrum. If one uses the single-sided spectrum
instead, as we did for frequency spectra in eq. (5.8), the numerator of eq. (5.12) should be 2, as in 5.9.
This expression is easily verified by taking a wave spectrum E(kx , ky ), simulating the ocean surface
map ζ(x, y) using random phases, and computing the variance over square tiles of side L, as illustrated
in Fig. 5.3. The theory, combining eq. (5.12) and eq. (5.10) gives a good approximation, in cases where
the tile size is much larger than the dominant wavelength.

5.1.3 Along-track variability in altimeter data

Replacing our boxcar window for a square tile of side length L with a disc of radius r we can imagine
that the local height is some convolution of the envelope
r
2
Hr (x, y) = 4 (η ⊗ gr )(x, y). (5.14)
π

In practice a single altimeter gives estimates along a single track in the (x, y) plane and the filter function
gr can be a little complicated. For Delay-only altimeters the detailed form of gr was derived in De Carlo
5.1. WAVE ENVELOPE, LOCAL AMPLITUDES AND THEIR STATISTICS 57

std(HL)/mean(HL)

10 1

Qkk=43 m
10 2 average
theory
Qkk=15 m
average
10 3 theory
103 104
L (m)
Figure 5.3: Variability of HL in simulated surfaces from 2 different wave spectra, one with
Qkk = 43 m and the other with Qkk = 15 m. These two spectra are shown below in Fig.
5.5. for each spectrum, 10 surface realisations were generated using random phases. This figure
was generated with the notebook https://github.com/ardhuin/waves_in_geosciences/blob/main/
NOTEBOOKS/chapter_groups_figure_Qkk_stdH_2D_maps.ipynb

et al. (2023), and for our purpose a good approximation is a a Gaussian filter of radius ra = rC /4.5,
giving a filter Gra = exp (−k 2 ra2 ) for the power spectrum in Fourier space.
Using the double-sided wave spectrum E(kx , ky ) of the surface elevation, defined for (kx , ky ) in the
entire wavenumber plane and centrally symmetric, the region of the envelope spectrum for k ≪ kp , with
kp the wavenumber peak, is proportional to
Z ∞Z ∞
Ψ2 (kx , ky ) = 8 E(u, v)E(u + kx , v + ky )dudv, (5.15)
−∞ −∞

in which Ψ2 is also double-sided. From eq. (5.14), the spectrum of Hr is

32 32 8 − 2π
ΨHr (kx , ky ) = Ψ(k)Gr (kx , ky ) = Ψ2 (kx , ky ) Gr (kx , ky ) (5.16)
π π Hs2
| {z }
ΨH0 (kx ,ky )

with Hs the usual significant wave height and Gr = Gra when considering a single altimeter measurement.
We will now average data along the track, which is the usual practice, at least over 0.05 s corresponding
to the 20 Hz rawest data downlinked for most instruments, and corresponding to 350 m distance along
the track. For delay-only measurements, these data are fairly noisy and most users typically use at least
1 Hz data (averaged over 7 km) or longer time averages. For simplicity we take the satellite track along
the x-axis, and we consider an averaging length d1 . Integrating Ψ for kx > k1 , amounts to integrating
ΨHr (kx , ky ) to get the expected variance up to the cut-off wavenumber along kx , k1 ≃ π/d1 , giving
var(Hr , d1 ), the group-induced fluctuations of H
bs
Z ∞Z
var(Hr , d1 ) = ΨH0 (kx , ky )Gr (kx , ky )dkx dky . (5.17)
−∞ kx >|k1 |

Here again, if the filter Gr is zero outside of a very small range of wavenumbers, we can approximate
ΨH0 (kx , ky ) ≃ ΨH0 (kx = 0, ky = 0) and take this value out of the integral, and the integral is related to
58 CHAPTER 5. WAVE GROUPS AND FLUCTUATIONS OF WAVE PARAMETERS

the effective area of the filter in the wavenumber plane which we approximate as a disk of radius 1/ra ,
with an area π/ra2 , given by the integral of Gr over the entire wavenumber plane, and we remove a band
of width 2k1 and length 2/ra because we are only averaging over d1 so that wavelengths longer than 2d1
cannot contribute to our signal and must be excluded. This gives,
Z ∞Z
var(Hr , d1 ) ≃ ΨH0 (0, 0) Gr (kx , ky )dkx dky
−∞ kx >|k1 |

1/ra 2 − 4k1 /ra

≃ ΨH0 (0, 0)π
32 8 − 2π
Ψ2 (0, 0)π 1/ra 2 − 4k1 /ra


≃ 2
π Hs
≃ Q2kk Hs2 (8 − 2π) 1/ra 2 − 4k1 /ra ,


(5.18)

where we have defined a two-dimensional spectral peakedness Qkk which is measured in meters,
RR 2
2 E (kx , ky )dkx dky 32Ψ2 (kx = 0, ky = 0)
Qkk = RR
2 = . (5.19)
E(kx , ky )dkx dky Hs4

This expression gives the approximate value for the standard deviation,
p
std(Hr , d1 ) ≃ Hs Qkk (4 − π) [2/ra2 − 8k1 /ra ]. (5.20)

This variability of Hra , which we have defined as the contribution of wave groups to the variability
of measured wave heights H b s is thus the product of three factors: the significant wave height Hs , the
shape of the wave spectrum as quantified by Qkk , and the effective range of spatial scales over which the
variance is integrated. That last factor is a function of the smoothing effect of the altimeter, represented
by the filtering scale ra , and the distance d1 = 2π/k1 over which we consider the variability.

Ĥs=9.3 m

std(Ĥs, 80 km)=0.49 m

std(Ĥs, 80 km) =0.88 m

a) b)

Figure 5.4: a) Map of wave heights in the North Atlantic at 09:00 on 14 February 2020, as provided
by the model hindcast of Alday et al. (2021), overlaid with circles located at the center of SWIM box
estimates for the L2-CWWIC wave spectra. Circles are sized by the L2-CWWIC Hs estimate and color
corresponds to std(Hs); b) corresponding measured Hs values as a function of latitude (y-axis) : black
small dots represent native measurements at 4.5 Hz, blue stars represent the 1 Hz averaged and grey
circles represent the Hs averaged over a box. Two boxes are selected for the case study: box A -
highlighted in light blue - is at 62◦ N, and box B - in dark blue - is at 44◦ N.

This work on the variability of wave height estimates was started after finding a much larger variability
on the south side of storm Dennis in the North Atlantic in 2020, as illustrated in Fig. 5.4. We used
CFOSAT because the SWIM instrument has a nadir (vertical) beam working like an altimeter, and
slanted beams that give a measure of the directional spectrum which we can use to compute Ψ, the
5.1. WAVE ENVELOPE, LOCAL AMPLITUDES AND THEIR STATISTICS 59

spectrum of the envelope. The maximum value of Hs reported by CFOSAT in the storm was estimated
at 17.9 m when using a the average over one measurement cycle (there are 4.5 such measurements per
second), or 17.9 m when averaged over 1 s. More interesting to us was, for the same mean wave height
of 9.3 m, a very small variability on the north side, and a twice larger variability on the south side.
Taking the measured wave spectra and simulating the surface and its envelope, as done in Fig. 5.5
makes it easier to visualize the importance of the spectral shape for the defining the spatial variability of
wave heights Hr . Both panels (c) and (d) correspond to spatially homogeneous sea states with a uniform
Hs . However, any measurement of waves will be sensitive to the ”lumps” of high values and low values
that are best seen in the envelope map of Fig. 5.5.f. The only way to remove this effect of wave groups
in the measurements would be to average in time as these groups propagate fast and randomly appear
and disappear.

Figure 5.5: From wave spectra (top) to surface elevation envelope (bottom): the surface elevations maps
in the middle line are simulated using the CFOSAT-derived spectra (these are L2S product corrected to
give the same wave height as the nadir beam), with random phases. Combining the real and imaginary
parts of the simulated surface gives the envelope. The left column is for a broad wave spectrum with
a lower peak period. The right column has the same wave height but a long peak period and narrower
spectrum.

We can use our theory to verify that in the case of narrow spectra, the expected effect of wave groups
gives a standard deviatin of the estimates of wave heights Hr that is similar to the standard deviation
60 CHAPTER 5. WAVE GROUPS AND FLUCTUATIONS OF WAVE PARAMETERS

of the measurement over a 80 km segment along the satellite track (Fig. 5.6). On the contrary, for

18

16

14

12
Hs [m]

10

8
Hs by box ± 2 std(Hs)
6 observations
estimations for WG:
4 Hs± 2 std(Hr,d1), eq. (4.13)
Hs± 2 std(Hr,d1), eq. (4.16)
2
6

4
:0

:0

:0

:0

:1

:1

:1

:1

:1
09

09

09

09

09

09

09

09

09
2020-02-14

Figure 5.6: Values of measured Hs , averaged over 80 km - black circles - , and corresponding std(Hs) -
black dash-dotted lines - in the satellite data, for the CFOSAT track shown in Fig. 5.4. Estimations of
std(Hr , d1 ) are also represented - in red and blue.

the broader wave spectrum, the effect of wave groups is much smaller than in the measurements: it is
possible that the data contains true gradients of the underlying Hs , and we also know that measurement
noise (known as speckle) is often the dominant source of variability in the measured values. This is one of
the main reason for the development of Delay-Doppler altimetry which has a much lower level of speckle
noise due to the averaging of different indepedent Doppler ”looks” for the same measurement.
In their paper De Carlo et al. (2023) went on to show that one can remove the predicted ”group-
induced noise” in the wave height measurements to produce maps of wave height ”noise” that now clearly
show that small scale variability in wave heights is clearly associated with regions of strong mesoscale
currents, as we will see in chapter 10.
Chapter 6

Wave data from satellites: Skylab

(1975) to SWOT (2025)

6.1 Radar remote sensing

Since the invention of radar, the sea was found to be an important source of echoes, at all radar frequencies
(from decametric to micro waves). This is due to the dielectric properties of sea water. An active radar
measures the electromagnetic power received by its antenna. Using the radar equation, this power (in
Watts) is normalized by the antenna-target distance, the antenna size, and the emitted power. This
gives the normalized radar cross section (NRCS), often represented by the symbol σ0 . σ0 depends on the
surface geometry but it alsy varies with the radar frequency and polarization, and observation direciton
(azimuth and incidence angles). The phase properties of the signal can alsl be used to estimate velocities
at the ocean surface: current and wave velocities.
Radar altimeter data are, the source of measurements of Hs available at global scale and used for
operational wave forecasting through data assimilation. Contrary to other observation systems, Hs is
directly determined, without the use of a spectral analysis. The usual arrangement is a radar antenna
looking straight down - at the nadir - on the the sea surface. The choice of the radar frequency is
dictated by a number of considerations, including atmospheric absorption – we want to be able to
measure a returning echo from the sea surface – and the size of the antenna – low frequencies give large
wavelength that require a proportionally large antenna to have a narrow beam. The radar wavelength
and frequency are related by the speed of light c, namely fr = c/λr .
Ocean waves have been monitored from space continuously since the lauch of ERS-1 using altimeters
and SARs, as summarized in figure 6.1

6.1.1 Conventional or ‘delay’ altimetry

The first main principle of the analysis of the radar echoes is the determination of the distance, usually
called range, based on the delay of a radar pulse to travel from the transmitting antenna to the target and
back to the receiving antenna. Usually the two antennas are the same piece of hardware and this is called
a ‘monostatic system’. Because the radar pulse has a finite duration which limits the resolution of the
time measurement, it is customary to use a varying transmitted carrier frequency ft that is modulated
as chirps: ft is increased linearly between f0 and f0 +B during a radar pulse. As a result, the carrier
frequency of the received signal fr can be used to determine the precise travel time of the received pulse,
with each frequency associated to a different range r, with a resolution in range dr that is determined by
the frequency bandwidth B, with dr = c/(2B). The actual signal at frequency fr combines echos from
all ranges convoluted by the point target response (e.g. Halimi, 2013).
for all types of radars, it is thus better to use a larger bandwidth, but this is usually limited by
atmospheric absorption windows or telecommunication regulations. Hence, for satellite altimeters, B for
the Jason altimeters is 320 MHz giving dr = 48 cm. SARAL/Altika used a wider bandwidth of 500 MHz
giving dr = 30 cm. Because of issues with rain attenuation, all altimeters since GEOS-3 (1975-1979)
have used Ku-band, except for the ongoing SARAL/AltiKa mission which uses only Ka-band.
Each radar pulse emitted by the radar antenna is reflected by an ocean area that expands with time,
starting from the blue disk in the middle of figure 6.2.c, and expanding to the outer rings. The shape of
the blue disk is only correct for a flat sea surface, and is distorted by waves as the crests give a shorter

61
62 CHAPTER 6. WAVE DATA FROM SATELLITES: SKYLAB (1975) TO SWOT (2025)

(A) 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 (B)
Hs (m)
GEOS-3
Geosat today
ERS-1 Coverage in 1 day:
Topex-Poseidon February 5, 2014
ERS-2
GFO Missions
Jason-1 completed
Envisat
Jason-2
HY-2A
Sentinel-1B
Cryosat-2
SARAL
Sentinel-1A (C)
Jason 3
Sentinel-3A
Gaofen-3 Missions
Sentinel-3B in operations
CFOSAT
HY-2B
H
HY-2C
Sentinel-6A
HY-2D
SWOT
Sentinel-1C Missions
Sentinel-1D approved
Sentinel-6B
Cristal-A

Figure 6.1: Satellite missions with sea state monitoring objectives.

Time coverage of satellite missions from 1985 to 2030, including nadir and near-nadir altimeters (solid
bars), and missions monitoring ocean wave spectra (open boxes) using C-band Synthetic Aperture Radars
(red), and real aperture radars in Ku-band (black) or Ka-band (blue). The lighter color (grey and
blue) bars correspond to altimeters using Delay-Doppler processing. Source: CEOS database http:
//database.eohandbook.com/). (B) example of 1-day coverage for Hs measurements with 4 satellite
altimeters (C) snapshot of a ’fireworks’ plot, the showing the height (size of symbols) peak periods
(colors), and directions (barbs) of swell partitions derived from Sentinel-1A and Sentinel-1B wave mode
data. Such plots are produced routinely by CMEMS.

range and the troughs give a higher range. the radar receives echoes from wave crests firsts and wave
troughs later: this difference in travel time between crests and troughs spreads the echoes over time.
Brown (1977) showed how the shape of the ‘waveform’, i.e. the received power return, under a number of
simplfiying assumptions, is a convolution of the radar antenna pattern and the distribution of the surface
elevation ζ. As a result, the slope of the ‘wave form leading edge’ is proportional to Hs . Because the
rise-time typically spans a few range gates, the wave height typically comes from a region of the ocean
that is between 4 and 7 km in diameter. In the case of the largest sea state in figure 6.2.a, the return
power spreads over about 35 range gates, from number 26 to number 61, i.e. a distance of 10.5 m, close
to the root mean square wave height Hrms ≈ Hs /1.4. The power in range gates beyond 50 comes from
the sea surface that is not directly under the satellite but on a circle around it, and is still illuminated
by the radar beam.
Classical altimeter estimates of Hs are fairly noisy when the wave height is low becase in that case
Hs is determined by only very few range gates, and the power in each rage gate as a random noise, also
called speckle noise, caused by Rayleigh fading of ocean echoes that are coming from ranges spread over
a distance much greater than the electromagnetic wavelength hence with random phases that randomy
cancel (Quartly et al., 2001). De Carlo et al. (2023) showed that wave groups introduce fluctuations in
the estimated values H b s that are proportional to Hs and the spectral peakedness parameter Qkk , these
fluctuations can be dominant for the largest wave heights.
Another interesting parameter that can be derived from the waveforms is the mean square slope
(mss). Indeed, the backscattered power σ0 is nearly inversely proportional to the mss (Vandemark et al.,
2002). For display purposes the waveforms shown in figure 6.2 have been scaled: you can see that the
noise level (power in gates 11 to 21) which should be nearly the same for all sea states is scaled to lower
values for lower sea states that generally have low values of both Hs and mss. Because the mss is not a
very common parameter in applications, many authors have derived empirical estimates of peak or mean
periods from Hs and σ0 . Also, the mss is a good proxy for the wind speed (Cox and Munk, 1954).
It should be noted that the main application of the altimeters was the mapping of the sea surface
height (SSH) for the determination of the geoid, tides and dynamical topography. The measured SSH
can be much more precise than dr, thanks to averaging. Waves play an important role in the estimation
of the SSH because of a range bias induced by wave non-linearities that induce a correlation between the
surface slope (and this its brightness for the radar) and the surface elevation. This is known as the sea
6.1. RADAR REMOTE SENSING 63

190 800 Hs = 3 m
180
160
(a) SARAL
140 600
(b) Sentinel 3
normalized power

120
100 400
80
Hs = 14.8 m Hs = 15 m
60
200
40
20 Hs = 0.8 m
0 0
1 11 21 31 41 51 61 71 81 91 101 111 121 128 0 10 20 30 40 50 60 70 80 90 100
Range gate Range gate
azimuth direction

(d) Sentinel 3

(c) SARAL

4 km 4 km

Figure 6.2: Altimeter waveforms and footprints

Example of altimeter waveforms for different wave heights. (a) These waveforms were selected along the
ascending track of SARAL/AltiKa on February 5, 2014, beteen 05:29:49 and 06:20:07 UTC. Each wave-
form shows the the power measured by the radar as a function of time: time is discretized with intervals
of 2 × 10−9 s corresponding to 30 cm range intervals usually called ‘range gates’. The corresponding
wave heights are 0.8, 3.2, 4.6, 9.7, 10.7, 13.2, and 14.8 m. Data is available from CNES/Aviso ftp. Each
waveform shown here was obtained for 1 s of data, as the median of 40 conscutive waveforms. (b) Similar
waveforms from Sentinel 3 using delay-only (dashed) and delay-doppler (solid) for Hs = 3,5,7,9,11,13 and
15 m, for cycle 23 orbit 349, on 25 October 2017 in the Pacific. (c) Spatial coverage of footprints corre-
sponding to the 3-dB antenna lobe pattern, for an unrealistic flat sea surface with a uniform roughness,
with the first 11 range gates painted alternatively blue and white, starting from center. (d) Footprint
for Sentinel 3, with the 7 seven range gates colored blue and white, with the azimuth resolution of the
Doppler processing indicated by the grey stripes.

state bias, and on average it is of the order of 3 % of Hs (e.g. Minster et al., 1991).
Also, because Hs and SSH are estimated jointly using a parametric fit to the waveform, there are
important correlated errors in the two parameters (e.g. Dibarboure et al., 2014; De Carlo et al., 2023). As
a result, without any particular editing or fitting of the waveforms, the variability of Hs at scales shorter
than 80 km is expected to be mostly due to the error in the fitting procedure that can be associated to
non-uniform radar backscatter and other effects. Recent methods have been developed to reduce that
noise and filter it (Passaro et al., 2015; Quilfen and Chapron, 2019).
Details will vary with different instruments, and here we will use the most simple case of radar
altimeters. It was shown by Chelton et al. (1989) that the sea echoes that contribute to the rising part
of the waveforms, as shown on Fig. 6.2, are within a radius rC of the nadir, the point on the sea surface
that is vertically below the satellite, with

s
2ho Hs + 2δR
rC = (6.1)
1 + ho /RE

where ho is the satellite orbit height above the ground, RE is the Earth radius and δR is the range
resolution of the altimeter.
64 CHAPTER 6. WAVE DATA FROM SATELLITES: SKYLAB (1975) TO SWOT (2025)

Number of valid
Mean value super-obs.
for 2011 in 2011

Figure 6.3: Global coverage of satellite altimeters

Over one day, the 3 satellites Jason 2, Cryosat 2 and SARAL/AltiKa covered all the oceans and seas
with a density high enough to capture all the important storms. In a year, the full ocean is covered at
a resolution of 1 degree in latitude and longitude. Data provided by ESA and CNES and processed by
Ifremer. The spatial cover depends on the orbits shape. The number of tracks per 1 degree x 1 degree
box (bottom panel) varies from 20 to 160 over a year: the coverage is less frequent close to the pole
because Jason 2 has a more oblique orbit that does not cover the latitudes beyond 66 degrees. Also, sea
ice produces echoes that differ from those of water and prevent the estimation of the wave height.

6.1.2 delay-Doppler altimetry

Another principle that can be used to refine the position and/or to measure the velocity of targets
along the satellite flight path is the Doppler effect: fixed targets ahead of the radar are moving towards
it, and hence have a higher frequency, while fixed targets that are behind along the satellite path are
moving away, giving a lower radar frequency. Typical low Earth orbits give a satellite velocity around
V = 7 km/s. Hence, a Ka-band system at 36 GHz frequency will see a Doppler shift of the order of
V fr /c = 840 kHz, which is further reduced by the incidence angle θr . For a 0.5◦ antenna aperture, θr
will normally vary between 0 and 0.5◦ , and the Doppler will be limited to 7.3 kHz.
The main benefit of Delay-Doppler altimetry is the increase of the number of ”independent looks”
of the sea surface, which helps reduce Speckle noise, but cannot do anything about the true geophysical
variability induced by wave groups. Delay-Doppler altimetry has also been developed to separate echoes
along the track, which is particularly useful for measuring sea ice freeboard at the edges of sea ice leads,
hence its first implementation on Cryosat-2.
A slicing of the radar echoes according to their Doppler shift allows a high-resolution mapping of the
surface along the satellite track. This principle is used on Cryosat-2, Sentinel-3, and Sentinel-6 (Mike
Freilich) with a standard processing giving 300 m resolution along the track, as shown on figure 6.2.d.
In that case the basic data has now one extra dimension, making a ‘stack’ of waveforms. For practical
purposes, this stack is converted to a single ‘SAR mode’ waveform, such as the solid lines shown in figure
6.2.b. At such a resolution, the surface elevation due to very long swells propagating along the track can
be resolved. A natural limit to this along-track resolving power is the fact that the sea surface is moving
up and down with the waves. Indeed, an orbital velocity of w = 2 m/s gives a ufr /c = Doppler shift
that can be mis-interpreted as a difference in incidence angle θr such that (ufr /c) = (V sin θr fr /c), this
corresponds to a horizontal displacement uHr /V ≃ 200 m for a radar altitude Hr = 800 km.

6.1.3 Synthetic Aperture Radars (SARs)

As explained, in section 6.1.1, the separation of echoes in the range direction is easily achieved by
combining the delay and frequency modulation of the chirped radar signal, with a resolution in distance
that is thus controled by the frequency bandwidth. In the other direction (azimuth), the simultaneous
echoes can be separated by their Doppler shift. This is perfect if targets are fixed relative to the ground,
and a very good resolution can be obtained, both in the range and azimuth directions (about 10 m for
Envisat, 5 m for Sentinel 1 in wave mode and 1 m in spotlight mode for TerraSAR-X). This processing
produces a map of the surface backscatter in delay-Doppler coordinates. Unfortunately these positions
are distorted from true geographic positions when the surface is moving toward the radar. With a
vertical velocity w, targets are displaced by δ = w cos θi R/V along the azimuth direction. With a
Sentinel-1 altitude Hr = 693 km and a typical speed over ground V = 7.5 km/s, when looking at an
6.1. RADAR REMOTE SENSING 65

incidence angle of 23◦ the distance R is close to1 Hr / cos θi and the factor Hr /V is about 92 s−1 , i.e.
a very small velocity w = 10 cm/s gives a displacement in azimuth of δ = 9.2 m. A high speed train
traveling at 100 m/s along tracks at 45◦ with the range direction at an azimuth angle of 23◦ has a radial
velocity of 28 m/s, would be displaced 2.7 km in the azimuth direction, at a location that is 2 km away
from the tracks!
(a) light source (c) radar power from target Radar modulation for T=10 s (f)

sea surface elevation gradient of velocity dw/dx

shallow pool low R/V

(e.g. TerraSAR X)
deep pool high R/V
(e.g. Envisat)
pattern at bottom SAR image

(b) picture by Gregory Massal (d) Pieces of Sentinel 1A Wave Mode image (e)

Figure 6.4: Patterns in SAR images

Analogy between (a,b) light patterns at the bottom of a shallow pool, and (c,d,e) velocity bunching
effects in SAR images of ocean waves. (d) and (e) are taken from a wave mode Sentinel 1A image,
acquired on 9 September 2014 at 04:48:16 UTC, at 10 and 35 km inside of the sea ice. In (d), almost
all crests are doubled, for example in the region within the dashed circle. In (e) the lines are less bright
and not doubled (for example within the dotted circle). This is easily simulated, as shown in (f), with
the variations of image intensity expected from a sinusoidal monochromatic wave of wavelength 156 m
(Adapted from Ardhuin et al., 2017a).

In the case of waves motion a velocity bunching effect appears, the echoes on the image are shifted
from their actual position depending on the surface velocity towards the satellite creating a pattern of
brighter areas, where the displaced targets are ‘bunched together’, and darker regions. This mechanism
is often the main cause of the wave-induced σ0 modulation in the open ocean. In ice-covered water, this
is probably the only mechanism present that creates wave patterns in SAR images. This bunching in the
azimuth direction is very similar to the light patterns at the bottom of a shallow pool that are caused
by light refraction, as illustrated in figure 6.4.
In the case of a monochromatic wave train propagating in the azimuth (y) direction, with wavenumber
ky , and not located right under the satellite but at an icidence angle θi , the target displacement in the
SAR image due to the velocity towards the satellite is

δ = (W cos(ky y − σt) cos θi + U sin(ky y − σt) sin θi ) Hr /(V cos θ). (6.2)

where U and W are the amplitudes of the horizontal and vertical velocities given by eqs. (2.28)-(2.29).
Assuming a uniform radar power scattered from the sea surface σ0 , and taking the y dimension
along the azimuth, the SAR image intensity is the incoherent sum at the displaced positions y ′ of the
power coming from the true pixel positions y, it is thus given by the inverse of Jacobian of the SAR
displacements y → y ′ = y + δ, (see eq. 21 in Hasselmann and Hasselmann, 1991),

J = |dy ′ /dy| , (6.3)

for a monochromatic wave of amplitude a it is,

′
ISAR (y) = σ0 /J = σ0 / |1 − CAR sin(ky y − σt)| . (6.4)
1 That approximation assumes a locally flat Earth. The exact value is obtained from the law of cosines in the triangle

made of the satellite position, the target position and the center of the Earth, i.e. R2 + 2RRe cos θi + Re2 − (Hr + Re )2 = 0,
where Re is the radius of the Earth.
66 CHAPTER 6. WAVE DATA FROM SATELLITES: SKYLAB (1975) TO SWOT (2025)

The important parameter for image patterns is the coefficient c in Alpers and Rufenach (1979),
CAR = ky (W + U tan θi )Hr /V. (6.5)
For CAR << 1, eq. (6.4) can be linearized, but as CAR increases, the SAR displacements become
′
strongly nonlinear and for CAR = 1, the Jacobian is zero and ISAR becomes infinite just like the light
intensity at the focal point of a lens. In our figure 1.f, with a wave period of 10 s traveling in the azimuth
direction, CAR = 1 corresponds to an amplitude of the elevation p a = 0.42 m, which, for random waves of
the same energy would be a significant wave height Hs = 4 (a2 /2) = 1.2 m. For CAR > 1, each bright
line becomes a doublet. The two lines of the doublet progressively drift apart as the amplitude increases,
pulling the minimum intensity to lower and lower values, up to the point where lines from different
doublets meet, at CAR ≃ 4.6. Beyond this value there is no region of very low intensity anymore.
In practice, except far inside the sea ice cover, there are also short wave components in the wave
spectrum E(k). From a SAR processing point of view, these short waves are equivalent to Gaussian
random vertical oscillations of < v 2 > leading to random displacements in the SAR image that are larger
than their wavelengths and that do not produce any pattern. These short waves also reduce the contrast
of longer components. Hasselmann and Hasselmann (1991) gave a theoretical derivation of the impact of
random waves on a SAR image spectrum ES (k), their eq. (55), with a simplified derivation by Krogstad
(1992). In practice the short wave effect is a reduction in the image spectrum by an exponential factor,
ES (k) ≃ exp(−ky2 < v 2 > Hr2 /V 2 )El (k) = exp(−ky2 λ2c /(2π)2 )|MS |2 (E(k) + E(−k)) /2, (6.6)
in which El (k) is a linearized spectrum, based on a modulation transfer function MS that includes a
linearized velocity bunching term, a hydrodynamic term due to the short scattering waves modified by
longer waves, and a tilt term due to the change in local slope along long waves (Hasselmann et al., 1985a;
Hasselmann and Hasselmann, 1991). All terms depend on the incidence angle, and these last two terms
depend on the polarization of radar waves, horizontal or vertical, and on wind speed and wave age. The
image is completely blurred at the scale of the random displacements and the resolution of 1 or 5 m is
useless. √
The azimuthal cut-off wavelength is λc = 2π < v 2 >Hr /V , in which < v 2 > is the orbital velocity
variance. This cut-off effect is so dominant that it can actually be used to measure the surface orbital
velocity variance from SAR images (Stopa et al., 2016c). A minor difficulty is the separation of the
part of the wave spectrum that produces patterns in the SAR image and the shorter part that only
introduces blurring. Looking at many ERS SAR data, Kerbaol (1997) concluded that, in the case of a
wind sea, the velocity variance < v 2 > should be restricted to waves shorter than a factor fL times the
peak wavelength, with fL ≃ 0.33 for a mean short wave direction in the range direction and fL ≃ 0.15
in the azimuth direction.
Combining all these effects with some empirically derived MTFs, it was possible to estimate the
heights of swells within 25% of buoy measurements using wave mode data from Envisat (Collard et al.,
2009). The full significant wave height, including the waves shorter than λc that are not resolved in the
SAR image, can also be estimated by combining all image parameters (Li et al., 2011).
Several aspects of SAR processing are the subject of active research, including the measurement of
high winds or currents, and improvements in the estimates of wave parameters in particular in ice-covered
regions.
Since SAR images are characterized by high resolution (5 m in the Sentinel-1 wave mode, 10 m in
Interferometric Wide swath mode), and large coverage, they provide a unique opportunity to measure
the spatial patterns in the wave field, as shown on figure 6.5.

6.1.4 SAR Across-track Interferometry: SWOT and near-nadir range bunch-

ing effects
Compared to the SAR systems on Envisat and Sentinel-1 described above, the KaRIN instrument on
SWOT has two very different features: first it measures at incidences very close to the vertical (nadir),
and second it actually has 2 SAR systems forming a cross-track interferometric baseline. This means
that in addition to the usual SAR imagery (a little unusual with KaRIN due to their incidence angles)
we also get interferograms from the two SAR receivers on KaRIN, which provide a wealth of information:
ˆ the mean phase of the interferogram is related to the distance between the radar and the reflecting
target, hence the sea surface elevation, which contains the geoid, the dynamic height, tides, internal
waves... but also those wind-generated waves that are longer than the SAR pixel, as shown on Fig.
6.6.
6.1. RADAR REMOTE SENSING 67

Figure 6.5: Left: Sample of a SAR image, recorded by Envisat on March 9 2003, over French coast.
The grey level is a function of the radar cross section that is modulated by waves. Right: Full image
processed into wave spectra, with significant wave height and peak direction. AWA et SA1 are positions
of in-situ instruments, used for validation (Collard et al., 2005).

ˆ the noise of the interferometric phase is related to the distribution of these distances within a
resolution pixel, hence the significant wave height of the waves that are not resolved by the image.

When trying to analyse SWOT data at small scales, it is thus really important to understand what is
resolved, what is not, and how the unresolved surface elevations give spurious signals in the resolved part:
indeed the imaging of the surface at near-nadir angles meas that all targets that have the same range
will fall in the same pixel, even though they are not at the same (x, y) position. This leads to a fairly
nonlinear image-blurring or image-enhancing effect called ”range-bunching”: just like velocity bunching,
which distorts or enhances the image in the azimuth direction, range bunching leads to distortions and
blurring in range. Figure 6.6 shows some example of wave signatures in SWOT surface elevation maps,
that still have to be analyzed quantitatively. You may look at the work by Peral et al. (2015) before a
more detailed summary appears in this section or a new chapter.

6.1.5 The wave spectrometer and the matching wavefront technique

Whereas a SAR resolves the wave patterns in an image to produce a wave spectrum, it is possible to
measure the 2D wave spectrum from a combination of 1D wave spectra.
When echoes are combined from narrow strip in range (red band in figure 6.7.a), the Fourier transform
of these echoes in the range direction selects only the modulations by waves that are perfectly aligned
with the direction θ0 to which the radar is looking, all other components cancel out. Hence, the Fourier
analysis of radar echoes provides a 1D spectrum E(k, θ0 ) in the look direction. A successive acquisition
in different directions provides the full directional spectrum E(k, θ). This is the principle of the wave
spectrometer and it has been demonstrated with the airborne instrument RAWS, developed by NASA
(Jackson et al., 1985), STORM and KuROS developed jointly by CNES and CNRS (Hauser et al., 1992;
Caudal et al., 2014). The first satellite wave spectrometer is SWIM, and it flies on the China-France
Ocean Satellite (CFOSAT, Figure 6.7), which was launched on October 29, 2018 (Hauser et al., 2021).
SWIM is unique in providing wave spectra co-located with classical altimeter measurements, allowing
a a much better understanding of historical altimeter data and their along-track variability (De Carlo
et al., 2023), as detailed in chapter 5.
It is also possible to analyze the Doppler of the backscattered signal, as demonstrated with KuROS
(Caudal et al., 2014), but this is best done with a narrower antenna pattern as used from space. This
additional measurement allows to remove the 180 degree ambiguity on the wave propagation direction,
but it also provides an indepedant measuremnt of the wave spectrum via the orbital velocities, and finally
it contains the signature of surface currents. This Doppler capability to measure currents was included
in the SKIM concept (Ardhuin et al., 2019) and demonstrated with airborne measurements (Marié et al.,
68 CHAPTER 6. WAVE DATA FROM SATELLITES: SKYLAB (1975) TO SWOT (2025)

(a) (b) (c) (d)

surface
elevation PSD
dB re 1 m4

40

20

(e) (f) (g) (h) 0

Figure 6.6: Following swells with SWOT, from storm to the antipodes
Top: small pieces (20 by 20 km) of SWOT data, bottom: corresponding surface elevation spectra
E(kx , ky ), these are double-sided spectra. Left: in the storm on June 6 2023, south of Kergulen is-
lands with strong range bunching effect. Next: large swells South of Australia, June 8. Next: refraction
and diffraction over Antigonia seamount, south-east of New Caledonia, June 11. Right: small amplitude
swells at ocean station Papa, Gulf of Alaska, on June 18, after 17,000 km of propagation.

2020).

(a) (b)

true 100 km
North
0 dR
range R

Figure 6.7: Measurement principle the SWIM radar that on CFOSAT. (a) geometry of the measurements
for one cycle in direction θ0 . The resolution in range dR is of the order of 10 m, but the azimuth resolution
is 18 km. Using different incidence angles θI (0, 2, 4, 6, 8 and 10 degrees) and rotating around the nadir
provides estimates E(k, θ0 ) in all direction θ0 . (b) Coverage of the SWIM instrument on CFOSAT: each
colored dot is the center of a 19 km diameter footprint.
Chapter 7

Measured wave evolution: main

parameters and wave spectra

A casual observation of the sea is enough to figure out that, the stronger the wind, the higher the height
H and length L of the waves. From one location to another, it is also obvious that L and H change with
the water body: a large lake can hold bigger waves than a pond, and waves in the Pacific, the largest
water body on Earth, can reach close to 40 m high, with wavelengths above 600 m. Finally, within the
same water body, the heights and lengths increase when moving downwind. The present chapter aims
at providing a quantitative summary of observed wave properties, including useful empirical prediction
formulae. These were the basis of wave forecasting methods (e.g. Pierson et al., 1955) before the advent
of numerical models (Gelci et al., 1957).

7.1 Wind-sea growth

A wide range of visual observations were systematically gathered by Sverdrup and Munk in 1941 in order
to come up with a wave forecasting method for the U.S. Navy, faced with the delicate task of landing on
the swell-battered beaches of Morocco. This work, initially classified, was only published as a landmark
report after the war (Sverdrup and Munk, 1947). The miscelaneous observations were organized thanks
to dimensional analysis with wave variables H and T expressed as a function of the wind speed U10 as
measured at 10 m above the sea surface, the fetch X, the duration t over which of the wind has blown
and the acceleration of gravity g. The fetch X is the length of the region over which the waves have been
generated.

Great 25
y Britain
20

Wind France 15
wind speed
Land (m/s)
200 km 10

x(fetch) 5

Spain 02/28/2010 00:00 UTC 0

Figure 7.1: Definition of fetch.

Left: idealized conditions, Right: a real case corresponding to the 2010 Xynthia Storm, which caused
severe coastal flooding in France: the wind speed varies rapidly in space and time, making it very difficult
to define a fetch for an equivalent constant wind. The Xynthia storm crossed the bay of Biscay in 6
hours going to the North-East, as a result the waves had little time to develop.

In the open ocean, X is not easy to define precisely, and it can be taken as an order of magnitude, in
relation with a representative wind speed U10 : the region where winds are very high is typically smaller
than the region where winds are lower, so that the choice of the wind speed will modify the fetch.
The use of U10 as the variable representing the strength of the wind is also fairly arbitrary. Many

69
70CHAPTER 7. MEASURED WAVE EVOLUTION: MAIN PARAMETERS AND WAVE SPECTRA

studies have debated the possibly better choice of the friction velocity in the air, u⋆a = (τa /ρa )1/2 . Since
wind speed measurements are usually converted to U10 and more often measured than u⋆a , it turns out
to be more practical to use U10 .
In the most simple case, represented in figure 7.1, a uniform wind speed blowing perpendicularly
offshore from a straight shoreline. In such conditions Sverdrup and Munk found that the dimensionless
wave energy
E ⋆ = Eg 2 /U10
4
(7.1)
and wave period
fp⋆ = fp U10 /g (7.2)
where E and fp are the free surface elevation variance and the frequency of the elevation spectral peak,
respectively, varies as a function of the dimensionless fetch

X ⋆ = Xg/U10
2
(7.3)

and duration
t⋆ = tg/U10 . (7.4)
⋆ 5
Figure (7.2) shows values measured at a time t > 10 , for which the sea state does not depend on
duration anymore. The figure also includes a comparison with the average measured values during other
experiments Kahma and Calkoen (1992); Kahma (1981).
Energy of the wind sea E*

-3
10

Total windsea (observed)

High frequency windsea
KC1992 unstable
KC1992 stable
JONSWAP
-4 Kahma 1981
10 Fully developped:
Alves et al. 2003
3 4
10 10
Fetch X*
0.5
0.4
Peak frequency fp *

0.3

0.2

0.1

Figure 7.2: Growth of waves with fetch, measured during the Shoaling Waves Experiment (SHOWEX) in
1999, off the North Carolina Outer Banks. The wind is blowing at an angle of 70 degrees relative to the
shoreline, which generates a double peak in the wind sea for the shortest fetches. Very close to shore, the
high frequency part of the wind sea (in red) is in the wind direction, while an ’alongshore wind sea’ (in
black) grows to lower frequencies, in the direction of the largest available fetch. For comparison, several
empirical growth curves are superimposed, as given by Kahma et Calkoen (1992), and Kahma (1981)
from other datasets, and the expected ’full development’ proposed by Pierson and Moskowitz (1964), as
re-analysed by Alves and Banner (2003).
7.1. WIND-SEA GROWTH 71

7.1.1 Full development and wave age

For large fetches, the wave energy and peak frequency appear to tend to an asymptotic limit. This
is very difficult to verify for moderate to strong winds, because the required non-dimensional fetch
X0⋆ = 2.2 × 104 is already 220 km for U10 = 10 m/s. In practice we have very few observations for steady
and uniform winds over a non-dimensional fetch larger than X ⋆ > 104 . In search of such conditions,
Pierson and Moskowitz (1964) have carefully selected 55 records obtained from weather ships. From
these measurements they estimated the asymptotic values
E ⋆ = 0.00402 (7.5)
and
fp⋆ = 0.123. (7.6)
The stage of development of waves, limited by fetch or duration, can be defined by the wave age,
A = Cp /U10 (7.7)
where Cp = 2πfp /kp is the phase speed at the peak of the spectrum. This parameter can be used to
separate the wind sea, which is generated by the local wind and corresponds to young waves, with ages
less than 1.2, and swell, for which the local wind has almost no effect, and which corresponds to older
waves, with ages larger than 1.2. Donelan et al. (1992) showed that the wave growth stops or at least
becomes very slow, when Cp /U10 > 1.2, confirming the analyses of Pierson and Moskowitz (1964). This
means that, for a fully developped sea state, the dominant waves are propagating 20% faster than the
wind speed.

7.1.2 Fetch limitation

The region where the wind is faster than a given value of U10 is always finite. This can limit the wave
energy and peak frequency to a value lower than what would happen in a larger region. A practical
important problem is the actual value reached by the wave height and peak period is such a limited area.
With the non-dimensional fetch at full development X0⋆ ≃ 2.2 × 104 , all measurements lead to empirical
wave growth formulae that are close to
  ⋆ 0.33
X
Cp /U10 ≃ 1.2 min ,1 , (7.8)
X0⋆
 ⋆ 0.5
U2

X
Hs ≃ 0.26 10 min ,1 . (7.9)
g X0⋆
We remember that in deep water,
2πCp
Tp = , (7.10)
g
which yields Tp from Cp .
The empirical growth laws (7.8)–(7.9) are a practical set of equations to obtain a quick estimate
of the order of magnitude of the sea state, with heights within a factor two of the measurements, as
shown in figure 7.2. Many variants of thes equations have been published, with numerical values of the
proportionality coefficients that may vary by a factor 2 (Kahma and Calkoen, 1992). These differences
are probably due, among other things, to the variations of the wind which is never exactly stationary
nor uniform, and not exactly perpendicular to the shoreline... which is itself never quite straight and
infinitely long. A careful analysis by Young (1998) reveals that some of the differences between the
different datasets may be due to differences between the air and sea temperature, which modify the
properties of the turbulence in the atmospheric boundary layer. The details of how that impacts the
wave growth is still not understood. One possible factor is that the wind tends to be less regular (more
gusty) in unstable conditions when the water is warmer than the air. It is well understood that this
increased gustiness can lead to higher waves for mature waves with X ⋆ > 104 , as shown by Abdalla and
Cavaleri (2002).

7.1.3 Time limitation

If the wind has started to blow only a short time ago, with t⋆ < 105 , the sea state parameters can be
estimated, replacing X ⋆ with
X ′ = (t⋆ /70)1.3 . (7.11)
72CHAPTER 7. MEASURED WAVE EVOLUTION: MAIN PARAMETERS AND WAVE SPECTRA

FD
40 10 40 10
9 9
8
20 8 20 7
Fetch - limited 7 Fetch - limited 6
5
10 6 10
time (hours)

time (hours)
4
5
4 3
5 5 2.5
3 2
2 2.5 2 1.5
2
1 1 1
1.5
time - limited time - limited
0.5 1 0.5
0.5

2 5 10 20 50 100 200 500 1000 2 5 10 20 50 100 200 500 1000

Fetch (km) Fetch (km)

Figure 7.3: Estimation of significant wave height (Hs , contours) as a function of fetch and duration for
the idealized case of an infinitely long coast with a wind blowing perpendicularly offshore at a speed
U10 = 20 m/s. In the left panel the estimate is given by a numerical integration of the wave action
equation, defined in the next chapter, using parametrizations for the wind-wave growth and dissipation
from Rascle and Ardhuin (2013) and non-linear wave-wave interactions from Hasselmann et al. (1985b)
using the WAVEWATCH III model, with a third order numerical scheme Tolman (1995). The right
panel combines eq. (7.9) and eq. (7.11). Both give the time-limited growth for large fetches and the
fetch-limited growth for large durations, with no more growth when waves reach full development at
Hs ≃ 10.8 m. The dashed line is the same in both panels. In the left panel, the model gives a slow
increase of the wave height even after the time limitation has been exceeded: this is because the infinitely
long shoreline allows the growth of very oblique components that travel very long distances alongshore.
That effect is not represented in the empirical formula of eq. (7.11). It is a well documented effect for
oblique fetches when the wind is not perpendicular to the coast (Ardhuin et al., 2007), but there is no
description of this phenomenon for winds perpendicular to shore.

7.1.4 Double limitation

In practice, the sea state is often limited by both time and fetch. The sea state parameters are then given
the lowest of the two values for Hs and Cp /U10 between the one obtained with X ⋆ and the one obtained
with X ′ (see CERC, 1977). For an alternative estimation, one may read Hwang and Wang (2004). This
double limitation is illustrated by figure 7.3.

7.2 Swell
So far we have only discussed the wind sea, which is related to the local wind. However, the sea state often
contains an important swell contribution, which are waves radiating away from their area of generation.
We can define swells as the waves for which the source of energy from the wind is zero or negative, in
practice it corresponds to waves travelling at a phase speed C faster than the wind speed U10 , or at angle
relative to the wind θw − θu such that U cos(θw − θu ) < C. This definition is often extended to allow
the peak of a fully developped wind sea, which can travel at a phase speed of 1.2 U10 , to be classified
as wind sea. This extension is equivalent to adding the source of energy from the nonlinear wave-wave
interactions to the source of energy from the wind in order to separate wind seas and swell. Because the
wind is neither steady nor uniform the boundary between wind sea and swell is rather fuzzy.
Swells is most important in large oceanic basins, in particular near their eastern boundary as the
dominant winds generally blow from west to east where it can account for more than 90% of the wave
energy (e.g. Chen et al., 2002). The swell is the part of the sea state that cannot be amplified by the
local wind because they are too fast, or travel at an angle too oblique to the local wind. This swell can
be composed of distinct individual swells, travelling in different directions and with different dominant
frequencies, these swells are the result of the evolution of wind seas that propagate out of their generation
area. Swells, when they exist, generally have longer periods than the wind sea because short waves are
rapidly dissipated away from their generation area. They are thus the result of storm conditions, with
winds strong enough to generate large period waves, as given by eq. 7.8. The stronger and larger the
7.2. SWELL 73

storm, the longer the swell period.

direction of propagation (degrees)

0
0.04 0.08 0.12 0.16 60 120 180 240 300 360

(a)
f (Hz)

21
11
13
15
17
19

23
25
27

31
1
3
5
7
9

29
Days (July 2004)

Figure 7.4: Example of the evolution of wave energy and mean direction as a function of frequency and
time, over one month, measured at the Christmas Island buoy (Kiritimati, Kiribati), in the middle of the
Pacific. Contour values equally spaced between 0.1 and 1.4 correspond to the logarithm of the spectral
density E(f ), colors correspond to the mean direction for each frequency. The oblique dashed line from
15 July at 0.05 Hz to 18 July at 0.8 Hz correspond to eq. (7.12) with a distance Rα = 6100 km between
storm and observation point. The intersection between this line and the axis f = 0 gives the date of the
storm: July 9, when a severe storm was indeed present south of New Zeland. (taken from Collard et al.,
2009).

Modern swell studies started during the colonial war between France and Morocco in 1906. In the
absence of safe harbors on the atlantic coast of Morocco, the disembarkment of troops and supplies
used small shuttle boats that could be destroyed by heavy swells. Such an event made the harbor of
Casablanca unavailable for several months. Swell forecasting for Morocco became a an important matter,
and the first method was based on the propagation of swells from the mid-atlantic Azores Islands where
visual observations were made several times a day (Gain, 1918). This method was used in the swell
forecasting office of Casablanca in the early 1920s (Montagne, 1922), where the first modern numerical
wave models were invented (Gelci et al., 1957).
In the Pacific or Indian oceans, it is very common to find at the same time and place several swells
coming from distinct storms (e.g. Figure 3.4), that may have happened 10,000 km away and a week before
(Darbyshire, 1957; Munk et al., 1963). Owing to the large distances travelled by swells, the sphericity of
the Earth must be taken into account. The re-derivation of linear wave theory in that case shows that
the swell that propagated along a straight line on a flat ocean, now travel on the shortest path on the
sphere, which are the great circles: the circles that have their centers at the center of the Earth, such as
the meridians.
The height and period of swells depend on the height and period of waves in the strom, and the
propagation distance outside of the storm. Storms generate waves with a wide range of periods up to
about 1.2 times the peak period Tp . This mixture of waves ‘disperses’, and because the group speed
Cg = gT /(4π) in deep water, is a function of the period T , the largest period swell arrive first, followed
by shorter swells. At very large distances, the storm from which the swell radiates can be considered a
point in space and time. The evolution of the swell peak period at a remote observation position follows,
4πRα
Tps (t) = , (7.12)
g(t − t0 )
where R is the Earth radius, t0 is the time of the storm, and α is the spherical distance, i.e. the angle –
in radians – at the center of the Earth between the storm and the observation point. This relationship
very well verified, as shown in figure 7.4.
74CHAPTER 7. MEASURED WAVE EVOLUTION: MAIN PARAMETERS AND WAVE SPECTRA

Swell heights decrease during propagation due to the dispersion from the source, and also due to
dissipation. Dispersion is the main effect for swell periods larger than 12 s, and corresponds to a stretching
of the wave fronts in the direction perpendicular to propagation, just like circles becoming larger away
from a stone dropped in a pond. There is also a stretching of the wave train in the propagation direction
due to the different group speeds, the longer wave periods contribute to a longer wave train away from
the storm, while the shorter wave periods are travelling slower behind. This spatial spreading of the swell
field is illustrated with numerical model results in figure 7.5, taken from Delpey et al. (2010). Neglecting

(a) (d)

(b) (e)

(c) (f)

12 13 14 15 16 17 18 19 20 21 22 23 24 25
Tps (s)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
Hss (m)

Figure 7.5: Swell radiated from a storm.

(a,b,c) peak periods of the swell Tps and (d,e,f) significant swell heights Hss from a realistic numerical
model of the storm of February 24, 2004, centered at (160◦ E, 42◦ N). The maps correspond to February
27th (top) March 1st (middle) and March 6th (bas) at 00h UTC, which is 3, 6 and 9 days after the storm
(reproduced from Delpey et al., 2010).

dissipation effects, and for distances of the order of 4000 km or more from the storm center, the decrease
in swell height following the peak in space and time is given by

r
α0 sin α0
Hss (α) = Hss (α0 ) . (7.13)
α sin α

In practice, the height is further reduced by islands and continents where it breaks and dissipates on
the shore. Dissipation in deep water is locally weak but can add up to a significant effect over long
propagation distances (Ardhuin et al., 2009a). Even with these effects, it is possible to record swells
coming from the antipodes (e.g. Munk et al., 1963).
7.3. FREQUENCY SPECTRA 75

7.3 Frequency spectra

7.3.1 The early days
The first measurements of wave spectra from time series were performed in 1944, as part of the efforts
of Group W at the British Admiralty, after the amphibious assault on Normandy (Barber et al., 1946;
Ursell, 1999). These records, and the many others that followed, revealed that for frequencies above the
spectral peak, the decrease in energy takes always the same form. Phillips (1958) introduced the idea
of an equilibrium region for f > fp , and proposed that gravity was the only determining factor for that
part of the spectrum which was controlled by wave breaking. Dimensional analysis leads to the shape
−4
E(σ) = αP g 2 σ −5 or E(f ) = αP (2π) g 2 f −5 , (7.14)

where αP ≈ 0.008 is now called Phillips’ constant. The non-dimensional energy E(σ)σ 5 /g 2 is constant
in this model: the surface is thus fractal without any particular scale. In other words, these waves are
self-similar and have all the same shape, whether they are large or small. These ideas were also developed

(a)
(b) 122
0
87
10 60
39
E(f) (m2/Hz)

25
E(f) (m2/ Hz)

6
-1
10 4

-2
10
0.1 0.2 0.3 0.4 0.5
f (Hz) f (Hz)

Figure 7.6: Evolution of the wave spectrum with fetch.

(a) measured spectra on September 15, 1968 at 11h during JONSWAP, with, inset, the proposed pa-
rameters for the JONSWAP spectrum. Numbers indicate the fetch in kilometers. (b) spectra measured
on November 3rd, 1999, averaged from 13h to 17h during SHOWEX, using Datawell Mark III waverider
buoys. Note that the scale is logarithmic. In the SHOWEX case, the wind speed is 10 m/s, 20 degrees
from the shore-normal. The peak around f = 0.1 Hz corresponds to swell arriving from the Atlantic,
against the wind. The instrument at the shortest fetch reveals two peaks in the wind sea, at 0.25 and 0.45
Hz. The analysis of wave directions, not shown here, show that the first peak is propagating alongshore,
and the second is in the wind direction. The existence of the first peak is due to the oblique 20◦ angle
between the wind and the shore-normal.

by Kitaigorodskii (1962), and led Pierson and Moskowitz (1964) to propose an empirical shape for the
full spectrum, based on fully developed sea states,
"  −4 #
2 −4 −5 5 f
E(f ) = EP M (f ) = αP g (2π) f exp − . (7.15)
4 fp

At short fetch, it was soon realized that spectral shapes could be significantly different. The peak is
particularly narrower for young sea states. Also, the values of the spectral density E(f ) for a given
frequency f can be larger than those found for old seas: this overshoot of the spectral peak was first
discussed by Barnett and Sutherland (1968), and further investigated during the 1968-1969 Joint North
Sea Wave Project (JONSWAP, see Hasselmann et al., 1973), figure 7.6).
Based on these observations, a ‘peak enhancement’ was added to the Pierson-Moskowitz (PM) shape,
" #
−(f −fp )2
exp
2σ 2 f2
A/B p
E(f ) = EP M (f )γ . (7.16)
76CHAPTER 7. MEASURED WAVE EVOLUTION: MAIN PARAMETERS AND WAVE SPECTRA

This may look a bit complex, but each parameter plays a fairly clear role, illustrated in figure 7.6.a.
γ ≃ 3.3 is the ‘peak enhancement-factor’, if equal to 1 then the spectrum is a PM spectrum, σA/B is
the relative width over which this enhancement applies. It was found that σA ≃ 0.07 if f < fp and
σB ≃ 0.09 otherwise. Using (7.16) and (7.8) one can estimate Cp and fp = g/(Cp 2π) as a function of
the wind speed. This gives the average spectral shape measured in the North Sea during JONSWAP.

7.3.2 The modern era

More recent obvservations, starting with Toba (1973), have shown that, at frequencies between fp and
3fp , the spectrum was not following f −5 but f −4 . This shape can be obtained by including the wind
speed or friction velocity in the dimensional analysis, it also corresponds to a constant flux of energy
towards high frequencies due to the non-linear wave-wave interactions. In passing, we can see that it
is always easy to find a theory for anything after the observations have been made. This f −4 decrease
near the spectral peak is particularly well verified by the wave gauge array data of Donelan et al. (1985),
which was one of the first clean measurements of the spectrum in both frequency and directions. Donelan
et al. (1985) also proposed a spectral shape that reconciles the peak enhancement of the JONSWAP data
with the mature wave spectral shape of Pierson and Moskowitz (1964). The adjustment given here in
eq. (7.8) is an adaptation by Elfouhaily et al. (1997), giving the proper asymptotes for wave energies
and periods. Further analysis by Long and Resio (2007) have shown that there is in general a transition
from f −4 at frequencies above the windsea peak to f −5 at higher frequencies, and the frequency where
this transition occurs appears to be a function of wave age.
This discussion of the detailed shape of the wave spectrum at frequencies beyond 2fp may sound
unimportant because it affects a very small fraction of the wave energy. However, that part of the
spectrum supports most of the energy flux from the wind to the waves, and thus defines the roughness
of the sea, and thus the growth rate due to the wind for the entire spectrum. Also, these short waves
dominate the surface slopes, which strongly affect ocean remotely sensed properties such as sea level,
surface salinity and wind speed and direction. Measurements of the sun reflection by the ocean performed
by Cox and Munk (1954) give a very good estimate of the mean square slope (mss),

(2πf )4
Z Z
2 2


mss = kx + ky E(kx , ky )dkx dky ≃ E(f )df, (7.17)
g2
where the second equality uses the linear wave approximation. In order to have a finite value of the
mss, E(f ) must decrease faster than f −5 towards the high frequencies. Elfouhaily et al. (1997) and
Kudryavtsev et al. (1999) have used this argument and detailed remote sensing data to propose a para-
metric shape of the wave spectrum that is a function of wind speed and wave age, and includes a strong
decrease for k > 10kp , taking also into account the effect of surface tension (see also Dulov and Kosnik,
2009; Yurovskaya et al., 2013). A few proposed parametric spectral shapes for gravity waves are com-
pared in figure 7.7. In practice there is a very strong variability of the observed specta around these
average shapes due to spatial and temporal variations in wind speed and direction. It should also be
noted that for f > 4fp there can be a dominant contribution of non-linearities, as revealed in stereo-video
imagery (see figure 3.8 in the preceding chapter).
7.4. DIRECTIONAL SPECTRA 77

1.5 JONSWAP 1973

10
0
DHH 1985
Elfouhaily et coll. 1997
(m2 /Hz)

KMC 1999
1 -2
10
f -4
E(f)

-4
0.5 10 JONSWAP 1973
DHH 1985
Elfouhaily et al. 1997
KMC 1999
-6
0 10 - 2
0.1 0.2 0.3 0.4 0.5 0.6 -1 0 1
10 10 10 10
f (Hz) f (Hz)

Figure 7.7: Example of proposed parametric wave spectrum shapes, for U10 = 10 m s−1 et X = 50 km.
Left: using a linear scale and right, the same spectra with a logarithmic scale. The slope of the curves in
log scale gives the power n of the dependance E(f ) ∝ f −n . Note the transition from n = 4 for f < 2fp
to n > 5 at higher frequencies, broadly consistent with the analysis of Long and Resio (2007). We have
extended here the Donelan et al. (1985) spectrum beyond its range of validity, namely for f > 3fp , in
order to better visualize its f −4 decay.

7.4 Directional spectra

7.4.1 Early parameterizations
The distribution of wave energy as a function of directions, is still very much debated because of the
difficulty of measuring details of that distribution. Indeed, a wave buoy only measured 5 parameters for
each direction (e.g. Kuik et al., 1988), and even ADCP systems are generally too noisy to go beyond the
mean direction and possibly spreading at each frequency (Herbers and Lentz, 2010). There are, however,
important constraints on the high frequency directional distribution, with a surface slope variance which
is almost the same in the downwind and crosswind directions.
The frequency-direction spectrum E(f, θ) is often decomposed into a frequency spectrum and a
normalized directional distribution
E(f, θ) = E(f )S(f, θ) (7.18)
with the normalization Z 2π
S(f, θ)dθ = 1. (7.19)
0
The first parameterizations of S(f, θ) used the fact that S is a periodic function and that buoys can
provide the first terms in its Fourier decomposition,
1
S(f, θ) = + a1 (f ) cos(θ) + b1 (f ) sin(θ) + ... (7.20)
2π
This was soon abandonned because S can be negative for some directions if the series is truncated.
Longuet-Higgins et al. (1963) proposed another distribution that is always positive
S(f, θ) = cos2s ((θ − θm )/2) , (7.21)
which is symmetric about the direction of the maximum θm , and narrower for larger s. This shape is
still widely used, although s cannot be readily measured, whereas the directional spreading σθ (defined
in chapter 3), is directly related to co-spectra of measured displacements or velocities. The width of
the directional spectrum is generally minimum at the spectral peak, and increases towards both higher
and lower frequencies. The mean direction, even for the wind sea, can differ significantly from the wind
direction, in particular at short fetch (see figure 7.9).
Among other proposed forms, the one by Donelan et al. (1985) is based on theoretical solutions for
non-linear wave groups,
1
S(f, θ) ∝ (7.22)
cosh2 [β (θ − θm )]
78CHAPTER 7. MEASURED WAVE EVOLUTION: MAIN PARAMETERS AND WAVE SPECTRA

and observations suggest β = 2.44(f /0.95fp )1.3 for 0.56 < f /fp < 0.95 and β = 2.44(f /0.95fp )−1.3 for
0.95 < f /fp < 1.6. That particular shape has been used for the estimation of the wind direction from HF
radar data because they have non-zero values in the direction opposite to the wind direction, consistent
with radar data. Unfortunately they still have a single maximum.

D(f, )

Win
d dir
ect
f/fp io n

Figure 7.8: Average spectral distribution measured in Currituck sound for U10 > 7 m s−1 (From Long
and Resio 2007).

7.4.2 Bimodality of the directional spectrum

Indeed, detailed observations using arrays (Young et al., 1995; Long and Resio, 2007) or surface mapping
systems with radar or optical imagery (Hwang et al., 2000; Romero and Melville, 2010; Leckler et al.,
2015) have clearly revealed that for f > fp the wind sea generally has two peaks on either side of the wind
direction, as shown in figure 3.8 and 7.8. Such a distribution is called bimodal. For frequencies f > 3fp
these peaks are 60 to 80 degrees away from the wind direction, at least for young waves (Cp /U10 < 1/3).
There is no simple parametric form for these bimodal spectra, and their general shape for more mature
waves is not established. Also, numerical models have a hard time reproducing clearly bimodal spectra,
as discussed by Alves and Banner (2003). In general, for the dominant waves, we only have a good
knowledge of the mean direction and directional spreading. One example of these two parameters is
shown in figure 7.9).
As a result, applications that require a detailed knowledge of the directional spectrum, such as the
interpretation of double-frequency acoustic and seismic noise, have to deal with large uncertainties. In
fact, underwater acoustic data is one important source of measurements that can be used to better
constrain our knowledge of the directional wave spectrum (Tyler et al., 1974). This source of data is now
better understood (Ardhuin et al., 2013) and is still being explored (Farrell and Munk, 2008; Duennebier
et al., 2012; Peureux and Ardhuin, 2016). This aspect is detailed in Part 3.
At high frequencies, remote sensing using HF (e.g. Kirincich, 2016) or microwave radars can also
constrain the directional distribution, and specific spectral shape parameterizations have been proposed
by, for example, (Elfouhaily et al., 1997) and (Kudryavtsev et al., 2003) to provide wave spectra consistent
with radar observations. With more data available now in new microwave bands such as L band (e.g.
Yueh et al., 2013), and detailed wave shape measurements from stereo-video and polarimetric systems,
there will certainly be improvements in the coming years.
7.5. SUMMARY 79

0
10

E(f) (m 2 /Hz)
-1
10

-2
10
0.1 0.2 0.3 0.4 WR(FRF)
frequency f (Hz) X1
180 Bravo
150 X2
120 X3
90 X4
qm (f) (deg)

60 X5
30 X6
0
-30
-60
-90
0.1 0.2 0.3 0.4 0.5
80

60
sq (f) (deg)

40

20
0

Figure 7.9: Example of measured directional parameters

Measurements during SHOWEX on November 3rd 2003, in presence of a 1 m swell opposing the wind
sea. The wind direction is from 270◦ and the shoreline faces 70◦ . The different wave systems are clearly
separated by a local maximum of the directional spreading σθ .

7.4.3 Swell spectra

The spectrum of a swell system is simply a wind sea spectrum that has dispersed and dissipated. Due
to the dispersion, it is much more narrow than a wind sea spectrum, and this narrowness increases with
the distance from the storm. This narrow peak may be parameterized by a Gaussian, with a frequency
bandwidth that is proportional to the size of the storm and inversely proportional to the spherical distance
α from the storm (Collard et al., 2009). The directional width is also increased by the storm size and
reduced by the distance, with one important difference due to the Earth sphericity. Indeed, beyond one
quarter of the Earth circumference, the directional width increases if no land has blocked the swell, and
is proportional to 1/ sin(α). Hence, the decrease in wave height given by eq. (7.13) corresponds to a
narrowing of the spectrum, not to a reduction of the spectral density. Indeed, in absence of disspation,
for deep water and without current, the spectral densities E(f, θ) are conserved during propagation.
Namely E(λ0 , ϕ0 , t0 , f, θ0 ) = E(λ′ , ϕ′ , t′ , f, θ′ ), where the point of coordinates (λ, ϕ, t) is on the same
great circle as the point (λ0 , ϕ0 , t0 ), and such that the azimuth of the great circle changes from θ0 to θ.

7.5 Summary
7.5.1 Important parameters
We have seen that the most important factor that control the wind sea are
ˆ the wind speed U10
80CHAPTER 7. MEASURED WAVE EVOLUTION: MAIN PARAMETERS AND WAVE SPECTRA

ˆ the fetch X

ˆ the duration t over which the wind has been blowing.

ˆ The depth D, which was not discussed here. The reader may follow Young (1999).

Other parameters also have a quantitative effect,

ˆ the shape of the fetch area

ˆ the air-sea temperature difference Ta −Tsea and the larger gustiness when this difference is negative.

ˆ strong currents C (if U/C > 0.1).

ˆ rain. That latter factor is not well known and is probably not so important in general.

The parameters of the first list have effects that are well understood. For the second list, the complex
situations usually require the use of a numerical wave model in order to get a reliable estimate of the sea
state parameter... but even in the models not all of these effects are well understood and thus not well
parameterized in the models. The next chapter will present the main concepts used in numerical models
for the evolution of waves in deep water.

7.5.2 Spectral shape

Wave spectra estimated from measurements have a wide variety of shapes, in particular close to coastlines
where the fetch limitation on wave growth gives mean direction that can vary with frequency. In the
open ocean, multiple swell systems that come from remote storms are also often present. (e.g. figure
3.4). The wind sea exhibits a clearly marked peak, where directional spreading has a local minimum,
and a decrease of the spectral density E(f ) proportional to f −4 up to 2 to 3 times the wind sea peak
frequency. For higher frequencies, the spectral density decreases like f −5 or faster.
In the directional distribution, there is always some energy in all directions, as revealed by high
frequency radar data (e.g. Barrick et al., 1974; Tyler et al., 1974) but it can be very small. The wave
spectrum generally has a marked bimodality at frequencies between 2 and at least 4 times the wind sea
peak frequency, with two maxima on either side of the wind direction. This spectral shape is the result
of different processes that, as we will see in the next chapter, can be represented in an equation for the
evolution of the spectrum.
Chapter 8

Physics of spectral wave evolution:

deep water

The present chapter will give some general explanations on the physical processes that explain the
observed sea state properties. It is now well understood that ocean waves derive most of their energy
from the wind, and lose most of it to the ocean turbulence as they break. One particular motivation for
understanding these physical processes is that beyond the wave spectrum, we may be able to quatify the
fluxes of energy and momentum in and out of the wave field. Indeed, the generation of waves by the
wind is thus associated to vertical fluxes of horizontal momentum and energy going from the wind to the
wave field. This flux of momentum generally accounts for more than 70% of the total momentum flux
going from the atmosphere to the ocean, this total flux is usually called the wind stress. That air-sea
momentum flux does not remain in the wave field, but rather it is lost by wave dissipation, mostly due
to wave breaking, and ends up in the ocean currents. Most of this momentum loss happens at the same
place where waves were generated: the wave field is thus largely ’transparent’ to this momentum flux.
However, a few percent of this momentum flux propagates away across ocean basins and is transformed
into changes in sea level – this is known as wave set up – and currents on the beaches where waves break.
These currents are the longshore currents. This transformation of wave momentum in the nearshore is
the topic of chapter 19.

wind direction
windward
slope (u2, w2) (u3, w3)
2
leeward air
p1
slope
p3
n1
n3
(u1, w1) water (u4, w4) 4

Figure 8.1: Fluxes at the air-sea interface, at four phases of the wave profile: 1 windward slope, 2 crest,
3 leeward slope, 4 trough. The energy flux from the atmosphere to the ocean is the work of the stresses
acting on the sea surface, [pn + τ ] · (u, w), with n the unit normal vector pointing to the water side.
∂ζ ∂ζ
As a result, if the air pressure is higher on the windward slope, then p1 w1 ∂x > p3 w3 ∂x and the
1 3
pressure-related flux is positive towards the water side. Likewise, if the shear stress τ is opposed to the
velocity vector (as here), the flux τ u induced by shear stresses is negative (i.e. energy goes from the
water to the air). Similarly, the momentum flux is the average stress acting on the surface. Because of
the surface slope, the vertical flux of horizontal momentum is the average of p∂ζ/∂x plus the average of
τx .

Besides the generation or attenuation by the wind and dissipation associated to breaking, a third

81
82 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

mechanism is very important for the evolution of the waves in deep water. It is the non-linear evolution
of the waves which can be understood as a wave-wave scattering process: waves components exange
energy and momentum, some components grow and others decay. In order to make things simple, we
will now consider separately each of these three processes.

8.1 Generation of waves by the wind

The fraction of the wave energy that is in the water is (ρw − ρa )/ρw ≃ 99.9%. Hence the transfer energy
from wind to waves involves a flux of energy through the sea surface. Because the surface is a material
surface, this flux cannot happen by advection, it thus requires the work of stresses on the surface, either
the tangential stresses τ or the normal stresses, i.e. the pressure p. In the case of tangential stresses,
the work is the correlation of the along-surface shear stress with the along-surface orbital velocity. The
work of normal stresses is the correlation of pressure times the velocity normal to the surface (see figure
8.1). A quantification of these energy fluxes thus require the identification of processes that produce
pressure and shear stresses on the surface. The first hypotheses on wave generation followed the theory
of hydrodynamic instabilities. The air flow over a water surface is a particular case of a sheared flow
in a stratified medium that may lead to the Kelvin-Helmholtz (KH) instability. This may indeed be
important for very high winds speeds (Soloviev et al., 2014), but it cannot explain the initial growth of
waves under moderate winds. In particular, KH theory predicts instabilities for an air-water interface for
wind speeds above 6.5 m s−1 (Jeffreys 1925), but ripples already appear on the water surface for winds
as low as U = 1.1 m s−1 (Kahma et Donelan 1988).
This initial growth of ripples is rather well explained by the effect of turbulence in the air advected
by the wind Phillips (1957). That process is very soon overtaken by the feedback caused by the wave-
induced pressure oscillations in the air, as the airflow is modified by the presence of waves. When waves
travel slower than the wind projected in their propagation direction, the pressure is slightly higher on
the windward face, typically of the order of a few Pascals, and lower on the leeward side.
We are thus going to generalize the Airy wave theory of chapter 2. To simplify the calculations we
will replace the wave amplitude a by a complex amplitude Z0 = aeiΘ0 , so that the sea surface elevation
is
ζ = ζ + R Z0 eik·x−σt ,
 
(8.1)
where we recall that R represents the real part of a complex number. In a first approximation, we will
assume that the atmospheric pressure is proportional but shifted in phase compared to the amplitude of
the surface elevation,
pa = pa + ρw gR (α + iβ) Z0 eik·x−σt ,
 
(8.2)
where α represents the in-phase oscillations and β the oscillations in quadrature.
Here we will take α = 0 for simplicity as we shall see that it is not important for wind-wave growth.
When β ≪ 1, we assume that the air pressure only makes a small O(β) correction to the solution we
found in chapter 2. Bernoulli’s eq. (2.13) at the surface now has a forcing term from the atmospheric
pressure. For propagating waves we can neglect γ(t), giving,

∂ϕ
≃ −gζ − pa /ρw − gR iβZ0 eik·x−σt .
 
(8.3)
∂t
Taking ∂(8.3 at z=ζ)/∂t +g×(2.14) we eliminate the unknown ζ.

∂2ϕ ∂ϕ
+ gσβR Z0 eik·x−σt = 0,
 
+g at z = ζ. (8.4)
∂t2 ∂z
Since the last term is already of order β, we can use the Airy theory result that is valid at order 0: eq.
(2.25) gives Z0 = iσΦ0 /g, and we recall that Laplace equation and the bottom boundary condition gives
us  
cosh (kz + kh) ik·x−σt
ϕ = R Φ0 e , (8.5)
cosh (kD)
.
We can thus re-write eq. (8.4) in terms of the surface complex amplitude Φ0 ,

−σ 2 + gk tanh(kD) + iβσ 2 = 0. (8.6)

8.1. GENERATION OF WAVES BY THE WIND 83

We thus have an order β correction to the dispersion relation, namely the frequency now has an imaginary
part. At first order in β, this is  
iβ p
σ = 1+ gk tanh(kD), (8.7)
2
and this gives an exponential growth if β > 0, or decay term if β < 0,
h √ √ i
ζ = ζ + R Z0 eβ gk tanh(kD)t/2 eik·x− gk tanh(kD)t . (8.8)

From
p a physical point of view it is more natural to keep our dispersion relation be redefining σ =
gk tanh(kD) and redefining the amplitude as slowly varying in time

a(t) = aeβσt/2 (8.9)

With these notations, the phase is the same as in chapter 2, Θ = k · x − σt + Θ0 , and the full solution
to first order in β is
a(t)
ζ −ζ = a(t) cos Θ − β sin Θ, (8.10)
2
ga(t) a(t)
ϕ = FCC sin Θ + βg FCC cos Θ, (8.11)
σ 2σ
a(t)
ξ3 = a(t)FSS cos Θ − β FSS sin Θ, (8.12)
2
a(t)
p − pH = ρw ga(t)FCC cos ψ − ρw gβ FCC sin Θ, (8.13)
2
da(t) βσa(t)
= , (8.14)
dt 2
d a2 (t) a2 (t) dE
= βσ giving = σβE (8.15)
dt 2 2 dt
where FCC = cosh(kz + kh)/ cosh(kD) and FSS = sinh(kz + kh)/ sinh(kD). For those alergic to complex
frequencies, the same solution is obtained by allowing the amplitudes to evolve slowly in time from the
beginning of the derivation, keeping track of the two time scales, slow for amplitude evolution and fast
for phase evolution. That method requires a careful book-keeping of the different terms that come from
the time derivative. For example,
∂ϕ a(t) a(t) a(t)
= −ga(t)FCC cos Θ + βg FCC sin Θ + βg FCC sin Θ − β 2 g FCC cos Θ, (8.16)
∂t 2 2 4
with the first two terms coming from the phase of eq. (8.12), and the amplitude evolution giving the last
two terms. We can see that at order β we neglect the last O(β 2 ) term and this is a solution to eq. (8.3).
It is an interesting exercise to also compute the Eulerian-mean flux of momentum at each depth,
⟨uw⟩, and prove that it is still zero at order β. The Lagrangian flux of momentum is given by the
mean pressure acting on the material surfaces, i.e. p∂ξ3 /∂x for the x-component of the momentum.
That Lagrangian flux has the same profile as the Stokes drift, given in the next chapter. This means
that when the wind generates waves, the (Lagrangian) wave momentum increases in the water column
(Ardhuin et al., 2008b).

8.1.1 Measuring and parameterizing wind-wave generation

With an air pressure that is proportional to the surface elevation, the energy growth in eq. (8.15) can
be written as a source term
Sin (f, θ) = σβE (f, θ) . (8.17)
The magnitude of the non-dimensional growth rate β is a key parameter to determine the energy balance
with dissipative and non-linear processes, and, as a result, the shape of the wave spectrum. Numerical
wave models all use semi-empirical parameterized expressions, with β a function of the relative direction
between the wind and the waves, and a function of the ratio of the wind speed and the phase speed.
In most model parameterizations, β is inspired from theoretical results (e.g. Miles, 1959; Fabrikant,
1976; Miles, 1996), with empirical adjustments to the few available measurements of pressure over waves,
or numerical simulations of air flow over waves, or observed wave groth (Plant, 1982), or measurements
of pressure-slope correlations (Snyder et al., 1981; Donelan et al., 2005, 2006).
84 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

8.1.2 Parameterizations based on observations

Snyder et al. (1981) performed an important field experiment in the bight of Abaco, in the Bahamas,
in order to reconcile the previous diverging observations by Dobson (1971) and Snyder and Cox (1966).
Their measurements performed under light to moderate wind speeds are summarized by a growth rate
 i
ρa h u⋆
β = max 0, 0.25 28 cos (θ⋆ − θ) − 1 . (8.18)
ρw C
1/2
p
where C is the phase speed ρa and ρw are the densities of air and water, and u⋆ = ⟨u′ w′ ⟩ = τa /ρa
where τa is the air-sea momentum flux per unit horizontal surface, usually called ‘wind stress’. This
parameterization is used in the ‘Cycle 3’ of the WAM model (WAMDI Group, 1988).
The measurements by Snyder et al. (1981) have been confirmed by other experiments, for example
in the North Sea by Hasselmann and Bösenberg (1991). Unfortunately these measurements at sea are
made only with moderate wave heights and wind speed, and only for waves around the spectral peak.
Besides, most wind measurements at sea consist of mean wind speed and direction at a fixed height,
typically 5 m, without the rapid fluctuations needed to estimate the friction velocity u⋆ . A link between
this wind speed and the wind stress is provided by assuming that the turbulent momentum flux ⟨u′ w′ ⟩
is constant with height (which is not so true near the surface in the presence of waves), and that the
mixing can be parametrized by an eddy viscosity of the form νT = l2 ∂U/∂z where the mixing length is
given by l = κz with von Kármán’s constant κ = 0.41. Under these assumptions, the wind speed profile
is a logarithmic as a function of height,
u⋆
U (z) = ln (z/z0 ) , (8.19)
κ
starting from 0 at the roughness height z0 . There are many discussions on the proper way to estimate
z0 , but a first reasonable guess is provided by the dimensional analysis of Charnock (1955),

z0 ≃ αCH u2⋆ /g, (8.20)

where αCH ≃ 0.015 is Charnock’s ’constant’.

From this type of expression, several adjustement have been proposed, in particular it appears that
the Charnock coefficient may vary with wave age, with an increase for young waves. This is how Janssen
(1991) parameterized the numerical results of a coupled wave-atmosphere boundary layer model.
For very young waves, with U10 /Cp > 3 there are also clear signs of strong air flow detachment from
the sea surface, which leads to an increase of β and decrease for extremely young waves when the air
flow is fully detached (Donelan et al., 2006; Babanin et al., 2007). This detachement of the air flow and
attenuation of the short waves by the wind (Soloviev et al., 2014) are possible explanations, together with
the effect of sea spray (Andreas, 2004), for the reduction of the drag coefficient Cd = u2⋆ /U10
2
in hurricane
conditions. This is a very active topic of research with important consequences for the understanding
and forecasting of extreme storm surges (e.g. Resio and Westerink, 2008).

8.1.3 Short waves and multiple-scale interactions

For the high frequency part of the spectrum, say f > 3fp , there are no direct measurements of the
pressure-slope correlations. The growth parameterizations are thus inferred from the observation of
wind-wave growth (e.g. Plant, 1982), and numerical simulations that are consistent with an energy
source term Sin proportional to u2⋆ cos2 (θ⋆ − θ). Besides, approaching the sea surface from above, a
growing fraction of the momentum flux is carried by the wave orbital motion, in the form of a correlation
uw. A likely consequence is that the short waves, which are affected by the air layers closest to the
surface, only feel a reduced wind stress as they are ’sheltered’ by the longer waves (Hara and Belcher,
2002). In the presence of long waves, we thus expect that the growth rate of shorter waves is strongly
reduced.

8.1.4 In summary
Wind is a source of energy (Sin > 0) for all spectral components such that U/(C cos θu − θ) > 1. In the
spectral plane, this corresponds to a region bounded by the straight line defined by U/(C cos θu − θ) = 1.
The other components, with U/(C cos θu − θ) < 1 are a sink of energy for the waves (Sout < 0), and thus
a source of energy for the wind.
8.2. WEAKLY NON-LINEAR EVOLUTION 85

80

70

60

b [(u /C) 2 r a /r w ] -1
50

40
2pb

$ 30

20

*
10

(a)
-10 (b)
-20
0 5 10 15 20 25 30
u* / C C/u
c/u,
*
of the wave growth rate from when IGQ =

Figure 8.2: Growth rate of waves propagating in the wind direction, combining observations and theory.
a. solid line: Miles theory as extended by Fabrikant and Janssen, dashed: measurements of Snyder
et al. (1981) as summarized by eq. (8.18), symbols: estimations compiled by Plant (1982), this figure is
taken from Janssen et al. (1994). (b) growth rate for a non-dimensional wavenumber kz0 = 10−4 (filled
triangles) compared to numerical model results with connected circles by Mastenbroek (1996) and the
Plant (1982) data (other symbols), that figure is taken from Belcher (1999).

The source term that represents the interaction between the atmosphere and the waves can be written
as
Satm (k, θ) = Sin (k, θ) + Sout (k, θ) = σβE(k, θ), (8.21)
where the non-dimensiona growth parameter β is of the order of 30ρa u⋆ /C/ρw for 1 < U/(C cos θu −θ) <
3, and a much smaller negative number, typically −2 × 10−6 for U/(C cos θu − θ) < 1. This negative part
appears to be non-linear with a magnitude of β that increases with the steepness of the swell.
In general it is expected that the growth rate of one component is also a function of the amplitude
of other components, but that effect is very poorly known.

8.2 Weakly non-linear evolution

Because waves are not exactly linear, they produce many interesting effects: presence of harmonics,
recurrence patterns, instabilities ... that are further discussed in Part 3. Here we will focus on the
evolution of the wave spectrum that comes from the exchange of energy between different spectral
components.
This spectral evolution is generally described as a wave-wave interaction process, which is a particular
case of wave scattering. In the case of a continuous spectrum, this effect is also known as ‘weak turbu-
lence’, as it exhibits Kolmogorov-type solutions with a cascade of energy towards the short waves as well
as an inverse cascade towards the longer waves. The rate of change of the spectrum due to this non-linear
effect was determined by Klaus Hasselmann (1960,1962) and Vladimir Zakharov (1968), starting from
the Euler equations and assuming a quasi-Gaussian sea state, and looking for the large time limit. These
two publications used different methods that are equivalent, as shown by Elfouhaily et al. (e.g. 2000)
and Resio et al. (e.g. 2001). The confirmation of this theory using more complete equations of motions
is fairly recent (Tanaka, 2001; Korotkevich et al., 2008). The experimental verification in the case of
a pair of monochromatic wave train was first performed by McGoldrick et al. (1966), who showed that
two wave trains of different frequencies and directions can create a third wave train with yet another
frequency and direction.
Hence, within the wave spectrum, there is a continuous exchange of energy between components
that strongly modifies the shape of the spectrum. This exchange is strongest for steep waves. The
only remaining doubts about this theory are its applicability to shallow water or in cases with strong
86 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

horizontal variations of depth or currents on the scale of the wavelength. In the latter case one cannot
take the large time limit and the interactions are generalized to include near-resonances (Stiassnie, 2004;
Annenkov and Shrira, 2006).

8.2.1 Wave-wave interation theory

When solving the Euler equations, we can keep the non-linear terms that were discarded by Airy, and
write the solution as an expansion in powers of the wave slope ε = ka. One then finds a first order
solution that is a superposition of Airy waves
X
ζ1 = Z(k, s)ei[k·x−sσt] (8.22)
k,s

and that can be introduced into the second order equations.

In order to make things more simple, let us imagine that the full boundary condition is given by,
 
∂
− gk tanh(kD) ζ = g∇ζ · ∇ζ. (8.23)
∂t2

At second order this gives the following forced harmonic oscillator equation
 
∂
− gk tanh(kD) ζ2 = g∇ζ1 · ∇ζ1 . (8.24)
∂t2

If there are components (k, σ) such that k = k1 + k2 et σ = σ1 + σ2 then the solution contain resonances.
Phillips (1960) showed that such components do not exist because the dispersion relation of gravity waves
does not have an inflexion point, hence there is no resonance at second order and the solution is bounded
with a simple expression of the second order amplitudes as a function of the first order amplitudes,
X X
ζ2 = A(k1 , k2 , s1 , s2 )Z(k1 , s1 )Z(k2 , s2 )ei[(k1 +k2 )·x−(s1 σ1 +s2 σ2 )t] (8.25)
k1 ,s1 k2 ,s

with A(k1 , k2 , s1 , s2 ) a constant coefficient given by

k1 · k2
A= . (8.26)
s1 σ1 s2 σ2 − gk tanh(kD)

The actual Euler equation (see Part 3) has a few extra terms but it is the same principle with a
similar result, only expression for A is more complex. This second order elevation is the generalization
of the first harmonic of a monochromatic waves: if the first order only contains waves of frequency f ,
the the second order has components with frequency 2f and 0.
Things get more exciting when resonances exist. Resonances at second order exist for dispersion
relation with an inflexion point, which is the case of gravity-capillarity waves or in the presence of sea
ice, but we will not discuss this here. For gravity waves Phillips (1960) showed that resonanced occur at
third order. The third order amplitude is a solution of a forced oscillator equation,
 
∂ X
− gk4 tanh(k 4 D) Z3 (k4 ) = B(k1 , k2 , k3 , s1 , s2 , s3 )Z(k1 , s1 )Z(k2 , s2 )Z(k3 , s3 )
∂t2
k1 ,k2 ,k3 ,s1 ,s2 ,s3

ei[(k1 +k2 +k3 )·x−(s1 σ1 +s2 σ2 +s2 σ3 )t] (8.27)

in which k3 = k4 −k1 −k2 . The right hand side is resonant for k4 = k1 +k2 +k3 and σ (k4 ) = σ1 +σ2 −σ3 .
This resonance condition is satisfied for an inifinite number of quadruplets (k1 , k2 , k3 , k4 ), as illustrated
in figure 8.3 for the case of deep water. Resonances also exist in shallow water, only the shape of the
curves are changed.
8.2. WEAKLY NON-LINEAR EVOLUTION 87

Figure 8.3: Geometric arrangement of wavenumbers that produce resonant interations in deep water,
and particular case of the quadruplet used in the Discrete Interaction Approximation (DIA), taken from
van Vledder (2006).

A particular case corresponds to k1 = k2 , as in figure 8.3. Any component k3 that lies on the thick
black curve shaphed like ∞, will interact resonantly with components k1 and k4 . This was verified in the
laboratory in the case where k3 and k1 are at right angles, with the creation of the new wave component
k4 (McGoldrick et al., 1966).
The fact that not all combinations of wavenumbers are resonant reduces the number of dimensions of
the interaction space from 6 to 3. In other words, the components with wavenumber vector k4 interact
with components k1 , k2 and k4 that follow some particular curves in spectral space. Assuming that
the surface elevation is Gaussian makes it possible to neglect the correlations of 4 different wave trains
(Hasselmann, 1962), giving a rate of change of the energy for time scales much larger that the wave
period,

∂E (k4 )
= Snl (k4 )
∂t Z
2
= |T (k1 , k2 , k3 , k4 )| δ (k1 + k2 − k3 − k4 ) δ (σ1 + σ2 − σ3 − σ4 ) (8.28)

× {E (k1 ) E (k2 ) [E (k3 ) + E (k4 )] − E (k3 ) E (k4 ) [E (k1 ) + E (k2 )]}

where δ is always zero except δ(0) = 1, and the coefficient T is an algebric expression similar to A in eq.
(8.26) but much more complex, and given by Herterich et Hasselmann (1980), or, with a simpler form,
by Zakharov (1999). The practical calculation of this coefficient and its integration requires a careful
handling of singularities (e.g. Gorman, 2003).
The positive term E (k1 ) E (k2 ) [E (k3 ) + E (k4 )] in eq. (8.28) comes from a third order amplitude
squared, and the negative term −E (k3 ) E (k4 ) [E (k1 ) + E (k2 )] is due to the correlations between first
and fifth order terms (see Part 3 for details). This negative part is critical to conserve the total wave
energy. In general the coefficient T decreases as the difference between wavenumbers is larger. In other
words, the interaction of very different wavelengths or directions is much weaker than the interaction of
components that are very similar. A rough estimate of the time scale of spectral evolution is ε4 Tp .
Faster evolutions due to non-linear effects exist that are not represented by Snl . These faster changes
are oscillations of the energies that do not contribute directly to the long term evolution of the wave
spectrum.
88 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

8.2.2 Conservation properties

The source term Snl conserves the total wave energy, namely,
Z
Etot = ρw g E (k) dk (8.29)

as well as the wave momentum vector, which will be discussed further in the next chapter,
Z
E (k)
Mw = ρw g k dk, (8.30)
σ

and the wave action,

Z
E (k)
A= ρw g dk. (8.31)
σ
As we will see in chapter 10, the conservation of wave action in general is related to the invariance of the
phase-averaged physical situation by a change of the wave phases. In particular, wave action - and not
wave energy - is conserved when waves propagate across current gradients without energy dissipation.
The conservation of the two scalar quantities A and E in a non-linear system imposes the presence of
transfers of energy, also known as ’cascades’, towards both short and long wave components (Zakharov
and Zaslavskii, 1982). Indeed, once integrated over directions, Snl (k) is the rate of change of the energy
for a given wavelength. Assuming that there is only a transfer of energy from components with k < kt
to shorter waves with k > kt , we would have both
Z kt Z ∞
Snl (k)dk = − Snl (k)dk (8.32)
0 kt

and
Z kt Z ∞
Snl (k)/σdk = − Snl (k)/σdk, (8.33)
0 kt

which is not possible since the division by σ makes the first integral relatively bigger than the second.
Hence there is a flux of energy towards both long and short waves, which explains part of the increase
in wavelength as the waves develop with time or fetch. Also, the conservation of momentum Mw ,
further imposes that the energy transferred towards high frequency cannot be in the same direction, but
rather in oblique or opposed directions. This property is tighly linked to the resonant conditions that
is a consequence of the dispersion relation. Investigating the evolution of a narrow-peaked spectrum,
Longuet-Higgins (1976)√ showed that the energy tends to flow away from the peak towards high frequencies
at directions arctan(1/ 2) rad ≃ 35◦ (figure 8.4). Under the effect of nonlinear interactions alone, a
narrow spectrum thus tends to have a split tail with two peaks separated by about 70◦ , which agrees
with the observations of wind sea spectra at frequencies 1.5fp < f < 2fp (Hwang et al., 2000; Long and
Resio, 2007), as shown in figure 7.8.
Without forcing from the wind and assuming that the dissipation occurs only at the smallest scales, the
wave spectrum tends to evolve towards a self-similar shape known as the Kolmogorov-Zakarov spectrum,
with a decay towards high frequencies that is proportional to f −4 . In reality, the presence of forcing and
dissipation at the dominant scales makes the wave spectra different from this self-similar solutions.

8.2.3 Other properties of wave-wave interactions

Besides the fluxes towards both ends of the spectral domain, the non-linear interactions also have a
very strong local smoothing effect. If one takes a spectrum near equilibrium and introduces a local
perturbation, the 4-wave interactions will very rapidly remove this disturbance giving a smooth spectrum.
This smoothing is faster for steeper waves. For a wind sea, this can happen in less than 100 periods,
(figure 8.5), which is much faster than the theoretical dimensional argument saying that Snl acts on a
time scale of ε6 Tp . In the case of swells, with a much weaker steepness, it is possible that the interactions
produce a significant broadening of the directional spectrum, without a noticeable shift in frequency
(T.H.C. Herbers, personal communication, 2001).
8.2. WEAKLY NON-LINEAR EVOLUTION 89

Figure 8.4: Non-linear interactions and bimodal spectral distributions.

The interaction of wave components with wavenumber vectors k1 , k2 , k3 and k4 , when k3 ≃ k4 .
Since k1 + k2 = k3 + k4 , the 4 wave vectors make a parallelogram. taking k3 and k4 symmet-
ric about q
the spectral peak kp and taking the x-axis in the direction of kp , one gets kp,y = 0
and k1 = 2 + ∆k 2 , the latter equation also applies to the three other wave
kp2 + 2kp ∆k1,x + ∆k1,x 1,y
√
vectors.
p Expanding for small values

of X1 = ∆k1,x /kp and Y1 = ∆k1,y /kp gives σ1 = gk1 ≃
gkp 1 + 0.5X1 − 0.125X12 + 0.25Y12 ,. The resonant conditions read X1 + X2 = 0 Y1 + Y2 = 0 and,
2 2
√ becomes 2Y1 − X1 = 8α. That is the equation of an hyperbola
for the frequencies σ1 + σ2 = 2σp (1 + α)
with asympotes at angles ± arctan(1/ 2) relative to the x-axis (figure adapted from Longuet-Higgins,
1976) .

8.2.4 Practical calculation of wave-wave interactions

The calculation of the full integral Snl is unfortunately a little too expensive today for operational wave
forecasting, except for large scale coarse models, due to the three-dimensional integral needed for each
spectral component and at each grid point where the source terms are evaluated. Such calculations
are thus confined to research applications. Operational wave models use some form of approximation,
and the Discrete Interaction Approximation by Hasselmann et al. (1985a) is the most common. That
approximation preserves the conservation properties of the full integral by using only a subset of the
resonant quadruplets k1 , k2 , k3 , et k4 . The resonance conditions k4 = k1 + k2 − k3 and σ (k4 ) =
σ1 +σ2 −σ3 mean that for a given k1 +k2 , k3 must be on the curve where k1 lies: each curve corresponds to
a fixed value of σ1 +σ2 . The DIA parameterization uses a single configuration k1 = k2 = k, k3 = (1+λ)2 k
2
√ imposes k4 = (1 − λ) k. This DIA, as adjusted by Hasselmann et al. (1985b),
and σ1 + σ2 = γσ, which
uses λ = 0.25 et γ = 2. The angle of the ∞-shaped curve near the origin with the x-axis, when k3 ≃ k1
is the same 35◦ angle as in figure 8.4. Each curve corresponds to a different value of γ. The original form
of the DIA is only valid in deep water.
The shape of Snl produced by the DIA can differ significantly from the exact solution. It has been
adjusted to give the right order of magnitude for the transfer of energy to low frequencies in a wind
sea, which is very important for the wave development. That choice, has some side-effects on other
unconstrained parameters such as the directional wave distribution which is too broad for 1.5 < fp <
f < 4fp .
Several intermediate methods have been developed with some already used in operational wave fore-
casting (e.g. Komatsu and Masuda, 1996).
90 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

Figure 8.5: Illustration of the smoothing effect of non-linear interactions.

A spectrum with a gap, evolves rapidly towards a smooth spectrum. Top-left: initial spectrum, right:
spectrum after 3 minutes, which is less than 100 dominant periods. Bottom: source terms for wind
generation Sin , wave-wave interactions Snl , and dissipation Sds . Figure taken from Young and van
Vledder (1993) .

8.3 Dissipation
Many processes contribute to the dissipation of wave energy, or more exactly, the conversion of mechanical
wave energy into other forms of energy, in particular turbulence in the water and air.
Wave breaking is generally the most important sink of energy for wind seas. In the case of periodic
waves, breaking results from an instability that develops near the wave crest when the orbital velocity
appraoches the phase speed1 . This criterion gives a maximum possible wave steepness that is

Hmax /L ≃ 0.14 tanh(kD) (8.34)

as first determined by Miche (1944a). The factor tanh(kD) happens to be also the ratio between the
amplitude of elevations a and the amplitude of the orbital velocities at the surface in the case of linear
waves. It is likely that a similar criterion applies to random waves. Indeed, water moving forward faster
than the crest will shortly find itself over air and ready to overturn. In deep water, the Miche criterion
(8.34), gives the Stokes limit, kamax = πHmax /L = 0.44.
Laboratory observations by Melville and Rapp (1988) and Stansell and MacFarlane (2002) show that
non-stationary breaking waves have orbital velocities u that approach the phase speed. One of the
important difficulties is to relate this orbital velocity to the wave spectrum, because, for real and thus
nonlinear waves, u increase much faster than a when approaching the breaking limit. Also, breaking is
defined for individual waves, which are not easily related to the spectrum. Many authors have sought
criteria for wave breaking based on the vertical acceleration, but this is not a good indicator of breaking
(e.g. Holthuijsen and Herbers 1986).

1 In the case of a stationnary wave, instability occurs when the vertical acceleration approaches g.
8.3. DISSIPATION 91

(a)

WIND

(b)
50 cm

(c)

Figure 8.6: Features of breaking waves from short to longer waves.

(a) Schematic of a micro-breaking waves viewed sideways, the vertical scale of the boundary layer is
exaggerated (from Siddiqui and Loewen, 2007, ©Cambridge University Press), (b) breaking of a short
wave with entrainment of air at the crest (from Koga, 1982), (c) evolution of a breaking wave in the case
of a plunging breaker with one wave profile every 0.04 s (from Bonmarin, 1989, ©Cambridge University
Press) .

8.3.1 A classification of breaking waves

Breaking waves of different sizes look different. The shortest gravity waves, with wavelengths 0.1 <
L < 1 m (i.e. f > 1.25 Hz) do not make any bubbles. This is because an increase in the area of the
air-water interface requires an energy that is the surface tension T times the excess surface, hence the
short waves do not have enough energy to make bubbles when the break. Instead, short waves produce
micro-breakers (Banner and Phillips, 1974) that are characterized by a strong curvature of the surface
and the generation of capillary ripples on the forward face ahead of the breaking point. These capillary
waves are strongly damped by viscosity that absorbs a large part of the energy lost during breaking.
Observations reveal that short waves break very often, with a probability that increases from 11% for a
wind speed of 4.5 m/s to 80% for 7.4m s−1 (Siddiqui and Loewen, 2007). These micro-breakers play an
important role in setting the surface temperature and the gas fluxes at the air-sea interface because they
disrupt the viscous surface layer (figure 8.6.a).
Longer waves are energetic enough to produce bubbles with a particular noise. This ambiant noise is
an important factor in the performance of sonar systems (e.g. Lu et al., 1990; Ma et al., 2005). Injecting
the bubbles at depths also requires a conversion of kinetic energy into potential energy, that can take
up as much as half of the energy lost in the whole breaking process (Lamarre and Melville, 1991). It
should be noted that long waves can contribute to the breaking of short waves, either because the long
waves are breaking or because they produce a straining of the short waves that locally increases the short
wave steepness. Wave breaking of all scales produce vorticity in the water column which is important
for upper ocean mixing, in particular for the diurnal cycle of sea surface temperature.
Several types of breakers are usually defined. When the breaking wave is quasi-stationary with a
gentle forward slope, it is a spilling breaker. A steeper waves that trap a tube of water near the crest, is
a plunging breaker (figure 8.6.c). Finally a waves that rapidly collapses on a steep shoreline is surging
breaker.
92 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

8.3.2 Parameterizations of dissipation due to breaking in shallow water

Dissipation of the total wave energy
Near the shore or on offshore shoals, the water is shallow enough that the dominant waves are not
dispersive. In that case the question of wave breaking is greatly simplified by considering the total
energy Et instead of the spectrum. Since waves near the shore often have a well-defined direction we
may consider that the wave energy is radiated in a single direction, that of the spectral peak. By analogy
with a hydraulic jump, as detailed in Part 3, the rate of energy dissipation per unit surface is
3
1 1 (BH)
ϵ(H, D, T ) ≃ ϵ1 (D − H/2, D + H/2) ≈ ρg (8.35)
L 4 DT
where B is a tuning factor that is close to 1... empirical adjustments are often useful to produce accurate
simulations, and this is one of the better constrained coefficients. Obviously, as B comes to the third
power, a small change in B can be a significant change in the modeled dissipation rate.

Effect of wave height distributions

Since the wave height H is very important for the dissipation rate, we should determine what is the
period and height of the waves that actually break. Experimental data shows that the breaking waves
have a probability distribution pB (H, T ) that is different from the distribution p(H, T ) of all – breaking
or not – waves. We can define a weighting function W that gives the breaking wave distribution from
the full wave distribution,
pB (H, T ) = p(H, T )W (H, T ). (8.36)
The dissipation rate per unit time and per unit horizontal surface is thus an average over all heights
and should equal the sum of all spectral components,
Z Z
ϵtot = ϵ(H, D, T )W (H, T )p (H, T ) dH = ρg Sdis (k) dk. (8.37)
k

In deep water, neither the orbital velocity nor the pressure gradient are uniform over the vertical.
The energy flux is thus clearly different from the shallow water value. Hence the depth D that appears
in the dissipation rate ϵ should be – at the very least – replaced by a relevant length scale e
h. Chawla
and Kirby (2002) proposed to use e h = tanh(kD)/k which goes to D when kD goes to zero. In order to
reproduce the wave evolution in both deep and shallow water, Filipot et al. (2010b) have redefined B as
a function of kD with B = 0.185/ tanh[(kD)1.5 ]. This leads to
s
1 3 gk
ϵ(H, D, T ) = ρgk (BH) (8.38)
8π tanh(kD)

The choice of e
h can be debated and should be determined from an energy balance such as eq. (8.35).

Breaking wave statistics in shallow water

Finally, the dissipation rate requires a definition of W (H, T ) and p (H, T ). In shallow water, the energy
is usually integrated across frequencies, and in that case we only need W (H) and p (H). The latter is
often taken to be the Rayleigh distribution pR (figure 1.4). For W , two different choices have been made.
The first choice by Battjes and Janssen (1978) is based on the idea that all breaking waves have the same
height corresponding to the depth limitation H = γD in which γ is a known constant, and the probability
of occurence of these waves is simply given by the Rayleigh distribution, namely QB = pR (h > γD).
The resulting functions p(H) and W (H) thus have a singularity at H = γD. On the problem of this
first approach is that it can lead to unphysical parameters such as QB > 1, which is not very nice for a
probability (Janssen and Battjes, 2006).
The second choice is an empirical determination of W from observations. After days of counting
waves passing by a fixed location in the surf zone, Thornton and Guza (1983) have proposed the empirical
expression
 4
Hrms
W (H) = (8.39)
γD
8.3. DISSIPATION 93

where γ plays a role similar to the γ in Battjes and Janssen (1978). One reason why they chose this
particular form, is that it allows an analytic integration of eq. (8.37) that gives,

3 1/2 B3 7
ϵtot = π ρg 4 5 fp Hrms . (8.40)
16 γ D
in which fp is the peak frequency.

8.3.3 Breaking wave statistics and dissipation rates in deep water

Many applications, including underwater acoustics, remote sensing and the investigation of air-sea fluxes,
rely on some characterization of breaking waves. The investigation of breaking waves has been limited
for a long time ot the estimation of the whitecap coverage in which the active part and the residual
foam was separated (e.g. Monahan and Woolf, 1989). This coverage was found to be strongly related
to the wind speed, with an increase from very low values for pour U10 < 7 m s−1 , to about 1 % for
U10 ≃ 10 m s−1 , 6 % for U10 ≃ 20 m s−1 , and much more for yet higher wind speeds. In fact, these
variations in foam coverage and aspect are the basis of the Beaufort scale (see table 1.2) that is still used
to determine the wind speed from a visual inspection of the sea.
Wave breaking can also occur without any wind, due to the convergence of wave energy associated
to current gradients or the bottom topography. Most authors have distinguished a ’depth-induced’ wave
breaking that occurs near the shore in the ’surf zone’, from the ’whitecapping’ that occurs in deep water,
usually in the presence of wind. This distinction can be fuzzy in the case of shallow tidal flats, which is
why we preferred to have a single breaking definition (Filipot and Ardhuin, 2012).
The fraction of sea surface covered by foam or the breaking probability used in shallow water is not
sufficient to fully characterize wave breaking. Hence Phillips (1985) proposed a spectral description of
breaking and introduced the breaking spectrum Λ, which is the density of breaking crest length per unit
surface and per unit vector speed c of the breaking fronts. With this definition, the length Λ(c)dcx dcy
is the total length of all breaking fronts per unit sea surface that move at a speed c = (cx , cy ) within a
speed interval dcx and dcy . In general c is about 0.8 times the phase speed C of the dominant waves
that are breaking, due to modulation effects (Banner et al., 2014). Such a decomposion of the breaking
waves was motivated by the observation by Duncan (1981, 1983) in the laboratory that the dissipation
is proportional to c5 .
Other interestingRquantities derived from Λ include the fraction of the sea surface wiped by breaking
fronts per unit time cΛdcx dcy . Assuming that the breaking waves have self-similar shapes, Λ can also
be used to provide a an estimate of the mean foam thickness (Reul and Chapron, 2003).
The investigation of breaking wave properties has thus focused onthe estimation of Λ from observations
or models. This is not an easy task, requiring high resolution video data in both optical and infrared
as illustrated in figure 8.7 (e.g. Sutherland and Melville, 2013). More classical measurements of surface
elevation time series have also been associated to a visual or acoustic indentification of breaking waves
(Banner et al., 2000; Babanin et al., 2001; Manasseh et al., 2006). These studies have led to the conclusion
that the breaking of dominant waves is not associated to a fixed threshold, but that statistically, the
probability of breaking waves can be predicted from the energy level of dominant waves. There is a fixed
threshold, when the energy is put in non-dimensional form, over which breaking occurs. There is thus a
link between the breaking probability and the ’saturation’ spectrum B(k) introduced by Phillips (1985)
and defined by
Z θ+∆θ Z k+∆k
B (k, θ) = cosp (θ − θ′ )k 2 E(k, θ′ )dkdθ′ . (8.41)
θ−∆θ k−∆k

As pointed out by Phillips (1984), using a local definition of B with ∆θ and ∆k very small makes
sense only if the spectrum is relatively smooth. Indeed, for a narrow spectrum, B could be very large with
very small amplitude waves, even going to infinity for monochromatic waves. In their analysis Banner
et al. (2000) used ∆θ = π, p = 0 and ∆k ≃ 0.6k. In fact, it is difficult to investigate breaking statistics
for small values of ∆θ, due to the greater statistical uncertainty. In practice the breaking probability
is associated to the presence of steep waves, and these can exist only if wave trains with neighboring
wavenumbers and directions can interact to form wave groups that live long enough to let the waves
evolve towards breaking. The basic interaction is first a linear superposition, and the eventual evolution
up to breaking is obviously nonlinear (e.g. Song and Banner, 2002).
Besides, the fact that B has no dimensions suggests that breaking mainly depends on the shape of
the waves, while other factors (wind, current ...) are only secondary. After a first demonstration for
94 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

(a) (b) (c)

(d)

160 m

(e)

Figure 8.7: Breaking fronts

(a) Detection of breaking front, in black, using infrared imagery of micro-breakers (picture from Jessup
and Phadnis 2005) , (b), (c), (d) foam coverage for wind speeds of 7, 10 and 14 m s−1 (Melville and
Matusov 2002), and (e) schematic defining Λ (from Reul and Chapron 2003).

dominant waves, a threshold for B has also been proposed for the shorter compoents (Banner et al.,
2002).
We note that if B is constant then the wavenumber spectrum integrated over directions decays like k −3
and, assuming that waves are linear, the frequency spectrum decays like f −5 . Besides, a constant non-
dimensional spectrum means that all the scales have the same shape: the waves are self-similar. We had
found before that non-linear wave-wave interactions tend to impose a f −4 shape to the spectrum. Well,
this shape is not possible beyond some frequency where it crosses the f −5 asymptote as the steepness of
the waves must be limited by breaking.

8.3.4 A spectral approach

In order to decompose the overall dissipation rate ϵtot across the spectrum one can use an empirical
approach by distributing the total dissipation with particular shape factor, e.g. using a distribution of
the dissipation proportional to f 2 , as is often done for depth-induced breaking in the surf zone. In that
case, waves are almost not dispersive and it makes sense to combine all components.
In other conditions, it makes sense to decouple the disspation of wave components that have very
different wavelengths and directions. In that case we can define a parametrization that is consistent with
the data of Banner et al. (2000).
A first step is thus to determine the breaking probability for different wave ’scales’: these scales are
spectral regions within which the frequencies and directions are close enough that their superposition
produce well defined wave groups with long-lived wave crests that have enough time evolve towards
8.4. SPECTRAL ENERGY BALANCE 95

(b) (c)

(d)

Figure 8.8: Breaking probability and saturation

The probability of breaking of dominant waves bT is linked to the steepness of the dominant waves ε
which is related to the saturation ε2 ≃ 4B (k) with ∆θ = π and ∆k ≃ 0.6k. The threshold over which
breaking is observed is ε = 0.055, which corresponds to B (k) ≃ 1 × 10−3 .

breaking. We can then link the breaking probability to B which, taking p = 2 is a non-dimensional
variance of the orbital velocity.
The second step is to atribute to each spectral component that contribute to a given scale the breaking
probability and dissipation rate. This gives a dissipation source term
Z
Sdis,s (k) = h(k′ − k)P (B(k′ )) q (B ′ (k′ )) dk′ E (k) . (8.42)
k

in which the integral over k′ is the deconvolution from the scales to the spectral components with a filter
h, with P the breaking probability for the a give scale, and q the dissipation rate per unit crest length.
This approach was formalized by Filipot et al. (2010a); Filipot and Ardhuin (2012).
In practice it may be more efficient to parameterize the dissipation rate directly from the spectrum,
avoiding costly sums as in eq. (8.42). Recent exemples include Romero (2019). How B or the dissipation
rate is defined from an integrated spectrum and how B is transformed in a dissipation rate can have
profound effects on the shape of the wave spectrum. All these parametrizations combine this kind of
spontaneous dissipation rate with the dissipation of waves that are steep, and an induced Sdis,c dissipation
rate, in which the effect of long waves on short waves is parameterized. One parameterization for that
effect by Ardhuin et al. (2010) assumes that the short waves are ’wiped out’ by the long waves. That
form is not sufficient to enhance the dissipation of short waves. Peureux et al. (2021) proposed that
short wave breaking is enhanced by the effect of long waves through a direction-dependent modulation
instability. More work is needed to verify how much that effect can explain the modulation of breaking
waves by long waves observed by Dulov et al. (2002), and the coefficients used by Romero (2019) probably
exaggerate that effect to arrive at realistic spectral shapes.

8.4 Spectral energy balance

Given the uncertainties on generation and dissipation processes the empirical parameters in these two
source functions are usually adjusted so that the integration of the full spectral evolution equation gives
a realistic spectral evolution, and also compensate for errors introduced by the DIA parameterization.
Models are generally tuned to reproduce observations of shore-normal fetch-limited growth (e.g. Kahma
96 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER

1
x 10-3
S (m2) or E x 10 -3 (m2/Hz)

0.5 EE(f)
S
S in
Sin
0 nl
SSnlds
-0.5 S
Sds
tot
U10=15 m/s, t=3h
-1 S
tot
-1.5
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
x 10-4 f (Hz)
S (m2) or E x 10 -5 (m2/Hz)

6 U10=15 m/s, t=48h

4
2
0
-2
-4

Figure 8.9: Source term balance

Source terms for a wind speed of 15 m/s with a uniform sea starting from rest at t = 0. (a) spectrum
and source terms after 3 hours, (b) after 2 days.

and Calkoen, 1992), slanting fetch growth (Ardhuin et al., 2007), and the global swell climate (Ardhuin
et al., 2010).
The wave energy equation is, in spectral form,
dE (k)
= Sin (k) + Snl (k) + Sdis (k) (8.43)
dt
Eq. (8.43) where the total derivative is a rate of change following wave packets along rays, is most often
written in Eulerian form. In the absence of currents it is,
∂E (k)
+ ∇x · (Cg E (k)) + ∇k · (Ck E (k)) = Sin (k) + Snl (k) + Sdis (k) (8.44)
∂t
in which ∇x and ∇k are divergence operators in physical and wavenumber space, respectively, Cg and
Ck are the corresponding propagation speeds. Cg is the vector group speed, which points into the
direction of the wavenumber k, and Ck is the rate of change of the wavenumber k, which is due to
the bottom slope and the Earth curvature. Indeed, waves follow geodesics, which are great circles on a
spherical Earth. Hence their direction, relative to the north pole, change during propagation.
Among the three source terms, Sin defines the range of frequencies where the wind sea is generated,
with phase speeds less than the wind speed, this is why it requires very strong winds to produce long
waves that will radiate as swells. At low frequencies Sin is weakly negative. The energy provided by
the wind via Sin is redistributed by the wave interaction term Snl with a flux to both high and low
frequencies. The low frequency flux makes it possible to have fully developed waves that actually travel
faster than the wind, with a peak frequency up to 1.2 times U10 . The dissipation term Sds removes the
excess of energy due to strong wave steepness.
The sum of the three terms gives the trend of the spectral evolution, which, integrated in time gives
the evolving spectrum, with an example on figure 8.9. All operational wave models use a discretization
of the energy balance equation (8.44), with finite differences in time.
8.4. SPECTRAL ENERGY BALANCE 97

Most differences between models, at least for the windsea part of the spectrum, are due to differences in
the parametrizations. Figure 8.10 shows and example of different parametrizations. Todays most accurate
model results have been obtained with the parametrizations by Rascle and Ardhuin (2013). There are
still problems at short fetch and for high frequencies, with a poor representation of the directional
distribution and an overestimation of the energy level at f > 3fp . This is why the parameterization by
Romero (2019) has been designed to correct some of these shorter wave problems.

Hs: 0.76m Hs: 0.93m Hs: 1.15 m

fp:0.28 Hz fp:0.25 Hz fp:0.23 Hz
Mean dir. at fp: 319° Mean dir. at fp: 305° Mean dir. at fp:307°
Spread at fp: 20.5° Spread at fp: 22.5° Spread at fp:21.5°
U10: 10.67 m/ s U10: 10.67 m/ s

0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00

TC JHHK WAMDI+WRT
0.00 0.25 0.50 0.75 1.00
E(f,q) (m2/Hz/Rad)

2 Sin Sin Sin

0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00

2•10 -4 1•10 -4 0 1•10 -4 2•10 -4

S(f,q) (m2/Rad)

0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00

2 Snl Snl Snl

0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00 0.60 0.40 0.20 0.00

2 Sds Sds Sds

Figure 8.10: Three sets of parameterizations, three different balances

Source terms for a wind speed of 10 m/s at a fetch of 40 km. Left: parametrization by Tolman and
Chalikov (1996) with much weaker input and disspation, center: WAM-Cycle 4 (Janssen et al., 1994),
right WAM-Cycle 3 in which the DIA parametrization has been replaced by an exact calculation of the
interactions. The Tolman and Chalikov (1996) terms have been multiplied by 2 in order to be in the
same range. Picture from Ardhuin et al. (2007).
98 CHAPTER 8. PHYSICS OF SPECTRAL WAVE EVOLUTION: DEEP WATER
Chapter 9

Waves and momentum

9.1 Stokes drift

The displacements of fluid parcels caused by waves is dominated by the periodic oscillations ξeh and ξe3
derived in chapter 2 for linear waves. For many application this first approximation may not be sufficient.
Let us examine what has been neglected. To be exact, the position (x(t), z(t)) of a fluid parcel is the
sum of its velocities at the successive positions (e.g. Phillips, 1977, p. 43),
Z t
(x(t), z(t)) = (x(0), z(0)) + (u (x(t′ ), z(t′ ), t′ ) , w (x(t′ ), z(t′ ), t′ )) dt′ . (9.1)
0

To the first order in steepness ε = ka, the velocity equals the velocity at the initial position,

(u (x(t′ ), z(t′ ), t′ ) , w (x(t′ ), z(t′ ), t′ )) ≃ (u (x(0), z(0), t′ ) , w (x(0), z(0), t′ )) . (9.2)

Integrating in time this gives a periodic motion, (x(t), z(t)) = (x(0), z(0)) + (ξh , ξ3 ) as given by eqs.
(2.31)–(2.32).

0 T=5 s, H = 0.6 m 0.5 0

−1 −1
0
z (m)

−2 −2
−0.5
−3 −3
0 2 4 6 8 10 12 0 0.1
x (m) velocity (m/s)
2 4 2

1
T=5 s, H = 2.0 m 1
2
0 0
z (m)

0
−1 −1
−2
−2 −2

−3 −4 −3
0 2 4 6 8 10 12 14 0 0.5
x (m) velocity (m/s)

Figure 9.1: Left: Horizontal velocity field at t = 0 and particle trajectories integrated over 2 Eulerian
periods. Solid lines are isotachs and dashed lines are streamlines in the frame of reference moving with
the wave phase speed. Right: vertical profile of Eulerian mean water velocity (dashed, velocity is set to
zero in the air for computing the average), Generalized Lagrangian mean (red), Lagrangian mean (black)
and Lagrangian mean from the linear approximation (blue). Both top and bottom panels are computed
with streamfunction theory to 80th order (Dalrymple, 1974). Non-linear terms are only significant in the
bottom case.

However, the exact calculation (figure 9.1) shows that the motion is not periodic. This is because
the velocity varies over the displacement distance. That variation introduces a correction of the position

99
100 CHAPTER 9. WAVES AND MOMENTUM

that is of second order in the wave amplitude. This is given by the following Taylor expansion
u (x(t′ ), z(t′ ), t′ ) = u (x(0), z(0), t′ )
+u2 (x(0), z(0), t′ )
∂
+ξh (t′ ) · ∇u (ξh (0), ξ3 (0), t′ ) + ξ3 (t′ )
u (ξh (0), ξ3 (0), t′ ) + O(ε3 ), (9.3)
∂z
where u2 is the gradient of the second order potential ϕ2 , which comes in the solution of nonlinear wave
equation (see Part 3). However, we do not need to worry about that term because the average over time
of u2 is zero. We will thus only consider the last two terms which are given by products of linear terms.
First of all ξh (t) is 90 degrees out of phase with u, and ∇u is also 90 degrees out of phase with u. Hence
ξh (t) and ∇u are in phase and their product has a non-zero average, kσa2 cosh2 (kz + kh)/[2 sinh2 (kD)].
Physically this corresponds to the fact that the orbital velocity at the crest is in the same direction as the
crest motion, hence particles ride with the crest longer than they stay in the trough where particles move
opposite to the wave propagation. Likewise, ξ3 is in phase with ∂u/∂z, which corresponds to the fact that
the horizontal velocity increases vertically, and thus a particle goes forward faster when it is up compared
to when it is down. That other product gives also a non-zero average, kσa2 sinh2 (kz + kh)/[2 sinh2 (kD)].
These two effects add up to give the Stokes drift, defined as
Z T
1 cosh(2kz + 2kh)
Us ≡ u (x(t′ ), z(t′ ), t′ ) dt′ = σka2 (1 + O(ε)), (9.4)
TL 0 2 sinh2 (kD)
where TL is the Lagrangian period, namely the time it takes for a parcel starting from a crest to loop to
the next crest. The bottom panel in figure 9.1 clearly shows that the Lagrangian period is longer than
the Eulerian period: after two Eulerian period the particles that started at the crest are not yet back to
the crest, because they have moved forward ahead of the next crest.
In deep water (kD ≫ 1), Us goes to
Us = σka2 exp(2kz). (9.5)
This speed Us is the average drift speed of a water parcel and it is directed in the direction of wave
propagation. It is a mean Lagrangian velocity. Hence the parcel displacement is not exactly periodic and
the parcels move forward (figure 2.3). This drift velocity decreases strongly with depth, in deep water
this decrease is twice as fast as the orbital velocity, as shown on figure 2.3.
First of all, it should be emphasized that we have computed a second order drift from a linear wave
field. The top-right panel in figure 9.1 clearly shows that for a nearly linear wave the Stokes drift is
dominated by the contribution of linear wave field (blue profile). There is a widely held misconception
that Stokes drift is a property of non-linear waves. This is wrong. In fact, the Stokes drift is a quadratic
property, just like the wave energy. Linear waves have a Stokes drift, just like they have an energy.
This correspondence between Stokes drift and energy is a very profound physical property (Andrews
and McIntyre, 1978). The Stokes drift computed here is the pseudo-momentum of the wave field. When
integrated over the vertical it is
Z 0
w cosh(kD) a2 Et
M = ρw Us dz = ρw σa2 = ρw g = (9.6)
−h 2 sinh(kD) 2C C
with Et the wave energy per unit surface. Eq. (9.6) is a very general result in physics for all types
of waves. It is the same result as the momentum of a photon p = E/c, where c is the speed of light.
For non-linear waves of finite amplitude, the exact relationship is Mw = 2Ec /C, where Ec is the mean
kinetic energy per unit surface (Longuet-Higgins, 1984).
Finally, this transport can also be computed from the Eulerian velocity, integrated from the bottom
to the crest level, *Z + Z
ζ a
w
M = ρw udz = ⟨ρu⟩ dz (9.7)
−h −a

These two estimates of the transport correspond to the two different views of the Stokes drift. In the
Eulerian point of view, the mass transport only happens between the crests and the troughs, with a
profile that is a parabola for linear waves. In the Lagrangian point of view the drift happens over the
entire water column (figure 9.2).
Finally, we also note that there is also a Stokes drift in the air, which is also in the direction of
propagation and has a profile that decays vertically up, like the profile of Stokes drift in deep water. It
is easily computed by the same method that was used here in the water.
9.2. THE ‘SHEAR’ OF THE STOKES DRIFT 101

0 0

free surface -5 -5
water
-10 velocity -10
(Eulerian Lagrangian
-15 -15 mean
mean)
-20 -20
0 40 80 0 0.5 0 0.5
x (m) speed (m/s)

Figure 9.2: Mass transport from an Eulerian and Lagrangian point of views, computed here for an
amplitude a = 3 m, wavelength L = 100 m in 30 m water depth, using linear wave theory.

9.2 The ‘shear’ of the Stokes drift

Another interesting property of the Stokes drift is that it has a strong vertical shear, in particular in deep
water, but also for strongly nonlinear waves in shallow water (Miche, 1944a). This may seem paradoxical
that an irrotational flow, with zero vorticity, has an average with a strong shear. This apparent paradox
comes from the fact that the curl operator does not commute with the Lagrangian average. Another
funny property is that the Lagrangian average of dw/dx is also non-zero, although the motion is periodic
in x (Ardhuin and Jenkins, 2006).
The shear of the Stokes drift also has the very nice property to persist in the presence of strong
mixing, because the mixing is usually done by eddies that have time scales longer than the wave period.
Hence one should be very careful to apply to the wave motion the eddy viscosity ideas that are usual for
currents. Indeed, the assumption that the effect of turbulence is analogous the the molecular viscosity
scaled up by a factor 100 or more does not work with waves: if it were the case the waves would dissipate
over very short distance and swells would never be recorded from remote storms. The ‘eddy viscosity’
idea is thus a very dangerous idea in the presence of oscillations, and turbulence closures should generally
be visco-elastic and not just viscous (e.g. Miles, 1996).
This persistence of the vertical shear of the drift makes the ocean surface boundary layer special and
tends to tilt vorticity that is perpendicular to the wind into the wind direction, promoting the generation
of rolls aligned in the wind. These are known as Langmuir circulations (Langmuir, 1938), which is a key
component of the ocean mixed layer. These properties of the upper ocean will be further discussed in
chapter 14.

9.3 Random waves and practical estimation

Although the Stokes drift plays an important role, it is very difficult to measure in the ocean because
it requires the measurement of both the Eulerian velocity and the drift at the same place, and drift if
usually measured at the surface only with objects that can be also affected by the direct influence of the
wind or the radiation stress of the very short waves. Indeed, no object or surfactant drifts exactly like
the water surrounding it, and only non-intrusive methods using for example infrared technology (Veron
et al., 2008) or surface wave dispersion (Laxague et al., 2018) can provide unambiguous measurements.
As a result, almost all estimates of the Stokes drift are based on the measurement of the wave spectral
properties and the use of linear wave theory because the Stokes drift for a random wave field with no
phase-correlations between the wave components is simply derived from eq. (9.3), as done by Kenyon
(1969). The horizontal Stokes drift vector is thus,
Z ∞ Z 2π
cosh(2kz + 2kh)
(US , VS ) = 2σk (cos θ, sin θ)E(f, θ)dθdf. (9.8)
0 2 sinh2 (kD) 0
102 CHAPTER 9. WAVES AND MOMENTUM

The inner integral can be replaced by the moments a1 and b1 that are directly measured by wave buoys
(see chapter 4 for details),
Z ∞
cosh(2kz + 2kh)
(US , VS ) = 2σk (a1 (f ), b1 (f ))E(f )df. (9.9)
0 2 sinh2 (kD)
The broad directional spectrum at high frequency contributes little to the total surface drift (Peureux
et al., 2018). As a result, the vertical profile of the Stokes drift can have a strong shear but weaker than
predicted by simple parameterizations such as the one by Breivik et al. (2016), or when using parametric
spectra such as the one by Elfouhaily et al. (1997).
Finally we note that non-linear effects typically enhance the drift, such as shown in figure 9.1. But
there is no published theory giving the expression for nonlinear corrections to the Stokes drift for a random
wave spectrum. That can be derived using the surface elevation second order spectrum (Janssen, 2009;
Leckler et al., 2015).

9.4 Radiation stresses and the flux of wave momtum

Just like the wave energy is radiated by the wave field, the wave momentum is also radiated away.
Considering monochromatic waves (i.e. with a single period and direction) propagating along the x-axis,
there is a flux of momentum across any surface perpendicular to the propagation direction. By definition,
there are two ways to move momentum in the x direction,
ˆ by pushing things around: the pressure forces transmit momentum from one water column to the
next,
ˆ by advecting the momentum density per unit volume ρu with the velocity u.
The usual definition of the momentum flux associated to waves, and called the ‘radiation stress’,
is the flux of momentum when the waves are present minus the flux when waves are absent. It is a
strange definition because the interaction of waves and currents make it impossible to have the exact
same current without the waves, in practice it means that we assume the sea level to remain the same
and we just set the orbital velocity to zero and the pressure becomes the hydrostatic pressure. A more
rigorous definition is given in Andrews and McIntyre (1978) and discussed in Part 3. Anyway, let us
make this thought experiment of removing the waves, we define the first component of the radiation
stresses by a phase average of the wave effects,
Z ζ Z ζ
rad
Sxx = p + ρw u2 dz − b2 dz,
p0 + ρw u (9.10)
−h −h

where the second term correspond to the pressure p0 in the absence of waves (but with the same mean
sea level) and u
b is the the mean (current) velocity.
When the waves are present, the pressure is obtained by integrating the vertical component of the
momentum equation (we take v = 0 because waves propagate only along the x-axis)
Z ζ
∂ρw w ∂ρw uw ∂ρw w2

∂p
+ + + + ρw g dz = 0 (9.11)
z ∂t ∂x ∂z ∂z
The first term yields
Z ζ Z ζ
∂ρw w ∂ ∂ζ
dz = ρw wdz − ρw w (ζ) , (9.12)
z ∂t ∂t z ∂t
the second yields,
Z ζ Z ζ
∂ρw uv ∂ ∂ζ
dz = ρw uwdz − ρw u (ζ) w (ζ) (9.13)
z ∂x ∂x z ∂x
the third yields
ζ
∂ρw w2
Z
dz = ρw w2 (ζ) − ρw w2 (z) , (9.14)
z ∂z
and, assuming p (ζ) = 0, the fourth term gives the pressure at elevation z,
Z ζ
∂p
dz = −p (z) . (9.15)
z ∂z
9.4. RADIATION STRESSES AND THE FLUX OF WAVE MOMTUM 103

Gathering all this and using the surface kinematic boundary condition (2.5) gives,
Z ζ Z ζ
∂ ∂
p = ρw g (ζ − z) + ρw wdz + ρw uwdz − ρw w2 . (9.16)
∂t z ∂x z

We now take the average over a wave period and because linear wave theory has uw = 0, we find


p = ρw g ζ − z − ρw w2 . (9.17)
rad
We can now compute the different pieces that make up Sxx . It is straightforward to generalize the
calculation to the flux of xα momentum in the xβ direction, where both xα and xβ can be either x or y.
Namely,
Z ζ Z ζ Z ζ !
rad
Sαβ = ρw uα uβ dz + δαβ p − p0 dz + pdz . (9.18)
−h −h ζ

rad rad
We note that the pressure only come in Sxx and Syy because pressure is a normal stress. In our case
rad rad
of v = 0 and with waves propagating along the x axis, we clearly have Sxy = Syx = 0.
rad
Now replacing the velocity and pressure in Sαβ with linear wave theory, the first piece

Z ζ
ρw ui uj dz (9.19)
−h

only comes into Sxx , and we actually have calculated almost the same integral in chapter 2. Indeed, the
energy flux is the same integral with pu instead of u2 . For linear waves, u = p/Cρw , where C = ω/k is
the phase speed. Hence the first piece of Sxx is equal to Et Cg /C. This is very nice, this is the flux of
wave momentum, just like Cg Et is the flux of wave energy.
For the last piece and for z ≃ ζ we may replace p by ρw gζ. That piece yields ρw gζ 2 /2, which is equal
to the potential energy, namely Et /2. Finally, we use the linear wave expression for p to get the second
rad
piece of Sxx ,
Z ζ Z ζ
a2 k
p − p0 dz = ρw gDζ − ρw g sinh2 (kz + kD) dz (9.20)
−h sinh (2kD) −h

using now the fact that

sinh2 x = (cosh 2x − 1) /2 (9.21)
we obtain
Z ζ
1
sinh2 (kz + kh) dz = (sinh (2kD) − 2kD) (9.22)
−h 4k
and our second piece of Sxx becomes
Z ζ  
ρw g kD
p − p0 dz = ρw gDζ + E 2 −1 . (9.23)
−h 2 sinh (2kD)

Finally, the third piece is

ζ
ζ2
Z
E
pdz = ρw g = ρw g . (9.24)
ζ 2 2
rad
It is common practice to remove the ζ term from Sαβ because this is the hydrostatic pressure that
varies with the sea level. Finally we have,
   
rad Cg kD Cg Cg 1
Sxx = ρw gE + = ρw gE + − (9.25)
C sinh (2kD) C C 2
rad E 2kD
Syy = ρw g (9.26)
2 sinh (2kD)
 
E Cg
= ρw g 2 −1 (9.27)
2 C
104 CHAPTER 9. WAVES AND MOMENTUM

Hydrostatic pressure gradient

z=ζ1 Mean sea level

Wave-induced z=ζ2
pressure S2p
S 1p
Cg2 E2 / C2
Wave momentum flux
Cg1 E1 / C1

H
Bottom, z = -

Figure 9.3: Momentum fluxes associated to waves propagating over a variable bathymetry from left to
right, towards shallower water. The radiation stress Sxx is the sum of the flux of wave momentum
ρw gECg /C and the pressure correction S p = 0.5ρw gE(2Cg /C − 1). This pressure correction goes to zero
rad
in deep water. Here a difference between Sxx = ρw gECg /C + S p at points 1 and 2 drives a lowering of
the mean water level from ζ = ζ1 to ζ = ζ2 . This phenomenon is called the set down.

rad
For waves propagating in any azimuth θ relative to the x-axis, the non-isotropic part of Sαβ is modified.
rad rad
The orbital velocity u becomes u cos θ and the v component becomes v sin θ. Hence Sxy and Syx are
not zero anymore,
  
rad Cg E Cg
Sxx = ρw g E cos2 θ + 2 −1 (9.28)
C 2 C
  
Cg E Cg
rad
Syy = ρw g E sin2 θ + 2 −1 (9.29)
C 2 C
rad rad Cg
Sxy = Syx = ρw gE sin θ cos θ. (9.30)
C
This result was obtained using linear wave theory. For periodic nonlinear waves, S rad is actually very
close to the linear theory result, with a difference that is less than 5% for kD < 0.3.
In general the radiation stress is thus the sum of the wave momentum flux and a correction for the
change in mean pressure in the presence of waves. If the wave field varies in space, a divergence of the
radiation stress is like a force exerted by the waves on the mean flow. In the absence of wave breaking
the usual response to this force is a change in mean sea level, as illustrated in figure 9.3. This type of
effect is discussed in more detail in chapter 19 which deals with nearshore hydrodynamics.
Chapter 10

Wave-current interactions

10.1 Current effects on waves: propagation

10.1.1 A uniform current
For waves over a flat bottom, in the absence of any friction, the shape of waves and their motion is
unchaged by a Eulerian uniform horizontal current U. This is easily seen by changing the reference
frame to a reference frame in which the current is zero. This is obtained with a position x′ defined as

x′ = x + Ut. (10.1)

In particular the phase in the moving reference frame is given by eq (10.2), and it is transformed back
to the fixed reference frame as

Θ′ = k · x − k · Ut − σt + Θ0 , = k · x − ωt + Θ0 (10.2)

where we have defined the absolute radian frequency,

ω = σ + k · U. (10.3)

The effect of a uniform current is thus a simple Doppler shift of the phase, which gives a modification
of the phase speed and group speed,

C′ = C+U (10.4)
C′g = Cg + U, (10.5)

where C′ and C′g are the phase and group speed in the fixed reference frame. σ is called the relative radian
frequency. We note that the general definition of the group speed still holds, C′g = (∂ω/∂kx , ∂ω/∂ky ).

10.1.2 Effect of current changes along the propagation direction

Taking a stationary current U (x) in the direction of progation x, or in the opposite direction, varying
only in this same direction, the waves kinematics must adjust to this variation in current speed. For
slow variations on the scale of the wavelength, we use the approximation by Wentzel, Kramers, Brillouin
et Jeffreys (WKBJ), i.e. the waves are locally sinusoidal, and the conservation of the number of crests
(which holds for linear monochromatic waves), and which gives a constant absolute frequency ω because
the medium in which the waves propagate is constant (an interesting case where ω is not conserved is
when the water depth changes in time). As a result the intrinsic frequency σ must adjust. If the current
accelerates in the direction of propagation, then σ must be reduced because k · U is positive. Because
σ grows with k, the wavelength increases. Conversely, for waves propagating against an accelerating
current, k · U is negative, hence σ and k must increases to keep ω constant.
A first application can be done in the case of deep water waves with a known radian frequency σ1 ,
propagating from a region 1 without current, to a region 2 where a current of magnitude U is uniform
and opposite to the waves.
For linear waves, the number of crests is conserved, which can be written as a conservation of the
absolute frequency,
σ2 − k2 U = σ1 . (10.6)

105
106 CHAPTER 10. WAVE-CURRENT INTERACTIONS

Combined with the dispersion relation σ12 = gk1 and σ22 = gk2 , this gives us a second order equation
for the unknown relative frequency σ2 ,

U 2
σ − σ2 + σ1 = 0. (10.7)
g 2

This second-order polynomial equation has two solutions, but we shall only consider the one that gives
σ1 = σ2 when U goes to zero,
√
1 − 1 − 4α
σ2 = σ1 (10.8)
2α
with α = U σ1 /g, the ratio of the current velocity and the phase speed C1 = g/σ1 in region 1. We note
that there is no real solution for α < 0.25, in that situation the waves are blocked by the current and
cannot propagate in region 2. In the limiting case, α = 0.25, we have σ2 = 2σ1 and the local group speed
Cg2 is equal to U . In that case there is no wave energy flux, because the mean wave energy velocity is
Cg − U = 0.
We can further investigate the case when α ≪ 1. In that case, we may write the following expansion,
√
1 − 4α = 1 − 2α − 2α2 + O(α3 ) (10.9)

which gives

σ2 ≃ σ1 (1 + α), (10.10)
k2 = k1 (1 + 2α). (10.11)

This means that the intrinsic wave period σ is shortened in proportion to the ratio α = U/C1 and the
wavelength is shortened by twice that amount.
In order to know what happens to the wave height, we may consider the energy balance, as we shall do
later. But be careful, this situation is precisely a situation where the wave energy is not conserved, even
without any dissipation. What is conserved is the total energy in the system (waves + current), and in
practice the waves and currents exchange energy. In the absence of dissipation, this interaction happens
with the conservation of another quantity, the wave action, which we may define for monochromatic
waves as
gE
A= . (10.12)
σ
With a properly defined control volume, the equality of the action fluxes gives,

E2 E1
(Cg2 − U ) = Cg1 . (10.13)
σ2 σ1

which gives an amplification of the wave energy

σ2 Cg1
E2 = E1 ≃ E1 (1 + 4α + O(α2 )) (10.14)
σ1 Cg2 − U

in which three factors are important:

ˆ the increase of σ makes it necessary to increase E to keep A constant, this amplification comes
with a transfer of energy from the current to the waves, this gives the factor σ2 /σ1 ≃ (1 + α).

ˆ the fact that the advection velocity involves the current, Cg is replaced by Cg − U , this accounts
for half of the total effect because U/Cg = 2α.

ˆ the fact that the intrinsic group speed has been reduced because the wave have become shorter.

As a result, the change of wave height is only proportional to (1 + 2α), and the change in wave stepness
is proportional to (1 + 4α). The effect of currents can be much larger for time-varying currents (Ardhuin
et al., 2012a; Peureux et al., 2021). In practice, another effect can have large impacts on wave heights:
refraction.
10.1. CURRENT EFFECTS ON WAVES: PROPAGATION 107

10.1.3 Effects of current gradients in the transverse direction:

rays and refraction
Just like changes in water depth discussed in chapter 16, currents induce a modification of the phase
speed. As a result, any current gradient in the direction transverse to the wave propagation will cause
refraction, namely a gradual turning of the wave crests. The particularity of cases with current is that
the direction of the phase advection can be different from the direction perpendicular to the crests. The
directions in which rays bend follows Fermat’s principle, namely waves will always take the path that gives
the shortest propagation time. As a result, waves will focus in a jet that flows in the direction opposite
to the wave propagation direction (as shown in Figure 10.2) and diverge in a jet of same direction.
The equations of wave rays are the trajectories followed by wave packets in time, these are identical to
the propagation equations, before discretization, that are solved by spectral models like WAVEWATCH
III (equations 2.9 to 2.11 in Tolman et al., 2014)
dx
= Cg + u
b, (10.15)
dt
dk ∂σ ∂D ∂b
u
=− −k· , (10.16)
dt ∂D ∂s ∂s
 
dθ 1 ∂σ ∂D ∂b
u
=− −k· , (10.17)
dt k ∂D ∂m ∂m
where x is the horizontal position along the ray, D is the water depth, θ is the local intrinsic wave direction,
Cg is the vector intrinsic group speed, pointing in direction θ, s is a coordinate in the direction1 θ and m
is a coordinate perpendicular to s. These ray equations are the same as those used by Mathiesen (1987),
with the addition of finite depth and bottom refraction effects.
Landau and Lifshitz (1960) gave a simple result on the curvature of the rays followed by wave groups
in a medium with varying mean flow velocity. This was re-derived by Dysthe (2001) for surface gravity
waves. In the limit of a weak current, u b ≪ C, the radius of curvature of the rays is
1 Cg
R= = (10.18)
∂θ/∂s ∇×u b

where ∇ × ub is the vertical vorticity of the flow. As the wave directions are changed, the energy of
the wave can be focused by opposing current jets, explaining the occurence of very large waves in the
Agulhas current (Lavrenov, 1986).

Figure 10.1: Influence of the 4 different components of the current gradient U sin(Kx), V sin(Kx),
V sin(Ky), U sin(Ky), on waves propagating from left to right with a narrow spectrum and a period of
7 s (figure by G. Marechal).

In practice the importance of refraction on wave height depends on the current structure. Gradients
that are perpendicular to the dominant wave propagation and are coherent over long distances are most
effective in creating a variability in the significant wave height, as shown in Fig. 10.1. This analysis can
1 Due to the presence of the current, s differs from the along-ray direction.
108 CHAPTER 10. WAVE-CURRENT INTERACTIONS

be generalized to any current pattern to derive a map of wave heights from the map of currents (Wang
et al., 2023).
Figure 10.2 shows an example off the West coast of France where the very strong vorticity of tidal
currents, of the order of 0.001 s−1 is enough to bend wave rays for T = 10 s waves with a radius of
curvature of 8.6 km.
6
model without current
model with current
data
4
Hs(m)

0
(a)
23 26 29 1 4 7 10 13 16 19 22
Days (October 2008) Days (November 2008)

(b)
depth (m)

68

66

64
62
3

4
R=8.6 km

3
2

û (m/s) Hs (m)

2
1

1
(c) (d)
0

0
"Pierres Noires" buoy, WMO number 62069

Figure 10.2: Example of strong impact of currents on wave heights due to wave refraction by currents.
The top panel shows a time series of Hs recorded at the wave buoy ‘Pierres Noires’ (WMO number 62069)
and modeled with WAVEWATCH III, while the middle panel shows the water depth at the buoy. The
bottom maps show the (c) currents provided by the numerical model MARS2D and the corresponding
wave rays for T = 10 s computed by integration of eqs. (10.15)–(10.17), and (d) shows Hs and mean wave
directions, both for October 28, 2008, at 11:00 AM UTC (corresponding to blue arrow in a). Clearly,
the wave height has a strong tidal modulation which is due to currents. The typical curvature radius of
the rays is around 10 km in the current jet located south-west of the island of Ouessant. This jet peaks
1.5 hours after the high tide, and deviates the waves away from the Pierres Noires buoy, located 20 km
down-wave (Adapted from Ardhuin et al., 2012b).

10.1.4 Waves over vertically sheared currents

In the presence of vertical shear, the Laplace equation becomes the Rayleigh equation and the Doppler
shift k · U for linear waves can be estimated by solving it. Biesel (1950) gave solutions in the limit
kD ≪ 1 and for a constant shear, with a current varying from u b−D at the bottom to ub0 at the surface.
The phase speed is then
" 2 #1/2
U0 − U−D
C= + gD (10.19)
2
For more general current profiles and water depths, Kirby and Chen (1989) gave an approximate
solution in the limit of small variations of U compared to σ/k,
Z 0
σ cosh(2kz + 2kD)
C = +2 k·u
b dz (10.20)
k −D sinh(2kD)
10.1. CURRENT EFFECTS ON WAVES: PROPAGATION 109

10.1.5 Practical importance of currents

In practice, current effects on waves combines the ‘bunching’ or ‘concertina’ effect described in section
10.1.2 due to a converge in the current, with the refraction, associated with current vorticity. A third
effect is the enhancement of wave generation in waves against currents due to the relative wind. When
we use the frame of reference moving with the current u b , the wind speed in that frame of reference
changes from U10 to U10 − u b . In practice one should thus correct the wind speed provided by a weather
forecasting model to force the wind-wave generation... except for real adjustement of the wind to the
presence of the current which are generally not included in weather models. Hence, the true wind is
reduced by opposing currents so that the wind-wave forcing should be something between the modeled
U10 and U10 − u b , say U10 − rbu. Numerical experiments for large scale currents suggests that on average
r ≃ 0.5 is the right order of magnitude when adjusted to the fully coupled atmosphere-current solution
(J. Bidlot, personal communication 2010). This is very important for the ocean dynamics, and forcing
an ocean circulation model with r = 1 generally gives a work of the wind against the eddies that is too
strong and produces a strong underestimation of the eddy kinetic energy (Renault et al., 2016b), with
influences on the position of western boundary currents such as the Gulf Stream (Renault et al., 2016a).
In figure 10.3, we show an example of modelled current effects on waves in the Drake passage, between
Chile and Antarctica. The variability of wave heights is investigated by computed the spatial spectrum
of Hs . We find that the variation in wave heights at large scales is dominated by the effect of refraction,
whereas at scales around 10 km, the variability is mostly due to the relative wind and advection effects.
Although standard processing of satellite altimeter data at scales shorter than 80 km (e.g. figure 10.3.d)
is dominated by noise, adaptative denoising method are now producing data that validate this strong
effect of currents (Quilfen et al., 2018).
110 CHAPTER 10. WAVE-CURRENT INTERACTIONS

56° 56°

58° 58°

60° 60°

62° 62°
(a) (b) (c)
200 km 200 km 200 km
64° 64°
65° 60° 55° 50° 65° 60° 55° 50° 65° 60° 55° 50°

0.0 0.3 0.6 0.9 1.2 1.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5
û(m/s) Hs (m)
5 5 5
10 10 current
(d) (e)
(cm2 or cm2/s2 per cycle per km)

all 4 terms
4 4 only ref. ( =/ 0)
10 10 only adv. (a=1)
/
only bun. ( =0)
3 3 only r.w. (r=1)
10 10

2 2
10 10
1
1
10 10 only adv. & r.w.
( =0,a=1,r=1, =0)
AltiKa 2015 0
0
10 10 no refraction
no current /
( =0,a=1,r=1, =0)
long waves −1 only ref & adv.
−1 short waves 10 ( =0,a=1,r=0,
/ =0)
10
all waves only adv. & bun.
/
E(k)

−2 ( =0,a=1,r=0, =0)
−2 current 10 no current
10
−3 −2 −1 −2 −1
10 10 10 10 10
ky(cycles per km) k (cycles per km)

Figure 10.3: Maps for September 16 at 18:00 UTC for (a) surface current magnitude modeled by MITgcm
(b) the modeled significant wave height when the current forcing is included in WAVEWATCH III (c)
significant wave height without effects of currents and wind directions (arrow). The dashed box is the
region used for spectral analysis. (d) Spectra of the modeled zonal current and Hs along the north-
south direction, with contributions of waves of periods shorter or longer than 6 s, along-track measured
spectra from AltiKa is shown for comparison, and power laws k −2 and k −3 are shown in green. (e)
Omnidirectional spectrum of Hs and contributions of the current through the four different terms of the
wave action equation (10.42) can be revealed by progressively switching off the different terms: refraction
θ̇, change in wavenumber k̇, relative wind r, and advection by uE and vE in λ̇ and ϕ̇. Adapted from
Ardhuin et al. (2017b).

10.2 Wave effects on currents

To make it as simple as possible we first follow the approach of Phillips (1977), in which the velocity
field u is the sum of a current u b , that we assume uniform over the vertical, and a perturbation u′ such
that its average is zero, at least for the points of space that are always in the water.

10.2.1 Mean flow equations integrated over the vertical

A systematic discussion of wave effects on currents started with Longuet-Higgins and Stewart (1964),
and we follow their derivation. The three-component momentum vector ρ(u, v, w) can be advected in any
direction, giving a flux tensor ρui uj . When one considers only velocity fluctuations u′ , the momentum
flux due to the self-advection of momentum is the usual Reynolds stress −ρu′i u′j , which is used in the
analysis of turbulence. The divergence of this flux is equivalent to a macroscopic force that accelerates
the fluid particles, which is easily understood by considering the momentum balance over an elementary
cube.
Longuet-Higgins and Stewart (1964) introduced a similar tensor caused by waves, the radiation stress
S rad , that is confined to the two horizontal dimensions, it represents a flux of momentum that comes from
the average of the wave-induced Bernoulli pressure p + u2 . Compared to Taylor expansion methods, the
present approach is easier because the non-linear wave effects are directly included by using conservation
10.2. WAVE EFFECTS ON CURRENTS 111

equations.
Starting from the conservation of momentum in any horizontal direction α, with our notation uα is
either the u or the v component, 2
∂ ∂ ∂ ∂ 2 uα
(ρw uα ) + (ρw uα uβ + pδαβ ) + (ρw uα w) + εαβi fi ρw uβ = ρw ν , (10.21)
∂t ∂xβ ∂z ∂z 2
in which there is an implicit sum over the repeated indices, here β, and where εαji fi uj is the α component
of the vector product of the Coriolis parameter vector with the speed vector. In the following we classically
consider only the vertical Coriolis parameter f3 .
To be sure that we understand the implicit sum, an explicit form of the equation for the u component
is,
∂ ∂ ∂ ∂ ∂2u
ρw u2 + p +


(ρw u) + (ρw uv) + (ρw uw) − f3 v = ρw ν 2 , (10.22)
∂t ∂x ∂y ∂z ∂z
Integrating over the vertical, and using the boundary conditon at the surface and bottom, w(ζ) =
∂ζ/∂t + u · ∇ζ, yields
Z ζ Z ζ Z ζ
∂ ∂
ρw uα dz + (ρw uα uβ + δαβ p) dz + εαβ3 f3 ρw uβ dz
∂t −h ∂xβ −h −h
∂ζ ∂h ∂uα ∂uα
= pa − p(−d) + ρw ν |z=ζ − ρw ν |z=−h (10.23)
∂xα ∂xα ∂z ∂z
We now take the average over wave phases 3 .
b + u′ ,
The second integral in 10.23, gives, on average4 , using u = u
Z ζ Z ζ Z ζ Z ζ Z ζ
ρw uα uβ dz = ρw u
bα u
bβ dz + ρw u
bα u′β dz + ρw u
bβ u′α dz + ρw u′α u′β dz. (10.24)
−h −h 0 0 −h

We now define the α-component of the total mass transport vector as5 ,
Z ζ
Mα = ρw uα dz (10.25)
−h

and the wave-induced mass transport, also known as the Stokes transport6 , and the difference of these
two is the mass transport due to the mean flow
Z ζ Z ζ
Mαm = Mα − Mαw = ρw u
bα dz = u
bα ρw u
bα dz = ρw Db
uα . (10.26)
−h −h

We can thus re-write eq. (10.24) as

Z ζ Z ζ
1
Mαm Mβm + Mαm Mβw + Mβm Mαw + ρw u′α u′β + δα,β pdz


ρw uα uβ dz = (10.27)
−h ρw D −h
Z ζ
1
Mα Mβ − Mβw Mαw + ρw u′α u′β dz.


= (10.28)
ρw D −h

The conservation of the vertical momentum is, neglecting viscosity and turbulence,
∂ ∂ ∂ ∂
ρw w2 = −ρw g.


(ρw w) + (ρw uw + p) + (ρw vw + p) + (10.29)
∂t ∂x ∂y ∂z
2 Taking into account viscosity and the air pressure p allows to include the surface and bottom stress, which are not in
a
the derivation by (Phillips, 1977, page 62).
3 Phillips (1977) uses a horizontal average but as a result the obtained equatiosn are only valid at scales larger than the

wavelength. Using the phase average allows to keep sub-wavelength variations such as in the case of partial standing waves,
see Ardhuin et al. (2008c).
4 This is here that we use the fact that u
b is independent of z, otherwise we would have an extra term coming from the
vertical current profile, that would be equivalent to a horizontal mixing term (Svendsen and Putrevu, 1994).
5 This notation differs from Phillips (1977) who used M fα .
6 The decomposition 10.28 and this definition of the Stokes transport require, to be well defined, an analytical extension

of the velocity field u′ in the air. This is not the real velocity field which, like the density, is very different from this
extension. A nice way to avoid this issue is to use an average that follows the up-and-down motion of the sea surface (e.g.
Ardhuin et al., 2008b).
112 CHAPTER 10. WAVE-CURRENT INTERACTIONS

A vertical integration gives, after using the surface kinematic boundary condition,
Z ζ Z ζ Z ζ
∂ ∂
p(z) = pa + g ρw dz + ρw wdz + ρw uβ wdz − ρw w2 (z). (10.30)
z ∂t z ∂xβ z
in which the second term is the hydrostatic pressure, and the third vanishes only if the spatial average
is independent of time – which is not the case in the presence of waves travelling in opposite directions
– the fourth term vanishes for a motion that is periodic in time, and the last term is the mean dynamic
pressure which is zero only if the velocity is zero. If we make all these assumptions the bottom pressure
is hydrostatic,
p(−h) = ρw gD + p′ . (10.31)
As a result the mean value of the right hand side of (10.23) is
 
∂h ∂ 1 2 ∂ζ
τa,α − τb,α + ρw gD = τa,α − τb,α + ρw gD − ρw gD , (10.32)
∂xα ∂xα 2 ∂xα
where τa,α and τb,α are the α-components of the surface and bottom stress. The correlation p′ ∂h/∂xα
has been included in τb,α as it represents the form drag on the bottom, while the other part of τb,α is the
skin friction given by the average of the viscous stress.
We now have (Smith, 2006),

Mα Mβ − Mαw Mβw
 
∂Mα ∂ rad ∂ζ
+ + Sαβ + εαβ3 f3 Mβ = −ρw gD + τa,α − τb,α . (10.33)
∂t ∂xβ ρw D ∂xα

where the radiation stresses are defined by Phillips (1977) as the difference between the momentum flux
and the flux in the absence of waves,
Z ζ
rad 1
Sαβ = ρw uα uβ + δαβ pdz − ρw gD2 δαβ − ρw u
bα u
bβ D. (10.34)
−h 2

The expression for S rad using linear wave theory is given in chapter 9.
The same procedure applied to the mass conservation equation yields

∂Mβ ∂D
+ ρw = 0. (10.35)
∂xβ ∂t

Equations (10.33) and (10.35) are the basic equations used in nearshore hydrodynamics, in their most
simple form. They can be used to explain a wide variety of phenomena, from changes in the mean sea
level, along-shore currents in the surf zone, infra-gravity waves .... All these will be discussed in chapter
19. The assumptions made by Phillips (1977) are fairly restrictive. Removing many of these, we present
in Part 3 an extention of the wave-current interaction theory to three dimensions.
We note that the mass transport Mα can be expressed with a mean velocity Uα ad Mα = ρw DUα .
This mean velocity includes the Stokes drift, which usually has a strong vertical gradient, even possibly
in the surf zone (Ardhuin et al., 2008c) and the mean current which can also have a strong vertical shear.
This mean speed can thus be very different from a measured mean velocity (Eulerian mean), and also
different from a tracer mean speed if tracers are not homogeneously distributed over the vertical, as it is
the case for suspended sediment.

10.2.2 Total energy

From the momentum equation the energy equation can be derived. With a viscous dissipation rate ϵ per
unit volume and the stress tensor pδij + τij , in which i and j can be any of the three directions x, y and
z,
∂  ui ui  ∂ h  ui uj i
ρw + gz + ui ρw + gz + δij p + τij = ϵ. (10.36)
∂t 2 ∂xj 2
Following the derivation of Phillips (1977, page 63), and defining the total depth-integrated mechanical
energy per unit surface as
Z ζ  
1 1 M α Mα 1 2 1
Ea = ρw ui ui + gz dz = +ubα Mαw + ρw g(ζ − h2 ) − ρw u bα + ρw gE + E ′ , (10.37)
bα u
−h 2 2 ρ w D 2 2
10.3. ENERGY EXCHANGE AND WAVE ACTION 113

in which ρw gE is the wave energy per unit surface and E ′ is the turbulent energy.
Integrating (10.36) over depth and using eq. (10.34) for S rad , we get
∂Ea ∂
Uα Ea + Fα + ρUα gh2 + u rad


+ bβ Sαβ
∂t ∂xα
Z
ζ ∂ζ
= −[wp + τi3 ui ]−h + (uα p + τiα ui ) + ϵdz, (10.38)
∂xα
in which the energy flux Fα is
Z ζ  
1
Fα = uα ρw u′2
i + ρ w g(z − h) + p + τi α dz. (10.39)
−h 2

10.3 Energy exchange and wave action

We can then subtract the mean flow energy [Uα × (10.33)+(gh − Uα2 /2)×(10.35) ] from (10.38) to obtain
the wave energy evolution equation7 ,
∂E ∂ rad ∂b
uβ
ρw g + [ρw gE (b
uα + Cgα )] + Sαβ = ϕaw − ϕoc − ϕbf , (10.40)
∂t ∂xα ∂xα
where ϕaw = − wp + τi3 ′ u′ |
i z=ζ is the mean wind to wave energy flux per unit surface, ϕoc is the part of
−ϵ associated to wave breaking and dissipation in the water column, and ϕbf is the wave energy lost
through bottom friction, which will be further discussed in chapter 18.
rad
The term of particular interest to us is the gain or loss of wave energy Sαβ ∂b
uβ /∂xα , when they
propagate through current gradients, even without dissipative processes. This is indeed an exchange of
energy between waves and currents.
To simplify this term, one can introduce the wave action A = ρw gE/σ (Bretherton and Garrett,
1968; Andrews and McIntyre, 1978), and using the evolution equation for the wave number, we obtain
(see Phillips, 1977, for details)
∂A ∂ ϕaw − ϕoc − ϕbf
+ [A (b
uα + Cgα )] = . (10.41)
∂t ∂xα σ
The action like the energy and the momentum is a quadratic quantity. It can be decomposed into
a spectrum N (k, θ, ϕ, λ, t) = E (k, θ, ϕ, λ, t) /σ where ϕ and λ define the horizontal position, here the
longitude and latitude. The spectral wave action equation is (Komen et al., 1994)
∂ ∂   ∂   ∂   ∂   S
N+ ϕ̇N + λ̇N + k̇N + θ̇N = , (10.42)
∂t ∂ϕ ∂λ ∂k ∂θ σ
where S is the sum of the source terms that represent the interactions with winds, bottom, wave-wave
interactions and dissipation, and θ is the wave propagation azimuth, the angle between North and the
wave propagation direction. The propagation speeds in physical and spectral space are given by Tolman
(1990),
ϕ̇ = (Cg cos θ + vb) R−1 (10.43)
−1
λ̇ = (Cg sin θ + u
b) (R cos ϕ) (10.44)
∂ω cos θ ∂ω −1
θ̇ = Cg sin θ tan ϕR−1 + sin θ − (kR) (10.45)
∂ϕ cos ϕ ∂λ
∂σ k
k̇ = − · ∇D − k · ∇b u, (10.46)
∂D k
where R is the radius of the Earth, u b and vb are the zonal (towards the East) and meridional (towards
the North) components of the horizontal current u b.
In the absence of dissipative processes, S = 0 and the wave action is conserved. This is a general
physical result and the action is an adiabatic invariant. It is the same for a pendulum of varying length,
A = E/σ is conserved, not E. More exactly, the total wave action, integral of A(k, θ) over the spectrum,
is conserved for adiabatic processes that are independent of the wave phases. This is Noether’s theorem.
As we have seen above in section 10.1.2, the conservation of wave action can be used to investigate
the amplification of waves across a current convergence.
7 Here it is assumed that ub is a vertically-uniform current. In a more general case u
b is replaced by uA as defined by eq.
(2.56). There are still discussions about the interpretation of this, but it was derived by Andrews and McIntyre (1978).
114 CHAPTER 10. WAVE-CURRENT INTERACTIONS
Chapter 11

Interactions of waves and sea ice

Ice in the ocean is part of the ocean. In the case of sea ice it comes from sea water that freezes as it
is cooled by a cold atmosphere, with some addition of snow accumulating on the top. In the case of
icebergs, it generally comes from glaciers that feeds the ocean with pieces of ice that is compacted snow
that has fallen on continents, and contains no salt.
The interaction of waves and ice is a subject that is very complex. It is made more difficult by the
paucity of available measurements. Fortunately, research efforts have been amplified since 2010, in an
area where much is still to be discovered. These efforts are motivated by the rapid evolution and poor
climate projections of the sea ice cover, in particular in the Arctic. Sea ice has an important role in
isolating the ocean from the atmosphere and solar radiation, effectively shutting off heat and gas fluxes
and increasing the ocean albedo. As a result, the effects of waves, pushing the ice or enhancing ice
formation and melting, are important for understanding the ice edge dynamics and air-sea interaction
from weather forecasting to climate projections. Another reason for studying waves in ice, is that the
emerging Arctic ocean is a place of increased human activities, with new shipping routes and increasing
exploitation of natural resources. As a result, any activity there requires the development of wave
forecasting capabilities, in particular around the ice edge.
This chapter discusses several key processes of wave-ice interactions, starting with sea ice and finishing
with icebergs which is made special by its very large thickness. In all of these processes , one obvious
aspect is that ice is a solid that floats, but a solid that can take many shapes and forms. Ice also deforms,
and can break into pieces under the strain caused by ocean waves. Although we start with an account
of frazil and pancake ice, it should be remembered that these are probably not the most common ice
form. Still, it is estimated that 50% of the Antarctic ice (Gow et al., 1982), and less so in the Arctic, is
initially grown as frazil and pancakes. Hence, this early stage of development is a very important one.
Without going in too many details about the physical properties of ice and its consequences for
sea ice (see Weeks, 2010), it is important to note that sea water freezing results in the formation of
cristals of pure ice that then confine salt to brine pockets and a few other cristals involving, among other
carbonates. Also, the freezing of water with salinity above 25 PSU produces sea ice with low salinity
(typically 10–20 PSU, made up of ice cristals and brine) and increases the salinity of the surrounding
water. That more saline water is denser and will thus lead to some convection in the upper ocean mixed
layer. If the original water is brackish (defined here with a salinity under 25 PSU) then the more salty
water is less dense, and will be stably floating on less salty water, so that ice formation can develop in a
very thin surface layer (Weeks, 2010, p. 48). Here we will not discuss these brackish conditions that are
specific to large Arctic estuaries.
When the water is very calm it freezes from the surface as columnar ice crystals, forming large slabs
of congelation ice. The wave-induced perturbation facilitates the nucleation of ice cristals that grow into
a suspension of small platelets of ice, known as frazil ice. Due to their buoyancy, these frazil cristals
concentrate at the sea surface, like a snowstorm flipped upside down.

115
116 CHAPTER 11. INTERACTIONS OF WAVES AND SEA ICE

Figure 11.1: From frazil to pancakes.

(a) vertical plane view of frazil cristals viewed with polarized light, and their aggregation (copyright
Elsevier, taken from McFarlane et al. (1982). (b) accumulation of frazil in a surface layer under waves.
(c) Formation of an ice edge in frazil ice (From Martin and Kauffman, 1981, , copyright Cambridge
University Press). (d) Field of pancakes in a storm as recorded by E. Rogers on October 12, 2015 in the
Beaufort sea (Rogers et al., 2016). (e) Ice floes resulting from the agregation of pancakes, Beaufort sea,
November 1, 2015, picture by H. Shen.

11.1 From water to pancakes

11.1.1 Suspensions of frazil cristals and effective viscosity
Knowledge on frazil properties mostly come from laboratory experiments, such as those performed by
Martin and Kauffman (1981), in which sea water in a wave tank is cooled. As sea water approaches the
freezing temperature, around -2◦ C for a salinity of 35 PSS, disk-shaped ice cristals form with diameters
1–3 mm and thickness of 1 to 10 micrometers, as shown in figure 11.1.a. The volume concentration ϕ
of this suspension may vary in the range 0.15 < ϕ < 0.45 depending on the confinement of the water.
In particular, the dissipation of waves by the ice induces a compression force due to the convergence of
the wave momentum flux. This wave-induced force tends to increase the concentration at the ice edge
exposed to the waves.
Such a concentration of solid particles enhances the effective kinematic viscosity ν1 of the suspension
compared to that of the fluid component only which, at the freezing point is νw ≃ 1.83 × 10−6 m2 s−1 .
The ratio ν1 /νw for a concentrated solution of solid glass spheres is expected to vary like (Mooney, 1951)
ν1
= exp [2.5ϕ/(1 − 1.43ϕ)] . (11.1)
νw

This equation gives a factor 2 at ϕ = 0.2 and a factor 23 at ϕ = 0.45.

Representing the ocean as a two layer system with different viscosities, ν1 for the top layer of thick-
ness h1 , and νw for the underlying and deep layer, wave attenuation is exponential. A first approximate
solution was given by Weber (1987) with a very large viscosity in the surface layer gave a dissipation
dominated by the viscosity of the underlying water, independently of the top layer viscosity and thick-
ness, identical to the solution by Dore (1978) for the viscous friction at the air-sea interface. The general
11.2. DISPERSION RELATIONS 117

solution, including finite depth of the underlying layer is given by Keller (1998). The resulting dissi-
pation can be very strong for waves relatively short compared to the frazil layer, which explains the
strong attenuation of waves in figure 11.1.c. However, longer waves are not much attenuated once their
wavelength is much larger than h1 .
Now, the empirical coefficient 1.43 in Mooney’s eq. (11.1) is strongly dependent on the solid material
that is in suspension. Besides, the frazil cristals are not spherical and, for smaller wave amplitudes, these
disks tend to stick together (Martin and Kauffman, 1981).
The problem of flat ellipsoids in suspension, instead of spheres, was addressed by de Carolis et al.
(2005), who give an effective viscosity in the form of a power law
ν1
= (1 − ϕ)K . (11.2)
νw

Interpreting the wave attenuations measured by Martin and Kauffman (1981), de Carolis et al. (2005)
find that the exponent K is of the order of 15 to 20 to explain the effective viscosities in the range 0.002–
0.01 m2 s−1 . They conclude that such high values can only be explained by a dominance of solid-solid
interactions, including collisions, friction and some sticking, and these all very sensitive to strain rate.
It is thus expected that the effective viscosity is generally a non-linear function of the strain rate and
volume concentration.
As pointed out by Martin and Kauffman (1981), the thickness and concentration of grease ice is
modulated by wave-generated roll structures known as Langmuir circulations (see chapter 14). The ice
collects in the surface convergence zone of these rolls. Garrett (1976) had envisaged that a localized wave
dissipation in the convergence zone, in his case due to wave breaking over the stronger current, could
reinforce the rolls. The stronger dissipation due to thicker grease ice could play the same role. These
effects may be relevant for the enhancement of air-sea fluxes in leads (Esau, 2007).

11.1.2 Pancake ice

Natural grease does not remain a suspension of ice fragments of similar sizes and shapes. Instead, grease
ice evolves to the formation of pancakes. These are agglomerations of frazil that weld together. Once a
pancake has started forming, it can grow by the washing of frazil ice over the top of the pancake and its
subsequent freezing (Doble et al., 2003). The raised rim of the pancakes, clearly visible as a white skirt
in figure 11.1.c, come from their mutual collisions. Pancakes typically grow to diameters of the order
of 1 m, and the ocean can be covered by a few layers of a pancakes. Shen et al. (2001) discussed that
the maximum diameter of the pancakes observed in the laboratory is generally of the order of 1% of the
the dominant wavelength, possibly due to break-up by bending or stretching if the pancakes get larger.
This scaling of the pancake diameter with the wavelength was also confirmed by Roach et al. (2018) with
video acquired in the Beaufort Sea, but they could not conclude if bending or stretching was the main
mechanism that limits the size of pancakes.
Pancakes have thicknesses that rarely exceed a few centimeters, so that they easily raft and form a an
ice layer with several pancakes irregularly stacked. The dissipation of energy in this ice layer is certainly
dominated by ice-ice friction and collisions. Even though the local surface motions is consistent with
orbital wave motion without ice (the ratio of horizontal to vertical motions recorded by buoys is very close
to 1), there is a very strong dissipation of short wave components. Several models have been proposed
to represent this dissipation, ranging from the viscous layer model of Keller (1998), to more complex
visco-elastic models (Wang and Shen, 2010; Rogers et al., 2016). These latter models include different
modes of motion and Mosig et al. (2015) show that some of these are often similar to what is predicted
by a viscous thin beam model. Unfortunately, all these models are highly empirical, and there is no
method for determining a priori the viscosity and elasticity parameters from large scale environmental
conditions.

11.2 Dispersion relations

Whereas the condition of zero pressure at the free surface and the horizontal momentum balance gives
the usual ice-free dispersion relation σ 2 = gk tanh(kD), the presence of ice clearly modifies these two
conditions and thus the dispersion relation. A first effect, when ice is not broken in floes shorter than
the wavelength, is the resistance to stretching of the sea surface. This same effect is similar to the effect
of surface tension, and thus also leads to an increase of the phase speed for the shorter waves. One
118 CHAPTER 11. INTERACTIONS OF WAVES AND SEA ICE

parameterization of that effect uses a flexural rigidity L, and gives the dispersion relation

Lk 5 = ρw (g − σ 2 hi ρi /ρw )k − ρw σ 2 (11.3)

Otherwise, if the ice sheet is broken up in pieces, there is only the effect of the added ice mass, as
given by Fox and Haskell (2001), in which case the phase speed is decreased by the presence of ice,
σ2
k= . (11.4)
g − σ 2 hi ρi /ρw
There are many more complex theories for the dispersion relation (e.g. Meylan et al., 2018) that
can give faster of slower phase speeds depending on conditions. The observations generally find faster
propagation for short waves over thick ice, compared to waves without ice, for example for waves periods
shorter than 12 s in 1.5 m thick ice (Marsan et al., 2012). In the case of 0.3 m thick floes, Fox and
Haskell (2001) instead found a slower phase speed (compared to waves without ice) for periods in the
range 6 to 8 s (they did not measure shorter waves).

11.3 Dissipation for solid floes: basal friction

When freezing persists, pancakes weld together, leading to a single and continuous layer of ice. The
agitation due to waves that has led to the formation of pancakes results in a thicker layer of ice compared
to freezing conditions without agitation that produces columnar ice in which the heat is lost to the
atmosphere by diffusion through the ice layer.
With such a single ice layer heaving up and down with the wave motion but constrained in its
horizontal displacement, the dissipation energy can occur in the boundary layer below the ice. This
oscillatory boundary layer is similar to the air-side boundary layer discussed in Part 3, and the bottom
boundary layer discussed in chapter 18. For the case of ice, the problem was treated by Liu and Mollo-
Christensen (1988), with the additional effect of ice inertia that we neglect here. Here we consider the
finite viscosity of seawater at the freezing point, νw ≃ 1.83 × 10−6 m2 s−1 . For monochromatic waves of
√
radian frequency σ, a boundary layer of thickness νw σ develops in the laminar case, and the attenuation
rate of the wave energy in time is p
βice,bfr,v = −k νw σ/2. (11.5)
This is the same dissipation rate that was found by Phillips (1977) for waves under an inextensible layer
of oil (see also Weber, 1987). For a smooth under-ice surface, if we accept the similarity with the bottom
boundary layer, this laminar regime is expected to occur for Reynolds numbers Re = σa2 /ν < Rec , with
Rec = 1.5 × 105 (Jensen et al., 1989). Above this value the boundary layer transitions into a turbulent
regime. In that case the dissipation rate becomes quadratic,

βice,bfr,t = −fe uorb /g, (11.6)

where uorb is is the significant orbital velocity amplitude at the water-ice interface. The dissipation
factor fe is obtained from the under-ice roughness in the same way that it was obtained in the turbulent
bottom boundary layer in chapter 18.
For random waves, neglecting the effects of the ice layer on the water motion, we use the solution
given in chapter 2, sZ
∞
σ2
uorb = 2 E(f )df , (11.7)
0 tanh2 (kD)
with a significant horizontal displacement
sZ
∞
1
aorb = 2 2 E(f )df , (11.8)
0 tanh (kD)

and we use this definition of the Reynolds number, Re = uorb aorb /ν.
Because the superposition of linear waves gives a Rayleigh distribution of the amplitudes, the tran-
sition between laminar and turbulent can happen only for the highest waves, giving a smooth transition
of the average dissipation rate βc . In practive, considering a Rayleigh distribution gives a value of βc
that is very close to the following approximation

βc = (1 − w)βice,bfr,v + wβice,bfr,t , (11.9)

11.4. DISSIPATION FOR SOLID FLOES: FLEXURE AND THE IMPORTANCE OF ICE BREAK-UP119

ice compression ice streching

boundary
layer
velocity profile
under ice

6
10
Half−decay distance

5
10

Viscous
4 Turbulent
10
Combined rayleigh
Combined smooth

5 6 7 8 9 10 11 12 13 14 16 18 20
Wave period (s)

Figure 11.2: Top: schematic of velocities and boundary layer below an ice layer (light blue). Bottom:
expected decay distance X1/2 = Cg ln(2)/βc as a function of wave period T , for an under-ice roughness of
0.1 mm, and wave heights ranging from 0.5 to 5 m. The ‘combined Rayleigh’ solution uses a combination
of βice,bfr,v and βice,bfr,t with a weighting depending on a Rayleigh distribution of wave amplitudes, and
the ‘combined smooth’ solution is given by eq. (11.9). Adapted from Stopa et al. (2016c).

as shown in figure 11.2, where we have used a weight w = 0.5 [1 + tanh((Re − Rec )/∆Re ], and ∆Re =
2 × 105 .
It should be noted that even the laminar dissipation is relatively strong, with half-decay distances
under 400 km for periods shorter than 10 s, and 30 km under 5 s. This viscous dissipation is larger
than what was estimated from buoys in pancake ice in recent experiments (Ardhuin et al., 2018). Rogers
et al. (2016) reported imaginary wavenumbers ki = 2 ln(2)/X1/2 of the order of 2 × 10−5 for a period
T = 5 s, which corresponds to X1/2 = 140 km, 4 times larger than due to molecular viscosity under an
inextensible surface ice sheet. This is possible because the pancakes actually move horizontally and raft
to form multiple layers, so that the present theory for βc gives an overestimate of the wave dissipation in
pancake ice. In the case of densely packed thick floes, rafting does not occur and the horizontal ice motion
is much weaker than the vertical motion (Fox and Haskell, 2001), in which case the above expression for
βc should be applicable.

11.4 Dissipation for solid floes: flexure and the importance of

ice break-up
In other cases, under-ice friction appears insufficient to explain the observed attenuation. This is par-
ticularly the case for the large-scale attenuation of swells with periods 18 to 33 s reported by Ardhuin
et al. (2016), and illustrated in figure 11.3. In that study of waves measured in the pack, 1500 km
120 CHAPTER 11. INTERACTIONS OF WAVES AND SEA ICE

from the ice edge, the constant dissipation rate needed to obtain a reasonable agreement with the data
corresponds to a spatial decay of the energy, α = β/Cg , ranging from α ≃ 4 × 10−6 m−1 at T = 20 s, to
α ≃ 2 × 10−6 m−1 at T = 25 s. These are 12 times the effect of viscous friction below a smooth ice plate
as given by eq. (11.5). That seems impossible to explain by the ice morphology alone, with complicated
keel structures, often exceeding 10 m below the mean ice level (Doble et al., 2011).
relative Days (February 2007)
energy (dB) 11 12 13 14 15 16

25 30 35 40

Swell height in cm ( T > 23 s)

−20
14 wave period
−30

−40 17 T (seconds) 3500

−50
20
1

l2

−60
l

swell 1
el

ice
el

25 3000
sw

−70
sw

North edge
−80
data 33 pole
−90

10 15 20
2500
40 85°N

y (km)
30
2000
20 80°N swell 2

5
10
Greenland

0
2500 3000 3500 4000 4500 5000
0 10 11 12 13 14 15 16 17 x (km)

Figure 11.3: Left: Spectrogram of surface elevation and width σ2 of the directional wave spectrum
estimated from second moments a2 and b2 . The black rectangle highlights the time and frequency range
of the spectral peak on February 12 at 18:00 UTC, showing that the directional spreading is around 10◦
or less when the energy is maximum. Right: model snapshot giving some context to the measured data.
The two swell events are due to two distinct storms. Adapted from Ardhuin et al. (2016).

Also, dissipation by basal friction leads to unrealistic high values of swell heights, of a few centimeters,
crossing the Arctic from Fram Strait to Alaska (Ardhuin et al., 2016). For this reason, the dissipation of
energy associated to the deformation of the ice was studied by Ardhuin et al. (2018); Boutin et al. (2018);
Ardhuin et al. (2020). Several dissipation effects have been studied in the laboratory and in the field,
they include anelastic effects Cole (1995); Cole et al. (1998) and inelastic effects, related to dislocations
at large strain rates (Cole and Durell, 2001). These effects are consistent with a strong reduction in wave
attenuation as soon as the ice is broken up (Ardhuin et al., 2020), and may explain the very wide range
of attenuations found in remote sensing data across Antarctic sea ice by Stopa et al. (2018).

11.5 Scattering of waves by sea ice

Alternatively, following early experiments in the the Arctic (Wadhams et al., 1986), it was proposed
that the multiple reflections at the edges of ice floes, or at sharp gradients in ice thickness, could be a
significant of source of attenuation, similar to the Bragg scattering of waves by bottom topography that
is described in chapter 16. The ice-induced ”scattering” has been the subject of many theoretical works
(Squire, 2020). When scattering is the dominant source of wave attenuation in the sea ice it should be
associated with a strong broadening of the wave directional spectrum, with a wave spectrum becoming
nearly isotropic near the ice edge. There is, however, very little support for a general broadening of wave
spectra in sea ice. Sutherland and Gascard (2016) found a minor increase in directional spread off the
East coast of Greenland, but that is not sufficient to explain most of the wave amplitude attenuation.
Over the scale of the Arctic basin, the example shown in Fig. 11.3 clearly disproves the idea that
there is a significant scattering of swells by under-ice topography, which was proposed by Squire et al.
(2009). In the marginal ice zone, all remote sensing data, except in cases of crossing seas (Ardhuin et al.,
2015), show long-crested swells that are not compatible with a strong scattering.

11.6 Waves and icebergs

Either at large ice shelves around Antarctica or from smaller glaciers around Greenland, icebergs are
formed by the calving of ice streams reaching the oceans. Large icebergs are found in the Southern
11.6. WAVES AND ICEBERGS 121

10 km

satellite
line of sight

ve
ea n wa
m
ct ion
di r e

Figure 11.4: Example of a synthetic aperture image acquired by Envisat’s ASAR instrument at 53◦ S
155◦ W on December 26, 2008, 9:35 UTC. White areas are the sides of icebergs facing the satellite. Waves
propagate from the West (left of the image, see arrow) and a majority of icebergs align their longer sides
in the North-South direction. The dark areas to the right of bright spots are either the body of iceberg
in or the sea surface in the radar shadow behind the iceberg (narrow strips next to the iceberg), or areas
with low wind and/or greasy ice for which the backscatter is very low. A few icebergs are almost square
tabular bergs with their top seen in an intermediate grey level between the very bright illuminated sides
and the darker modulated sea. The “short” modulations (with a 270 m wavelength) in the grey level
throughout the image are the ocean waves. Only 25% of the full image is shown.
122 CHAPTER 11. INTERACTIONS OF WAVES AND SEA ICE

Figure 11.5: Left: climatological mean of the fraction of ocean area covered by icebergs. Right: maximum
difference, in meters, over the year 2008, between modeled wave heights without (Hs1 ) and with (Hs2 )
icebergs parameterized using the Ifremer database. Reproduced from Ardhuin et al. (2011).

Ocean, with the largest ever observed having the same area as Belgium. Recent large calving events
from the Larsen ice shelf have attracted headlines, but there are thousands of icebergs a few kilometers
in diameter (Tournadre et al., 2016) that deserve attention because they are the ones that give the
largest ”cross-section” for incoming waves, and are responsible for very strong wave attenuation in the
Southern Ocean (Ardhuin et al., 2011). Indeed, these floatings islands of ice act like giant break-waters.
At present we only have a statistical knowledge of their presence, with a monthly database of their
occurence developed by Ifremer (Tournadre et al., 2016).
As a result, the average estimated impact of these icebergs can be a reduction in wave height that
exceeds 0.5 m over vast regions of the Southern Ocean and is highly variable in time as clusters of ”small”
icebergs are caused by the break-up of the largest calved icebergs. In particular, the anomalously large
number of small icebergs in the South Pacific in 2008 is related to the breakup of iceberg C19a (Tournadre
et al., 2012).
Chapter 12

Numerical modeling in deep water

Numerical wave models are used for a wide variety of applications. These include navigation safety, ocean
engineering for marine energy (oil and gas or renewables) or ship design, coastal engineering. Wave models
are used both for forecasts and hindcasts, for deterministic simulations or ensemble predictions.
Because waves also have impacts on the atmosphere, ocean, sea ice, or sediments, wave models are
increasingly used in Earth System models, coupled with other dynamical models. This coupling is most
important in shallow waters (storm surges, wave-driven currents ...) and specific aspects of coastal and
shallow water wave modelling will be discussed in chapter 20. Our purpose here is to give indications
about the validity and limitations of wave models. As we focus on deep water, it removes the specific
issues of bottom friction, refraction, shoaling, making the question of wave evolution more simple. The
first question to consider is: what is deep water? From chapter 2, the ratio of the wavelength and water
depth represented by kD, is one criteria. Typically when kD > 2 the effect of shoaling and refraction
can usually be neglected.
This chapter only deals with spectral phase-averaged models, based on the wave action equation
(10.42), following the first numerical model of Gelci et al. (1957). The practical problem is to obtain the
most accurate result, either for a forecast for the coming days a climate projection over long terms, or a
restrospective simulation (hindcast), using limited computational resources and time. This leads to trade-
offs between expensive calculations, for example the 4-wave interactions, and cheaper parametrizations.
The choice of the numerical method also has a strong impact on the model cost, with benefits that may
be visible only for some parameters.
What are the main factors that control the model accuracy?

ˆ Because waves are generated by winds, and propagate in a medium characterized by a bottom
topography, currents and obstacles (small islands, sea ice ...), the quality of these forcing fields is
determinant, and the accuracy of winds is certainly the most important.

ˆ The second most important factor is the accuracy and behavior of the parametrizations use for all
the processes represented in the source terms S in eq. (10.42). These parameterizations should
be robust, meaning that they should work under all circumstances from calm seas to hurricane-
force winds, in the presence or absence of swells ... Many model errors can be traced to poor
parametrizations. No parametrizations is perfect but some definitely produce more accurate results
than others.

ˆ Finally, there is no method to solve eq. (10.42) to the accuracy of the computer round-off error
at an acceptable computational cost. The numerical integration methods are thus based on ap-
proximations that lead to numerical diffusion. Also, numerical limiters in the evolution are used
by many models, so that the numerical result may be very different from the actual solution of the
wave action equation (e.g. Tolman, 2002b). One infamous example is the effect of semi-implicit in-
tegration of the source terms in time when large time steps are used (Hargreaves and Annan, 2000).
This particular issue motivated the use of an adaptative time step for source term integration in
the WAVEWATCH III model (Tolman et al., 2014).

The overall model quality is thus the result of many choices. As a user, you should be very careful
about supposedly ‘default’ model settings: this means that somebody has chosen for you, however
reasonable that choice may be in usual conditions, it may be not be the best choice for your particular
application. As a user of a wave model you may not have the freedom to choose, and the choice of model

123
124 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

or forcing may be imposed because it is more or less easily available and easy to use. Whatever the
circumstances, you should be aware of the impact on the model results. Also you may end up frustrated
by model result that exhibit unphysical behavior: such as the stronger growth of wind sea in the presence
of swell or a broadening of the directional wave spectrum in shallow water. You may thus want to try
yourself and improve on the existing models. Hopefully the next sections will be interesting advice. As
for the swell effect of wind growth, we have not get a full physical model that would give the proper
reduction in wind sea growth (Garcı́a-Nava et al., 2012), but at least, going from the parameterization
by Janssen et al. (1994) to the one by Ardhuin et al. (2010) you will not see anymore the wind sea wave
height jump up when swell arrives.
As mentioned, the most important choices that define the model accuracy are the model forcing,
model parametrizations, and numerical schemes. Models may also assimilate measurements to correct
past estimates of the waves. However, contrary to atmospheric or oceanic circulation models, assimilation
is not necessary to obtain accurate forecasts and hindcasts. For all these choices, the most accurate result
is not necessarily obtained with the most complex choice, because some errors are often compensated by
other errors.
Finally, models are usually good only for the parameters for which they have been validated, and for
the ranges of these parameters in which we have data. Remember that there are very few measurements
of the full frequency-direction spectrum, so that the modeled spectral shapes can be really bad, especially
at high frequencies. When one predicts that the 100-year significant wave height off the west coast of
France is 18 m, based on a model that has only been validated up to 14 m, this is a real extrapolation.

12.1 Forcing
12.1.1 Bathymetry and islands
If you have never run a numerical wave model, it probably sounds trivial that the ocean depths and
shape of the shoreline should be known. In practice, though, it is not always easy. Global databases such
as ETOPO (Sloss, 1993) and GSHHS (Wessel and Smith, 1996) can have important errors, with large
error in water depth over continental shelves, and some islands misplaced by a few kilometers. Even
at global scales and coarse resolutions, it is important to take into account subgrid islands such as the
Tuamotus or Aleoutian islands in the Pacific (Tolman, 2003), as illustrated in figure 12.1.

12.1.2 Wind fields: analyses, forecasts, re-analyses

Although there are many measurements of surface winds based on satellite radiometers or scatterometers,
the coverage they provide is not enough to use a forcing field defined from these data only. Hence it is
most common to use winds from an atmospheric circulation model. In these models, the near-surface
wind speed, usually 10 meters, is not a an dynamical parameter because these models usually have a
vertical discretization in pressure levels that does not follow a constant height. This surface wind, possibly
corrected for stability effects in the form of a neutral wind, is in fact diagnosed from the atmospheric
boundary layer parameterization. As a result there are significant differences between models. For
example, for wind speeds exceeding 30 m/s, the winds provided by the European Center (ECMWF)
as part of their operational analyses were typically 10% slower than those of the U.S. weather service
(NCEP) for the year 2005. Given that there are very few reliable data at high wind speeds, it is difficult
to decide which is more accurate correct. As a result, the wave model parameterizations are adjusted
differently to the different forcing fields.
If you are a scientist, you will probably try to get the most accurate wind fields. If you are an
engineer, you may be in a rush or may not have access to the most accurate database. When it comes
to wind vectors, the most accurate model analyses and forecasts over the recent years and most of the
globe, are those provided by the European Center for Medium Range Weather Forecasting (ECMWF).
Although their data policy may soon change, these are usually not freely available. The anlysis is the
combination of model and observation that is expected to produce the most accurate estimate of the
state of the atmosphere, providing the initial conditions from which the forecast in produced.
ECMWF, like most Numerical Weather Prediction (NWP) organizations, produces both deterministic
analyses and forecasts, a single simulation, and ensemble predictions, which consist of a set of simulations
that is designed to explore the possible variations of the atmospheric state. ECMWF spatial resolution
of the deterministic product was reduced to 0.25 degree in 2008, 0.125 degree in 2011, and 9 km in March
 12.1. FORCING 125

average Hs from altimeters (m)

0 0.2 0.4 0.6 0.8 1.0 1.2
normalized wave height

Figure 12.1: Left picture: example of wave attenuation due to propagation across many islands, here
the Tuamotus, for a case of swell from the North-East, which almost never happens in this region,
as computed on a 3.6 km grid. These islands would not be resolved in a typical ocean global grid
(reproduced from Chawla and Tolman, 2008) and require a sub-grid parameterization as introduced by
Tolman (2003). The dashed box corresponds to the extent of the right panel which shows the mean
values of Hs recorded by TOPEX and Jason (reproduced from Andréfouët et al. (2012), showing the
very strong sheltering in real conditions with both windsea and swell, with swell usually coming from
the South-West.

1997 to 1998 1999 to 2000 2001 to 2002

4.5 1.25 2003 to 2004 2005 to 2006 2007 to 2008
Wind Speed RMSE (m/s)

2009 to 2010 2011 to 2012 2013 to 2014

4 2015 to 2016
1
3.5
Hs RMSE (m)

3
0.75
2.5

2 0.5

1.5
0.25
1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Forecast range (days) Forecast range (days)

Figure 12.2: Root mean square error (RMSE) as a function of the forecast range, 0 corresponding to the
analysis, in the operational systems ran at ECMWF. These modeled wind speed and wave heights are
compared to buoy measurements over a year centered in winder (August to July). Only every other year
is shown (Picture courtesy of J. Bidlot).

2016. Their system, like those at all NWP centers is constantly improving to take the best advantage of
computer power, new observations, new model parametrizations and assimilation methods.
You may find details about ECMWF’s Integrated Forecasting System (IFS) on their website with a
nice quarterly newsletter. It can be a good idea to check on the quality of other sources by looking at the
wind and wave model verification http://www.jcomm-services.org/Wave-Forecast-Verification-Project.html,
presented by Bidlot (2008).
Wind analyses and deterministic forecasts have made enormous progress over the last 20 years (e.g.
Janssen, 2008), thanks to an increase in computational power that allows higher resolution, more complex
parameterizations and sophisticated data assimilation. This has gone with a dramatic increase in number
and quality of observations. Figure 12.2 shows the reduction in errors in both analyses and forecasts for
winds speeds and wave heights in ECMWF operational deterministic forecasting system.
Because of this continuous improvement of model quality, a time series of operational hindcasts is
not homogeneous in time. Hence, a statistical analysis of extreme events based on operational analysis
will probably have spurious trends. In order to reproduce past events as well as possible, recent versions
of the atmospheric models have been re-run to produce re-analyses. In this case the only source of non-
homogeneities are the observations assimilated. It is indeed not possible to go back in time and add a
few satellites. For this reasons, some atmospheric reanalyzes start in 1978, when enough satellite data is
126 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

Figure 12.3: Average wind and wave height in the Northern Hemisphere, Southern Hemisphere and
Mediterranean sea, obtained in different runs of the ECMWF model in which the spatial resolution is
changed. The resolutions of Gaussian grids T106, T213, T319, T511, T639, et T799 correspond to 188,
94, 63, 39, 31, and 25 km. Wind speed and wave height values have been divided by the values obtained
for the coarsest of these grids. Taken from (Cavaleri and Bertotti, 2006, ©Elsevier).

available for a reasonable estimate of the atmospheric circulation. This is the case of the Climate Forecast
System Reanalysis (CFSR Saha et al., 2010) and ERA-Interim (Dee et al., 2011). Other re-analyses go
back to 1958 (Kobayashi et al., 2015), and Compo et al. (2011) have produced a re-analysis from 1871
to 2008.
As we have seen, the highest waves require very high winds, and the quality of re-analyses for
estimating past wave heights critically depends on the capability of the atmospheric models to re-
solve the gradients in small storms, in particular for tropical storms or polar lows. As a result, some
wave re-analyses like ERA40 are notorious for understimating wave heights because of underestimated
winds (Caires and Sterl, 2005). Unfortunately for that aspect, the more recent fitfh generation re-
analysis (ERA5) at ECMWF favored an ensemble approach, rather than single and higher resolution
model (Hersbach et al., 2020). The ERA-5 winds are likely biased low for wind speeds above 21 m/s
(Pineau-Guillou et al., 2018). A monthly report on wind and wave model accuracy can be found here
https://confluence.ecmwf.int/display/WLW .

12.1.3 Coastal winds

In coastal regions, closed or semi-closed basins surrounded by mountains, such as the northern Mediter-
ranean, the winds from global models are often inaccurate due to a lack of resolution of the land topog-
raphy that channels winds. A striking view of this is in figure 12.3 (Cavaleri and Bertotti, 2006). In this
case, a solution can be to use higher resolution local area models. However, it must be noted that the
quality of the nested grid result, owes much to the boundary conditions and the quality of the parent
model. Sometimes a statistical correction of coarse grid can be a better option (Signell et al., 2005).

12.1.4 Observed winds

Winds over the oceans are measured by altimeters, scatterometers (or their higher resolution cousins,
SARs) and radiometers with increasing coverage and accuracy. It is now possible to have data everywhere
over the oceans at a resolution of about 12 km, and every 12 hours. Although 12 hours is long compared
to the time scale of evolution of the wind fields and we cannot use this data as such to force a wave
model, it is possible to blend this with atmospheric models to improve the realism of modelled winds
(e.g. Bentamy et al., 2007).
Although some of that data is assmilated into atmospheric models, these models often reject the
highest winds that are often too far from their first guess. As a result, it is always a good thing to check
12.2. NUMERICS 127

the consistency of modeled winds against data. Although traditional scatterometers that use Ku or C
are not able to discriminate wind speeds beyond 25 m/s, the use the measured Doppler shift (Mouche
et al., 2012) or cross-polarizations (Vachon and Wolfe, 2011; Mouche and Chapron, 2015), has been
demonstrated with SAR data to be able to better constrain higher wind speeds, and will be used in
future scatterometers. The recent use of L-band, which is more sensitive to longer waves than Ku or C,
has also been shown to be sensitive to the highest wind speeds (Reul et al., 2006).

12.1.5 Currents
The main issue with ocean currents is to get accurate estimations of current fields, at the relevant resolu-
tion (Ardhuin et al., 2017b). This is relatively simple for tide-dominated environment for which numerical
models are reasonably goog (e.g. Ardhuin et al., 2012a), it is much more delicate for regions dominated
by quasi-geostrophic dynamics, where deterministic ocean circulation forecasts have limited skills, and
even more so where internal waves are the main source of surface current gradients (Osborne and Burch,
1980) due to the required high resolution to resolve these features. As a result, most operational wave
forecasting systems take into account tidal currents, at best.

12.1.6 Sea ice

Wave-ice interaction is the topic of chapter 11. To summarize it here, the presence of ice has a strong
impact on wave growth and disspation, and a knowledge of ice concentration is required in any ocean
basin that has some ice, the Arctic of course, but all three oceans (Atlantic, Indian and Pacific) are
strongly influenced by ice, at least in their southern hemisphere part, and many regional seas an lakes
also freeze up: the Baltic, Caspian, the Great Lakes... Experimental evidence shows that waves are not
generated by the wind in ice-covered regions, and the ice can damp the waves very rapidly, in particular
the high frequency components.

12.2 Numerics
12.2.1 Wave models are big
The problem of numerical wave modelling at oceanic scales is often a question of compromise between
spectral and spatial resolution. Indeed, our wave action spectrum is a 4-dimensional beast that we have
to integrate in time. All models today use a fixed discretization into Nf ≃ 30 frequencies and Nθ ≃ 30
directions. The wave action equation 10.42 is thus a set of Nf × Nθ ≃ 1000 hyperbolic advection
equations that are coupled via the source term S and the refraction and shoaling terms. Each equation
for a component (fi , θj ) is a 2-dimensional partial differential equation (PDE) that represents advection
equation with source term. This is a fairly standard problem to solve, with the transport velocity Cg + U
varying in space and possibly in time. If the spatial discretization is a regular mesh with only 100 by 100
points, hence a spatial discretization into 10,000 nodes, the overal problem has 10 million unknowns: all
the values of our gigantic wave action matrix. As result, wave models typically use a lot of memory. For
example at ECMWF, with the IFS system, the wave model uses about as much as the entire atmosphere.
Now, every time I think of it, it makes me mad that most of these unknowns are zeros: waves over 25 s
period do not happen very often, but they do happen once in a while in the biggest storms (Hanafin et al.,
2012). One trick could be to adjust the spectral grid to remove the components for which we have zeros
on the entire grid or in a subgrid. At the same time, the recent extension of wave models to add shoreline
reflections (Ardhuin and Roland, 2012) and sources of low-frenquecy infragravity waves (Ardhuin et al.,
2014) have replaced these zeros by small numbers that can be significant for some applications. Still, if
you do not activate these options, there should be a way to make models faster by avoiding to compute
zero plus zero times zero equals zero.

12.2.2 Spectral discretization

In frequencies, one needs to choose
ˆ the spacing between frequencies: it is usualy exponential, with fi+1 = Xf r fi where Xf r is a
constant close to 1.1. The reason for this geometric progression is that it simplifies the calculation
of the 4-wave interaction term Snl : in deep water the geometric arrangement of the interacting
waves is the same from one frequency to the next. The choice of Xf r = 1.1 is relatively coarse, as
128 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

a JONSWAP spectrum will have only one or two frequencies inside of the peak. However, going to
Xf r = 1.05 will double the number of frequencies, if you used 30 it now jumps to 60, and double
the cost of the model.
ˆ a minimum frequency fmin : this should be low enough to allow waves to develop. In the open ocean,
it is possible to get a peak period of 25 s (Hanafin et al., 2012), and you probably need a couple of
frequency bands below that to allow the spectrum to roll back, this means that fmin = 0.037 Hz
may already be too high. Ardhuin et al. (2014) have also gone as low as 0.003 Hz to model the
infragravity part of the spectrum which will be discussed in chapter 17. However, adding these
frequencies means using a lager number Nf of frequencies, and also the computation cost for these
low frequencies can be higher due to numerical constraints: these long period waves have larger
group velocities, and explicit schemes will require a smaller time step.
ˆ a maximum frequency fmax , this is dictated by the lowest winds that you want to represent properly:
to catch the peak of the wind sea you need to have fmax > 1.2Xf r g/(2πU10 ). For a fairly frequent
wind speed of 5 m/s, this means that fmax ≃ 0.4 Hz is not enough. Still, this was used recently
in many global models. Also, if you want to investigate properties of high frequency waves or
backscatter for remote sensing, you probably want fmax > 0.7 Hz. Finally, most wave model ignore
viscous dissipation (Dore, 1978) and capillary effects. These cannot be ignored for fmax > 2 Hz.

North
Hs: 1.60m 07/05/2008 09:00 UTC
330
fp:0.0663 Hz
300 Mean dir. at fp: 213.99 deg
Spread at fp: 28.48 deg
West
U10: 9.84 m/s
240 Curr: 0.00 m/s
210

South

150
0.60 0.40 0.20

120

East

60

30

North
0.1 0.2 0.3 0.4 0.5 0.6 0.7

0.0 0.3 0.6 0.9 1.2 1.5 1.8 0 5 10 15 20 25 30

E(f, ) (m2 /Hz/rad) E(f, )/f (m 2/Hz 2 /rad)

Figure 12.4: Example of spectral discretization: here these modeled spectra use 32 frequencies from
0.037 Hz to 0.72 Hz with Xf r = 1.1, and 24 directions. The same spectrum is shown in cartesian and
polar coordinates. Note that the polar representation is nice to show the waves in the direction from
where they are arriving, but it squeezes all the dominant waves near the center, making it difficult to see
the details. Here the spectrum in polar coordinates is divided by the frequency f which compensates a
little for the squeezing: the total energy is still the area in the plot times the value plotted.

For the directions, with a directional spread that is about 20◦ for the wind sea peak, a resolution of 15◦
(i.e. 24 directions), can be satisfactory with 2 points in the peak – the spread is the half-width of the
spectrum in the limit of a narrow Gaussian shape. As wind-waves turn into swell, the spread can be as
low as 5◦ , and it can be useful to have a higher directional resolution. Again this comes at a cost: more
direction means larger spectra, more memory and computational time.

12.2.3 Spatial discretization

Given the cost induced by spectral discretization, the spatial discretization can be optimized in several
ways. First of all, one should consider the necessary resolution. In the absence of currents, the gradients
in the wave field are given by the wind field and the shoreline geometry. There are thus several options,
from a regular grid in latitude and longitude to a triangle-based mesh (Benoit et al., 1996; Roland and
Ardhuin, 2014, e.g.), or mosaics of grids with different resolutions (Tolman, 2008) or quad-trees (Popinet
et al., 2010; Li, 2010). In particular, high resolution is desirable in hurricanes, and moving or adaptive
grids have thus been designed for this (Tolman and Alves, 2005; Popinet et al., 2010). All these options
are available in the WAVEWATCH III modelling framework.
12.2. NUMERICS 129

12.2.4 Integration methods

There are two large classes of methods used for this type of multi-dimensional PDE.

ˆ a global resolution with an iterative method: a method of lines as described in e.g. Patankar
(1980). Here, all terms of the equation are discretized first and thereafter integrated in time using
an ODE solver. This method has been applied in the SWAN model Booij et al. (1999). Since the
equation is rather stiff a fully implicit scheme is commonly employed for the time integration. In
turn, due to nonlinear source terms, this is accompanied by an iteration process.

ˆ A splitting method following Yanenko (1971). In that case the several terms of the equation are
integrated in succession,

∂N
+ advection = 0 (12.1)
∂t
∂N
+ refraction = 0 (12.2)
∂t
∂N
= S/σ. (12.3)
∂t

In the limit of small time steps, this is equivalent to the full integration. This approach is used in
WAM (WAMDI Group, 1988), and for most options of WAVEWATCH III (Tolman, 2002d), and
in TOMAWAC the first two steps are combined (Benoit et al., 1996).

The first approach can be interesting at very high resolution, typically in the nearshore, when used
with implicit methods that are not constrained to have very small time steps. It is used in SWAN (Booij
et al., 1999) and has been implemented in a research version of WAVEWATCH III (Huchet et al., 2015).
The great benefit of the second approach, is that the splitting allows to adapt the time step to the
time scale of evolution of each term. This is fully used in WAVEWATCH III, with an adaptative time
step for the source term integration, and different time steps for the advection and refraction.
In order to limit the computational time to something acceptable, wave models use limiters: the
rate of change of evolution is limited to ensure stability while keeping large time steps. These limiters
can completely change the solution (Tolman, 2002b; Roland and Ardhuin, 2014), so that it is a good
practice to test that a model gives similar results when reducing all time steps by a factor 2. When
a splitting approach is used, different limiters can be applied to different pieces of the equation. For
example, WAVEWATCH III has a limiter on refraction (waves cannot turn by more than ∆θ = 2π/Nθ
over one refraction time step ∆t,r : if the bottom slope of current gradient require a faster rotation then
the rotation speed is limited to ∆θ /∆t,r . Limiters for source term integration are a bit more complex.

12.2.5 Source term integration

The source terms link all spectral components together, including high frequencies that can evolve very
fast. For that reason and also because the source term balance Sin + Snl + Sds is generally not well
reproduced at high frequencies by parameterizations (e.g. Zieger et al., 2015), the spectrum above a
diagnostic frequency fd is usually prescribed to decrease like f −5 . This cannot reproduce the observed
broadening of the directional spectrum with increased frequencies (e.g. Banner et al., 1989; Leckler et al.,
2015) but it is a shortcut to keep the models within reasonable bounds. In the ECWAM model fd is set
to be the maximum of 2.5 times the mean frequency of the wind sea and 3 times the Pierson-Moskowitz
frequency (Bidlot, 2012).
A stable integration of the source terms is possible with an implicit integration,

S(t)∆t
N (t + ∆t ) = N (t) + . (12.4)
σ(1 − α∆t ∂S/∂N/σ)

With α = 0.5 this is accurate to second order in time, but α = 1 generally performs better for the
short wave components (Hargreaves and Annan, 2000). Still, one should be careful that source term
integration is sensitive to the choice of time step it if is constant. As a result, even at frequencies f < fd ,
a limiter is used for spectral evolution. The design of this limiter is discussed in Tolman (2002b).
130 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

12.2.6 Spatial propagation

The early wave models such as WAM (WAMDI Group, 1988) have chosen the cheap but robust first
order ’upwind’ scheme. In only one dimension, the wave propagation direction is given by a sign s, going
towards x > 0 if s = 1 and in the opposite direciton if s = −1. The upwind advection scheme is thus
∆t
N (x, t + ∆t ) = N (x, t) − [Cg (x)N (x) − Cg (x − ∆x )N (x − ∆x , t)] if s = 1 (12.5)
∆x
∆t
N (x, t + ∆t ) = N (x, t) + [Cg (x + δx )N (x + δx ) − Cg (x)N (x, t)] if s = −1. (12.6)
∆x
The name upwind expresses the idea that the information is taken from where it is coming. That scheme
is stable when the Courant number Cg ∆t /∆x is less than 1. This condition is known as the CFL condition
after the work of Courant, Friedrichs and Lewy (Courant et al., 1928).

0..8 12
74°N, 13°E 78°N, 83°E
Hs (mm, T > 20 s)

10

Hs (cm, T > 20 s)
0..6
1st order upwind 8
0..4 3rd order 6
upwind QUICKEST
4
0.2
2
0..0 0
6 8 10 12 14 16 18 6 8 10 12 14 16 18
days (February 2007) days (February 2007)

Figure 12.5: Time series of modeled wave height (computed over wave periods longer than 20 s) at two
locations of the Arctic, using the first order upwind scheme, or the third order upwind QUICKEST
scheme - including the Garden Sprinkler Effect correction of Tolman (2002a).

The main defect of the first order upwind scheme is its large numerical diffusion: a localized packet
of energy tends to disperse in space. This dispersion effect is not always bad because real waves are
dispersive, but the physical dispersion is only a function of the spectral resolution. A much reduced
diffusion can be obtained by using higher order schemes, such as the third order upwind QUICKEST
scheme of Leonard (1991). This is obvious in figure 12.5, which shows time series of wave heights in the
ice-free part of the Arctic, in February 2007. The two biggest storm of the year in the North Atlantic
occured on February 9 and 10, sending swell to the Arctic. This swell is modeled to arrive on 13 and 14
February off the East coast of Greenland at (74◦ N,13◦ W), and the two peaks are less separated with the
first order than with the third order scheme. Things are worse as waves propagate further, and the two
peaks are almost indistinguishable north of Siberia on February 16 and 17, at (78◦ N,83◦ E). As a result,
the investigation of the potential diffusion effect of sea ice on these swells (Ardhuin et al., 2016) would
not be possible with a first order scheme. That first order scheme is still used at ECMWF.
It is also troubling that the mean values are different, because the diffusion in the first order scheme
strongly depends on the orientation relative to the grid: the discrete directions that travel along the grid
axis are much less diffused. This shows up as north-south beams of energy in the case of slowly moving
storms, such as Tropical Cyclone Dora, shown in figure 12.6. To reduce that effect, the usual trick is to
shift the directions by ∆θ /2: instead of using 0, 15◦ , 30◦ ... one uses 7.5◦ , 22.5◦ , 37.5◦ ... Now, instead
of a maximum along the meridian, we get a weak minimum (see top-left and bottom-left panels in figure
12.6).
When numerical diffusion is strongly reduced, the dispersion of the wave field can be less than
expected from the physical dispersion, giving an anomalous effect of the spectral discretization, known
as the ‘garden sprinkler effect’ (GSE). Without any GSE correction, the upwind QUICKEST scheme
gives the beautiful flower pattern in figure 12.6: because each spectral component has its own speed
and direction, waves from a very small source separate as they propagate away from the source as
discrete blobs of energy, instead of a smooth pattern that would be obtained with a very fine spectral
discretization in frequency and direction. The solution to correct this GSE is to introduce some explicit
diffusion (Tolman, 2002a).
Such differences are much less visible for fast moving storms, such as the more powerful cyclone
Gamede which struck La Reunion island on February 25.
12.3. PARAMETERIZATIONS OF PHYSICAL PROCESSES 131

PR1 UQ

PR1 + shift UQ + GSEC

0 1 2 3 4 5 6 7

Hs (m)
Figure 12.6: Example of wave heights for Tropical Cyclone Dora, in the Southern Indian ocean, using 4
different numerical schemes. ’GSEC’ is the GSE correction scheme of Tolman (2002a). All model runs
use 24 directions.

12.3 Parameterizations of physical processes

Because of the complexity of wave evolution processes that were briefly described in chapter 8, and further
discussed Part 3, there is no generally agreed upon parameterization for wave generation and dissipation.
Also, the non-linear wave interaction that is based on the theory of Hasselmann (1962) is too costly to
be computed accurately for routine applications. Hence several parameterizations have been proposed to
reproduce the non-linear source function. The most common is the Discrete Interaction Approximation
by Hasselmann et al. (1985b), which keeps the energy, action and momentum conservation properties of
the full non-linear interaction, at the price of an overestimation of the flux of energy to high frequency,
among other side effects.
For wave generation and dissipation, the parameterizations proposed by the WAMDI Group (1988),
Komen et al. (1994), Tolman and Chalikov (1996), Banner and Morison (2006), Ardhuin et al. (2010),
Rogers et al. (2010), and many others, have tried to represent the magnitude of the wind to wave energy
and momentum fluxes, as well as the overall balance that gives the wave growth. We cannot describe
here in details all the flavours of the proposed parameterizations, nor give all details about their impact
in the model solutions, which is the topic of an abundant record of scientific publications. We will
try to summarize what we believe to be the most salient aspects and side effects of a few common
parametrizaion, looking here only at the resulting wave spectra and derived parameters. In the context
of coupled models (wave-atmosphere or wave-ocean or wave-sea ice...) one should probably focus more
on the fluxes of energy and momentum than on the shape of the wave spectrum. This will be discussed
in chapter 14.
What follows is mostly taken from a few recent papers by Rascle and Ardhuin (2013), Roland and
Ardhuin (2014) and Stopa et al. (2016a), and focuses on large scale oceans away from the ice-covered
regions. Modelling in and around ice-covered regions will be discussed in chapter 11. Although we mostly
discuss the parameterization proposed in these papers and in Zieger et al. (2015), as implemented in the
WAVEWATCH III model (version 5.16), other papers, such as van Vledder et al. (2016), have discussed
132 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

other models.
Figure 12.7 shows maps of systematic (bias) and total error in Hs from the same model ran with four
different parameterizations, compared to altimeter data. Altough it is difficult to associate a particular
error pattern to a single feature of the parameterization, the reduction in error from 1988 (top, WAM
Cycle 3) to 2013 (Rascle and Ardhuin, 2013), shows that some of the errors have been well identified. In
this case, they are mostly associated to swell dissipation and spurious swell - windsea interactions. The
next section discusses how such errors can be detected and corrected.

12.3.1 A model development and validation strategy

Historically, the validation of wave models started with the reproduction of academic cases followed by
fetch-limited growth and then application to a few real cases (e.g. WAMDI Group, 1988). Although
this approach is still valid, it is not sufficient: indeed, the parameterization proposed by Tolman and
Chalikov (1996) had input and dissipation terms 3 times smaller than those of Janssen et al. (1994), but
still it gave pretty much the same fetch-limited growth curve and performed reasonably well in regional
modeling cases. However, Tolman (2002d) had to strongly modify the dissipation for the swells to arrive
at reasonable results for oceanic basin scales, as shown in figure 12.7. This demonstrated, as now realized
by anybody who tried to improve the global models (e.g. Ardhuin et al., 2010; Zieger et al., 2015), that
the swell dissipation coefficient is the single most sensitive parameter in a basin-scale or global wave
model.
Today, with all the readily available data, including satellite altimeters, in situ buoys and wave
mode synthetic aperture radar spectra, it is possible to adjust the empirical parameters in wave models,
adjusting the full model (forcing, numerics and parameterizations). Indeed, the academic cases used with
very high spectral resolution are good for debugging a model but they tell very little of the true model
performance when using say 24 directions and 30 frequencies.
Many critics of the Tolman and Chalikov (1996) parameterization put forth the idea that it had too
many ‘tunable coefficients’, which was indeed a problem if you wanted to modify it, but was the simple
consequence that it was a fit to the detailed numerical calculations of air flow over waves performed
by Chalikov (1993). In contrast, Janssen (1991) had managed to summarize his numerical modelling of
wind profile coupled to the wave spectrum by replacing the surface roughness by a simple Charnock-like
relationship, which only reproduced part of the full model beahviour, namely a stronger wave growth
at short fetch. That made the Janssen (1991) parameterization of the wind input more amenable to
further modification. However, the Janssen (1991) parameterization is very sensitive on the shape of the
high frequency spectrum tail, so that the results it gives in Banner and Morison (2006) or Ardhuin et al.
(2010) are very different from the original form.
Given the complexity of the air-sea interface, it is not very surprizing to have many degrees of freedom
in the form of adujstable parameters. More important is the strategy to adjust them separately, typically
by using a set of well defined wave evolution conditions and measured parameters to which the model
can be adjusted. Following Ardhuin et al. (2010) we can propose the algorithm,
1. Start from a reasonable set of physically plausible processes with quantified magnitudes or at least
orders of magnitudes
2. Test wave growth and value of Hs under 10 m/s wind, after 24 hours: this should be of the order
of 2.3 m: this controls the difference of input and dissipation Sin + Sds
3. Reproduce the November 3, 1999 case of slanting fetch-limited growth, first analyzed in Ardhuin
et al. (2007). In particular the mean direction as a function of frequency depends strongly on the
magnitude of Sin . The directional spreading also gives an indication on the balance of source terms
and their directional distribution.
4. Adjust the sheltering and cumulative terms in order to reproduce the observed variability of mean
square slopes (from altimeters) and vertical acceleration variance (from buoys)
5. check on the decay of pure swells using one of the many cases sampled by SARs, for both large
and small amplitude swells.
6. Verify the biases of swell peak periods against buoy data, and frequency-dependent biases in the
energy content of the spectrum.
7. Once the deep water evolution is correct, later adjust bottom friction (see chapter 18), then sea ice
effects (see chapter 11).
12.3. PARAMETERIZATIONS OF PHYSICAL PROCESSES 133

12.3.2 Overall performance for common parameters

We will not show time series of model and observations, as can be obtained using buoy data. Such
illustrations can be found in all publications. Instead we will use the globally available estimates of
Hs provided by satellite altimeters. Figure 12.7 shows typical error statistics over the global ocean
when using ECMWF operational analysis for the wind forcing, NCEP operational analyses for the sea
ice concentration, and Ifremer/CERSAT iceberg statistics for the southern ocean (Tournadre et al.,
2012). All these simulations use the DIA parameterization for the non-linear evolution, but different
parameterizations for the generation by the wind and various dissipation processes.
All four simulations have a near zero mean bias, but there are regional biases, particularly with the
parameterization by Tolman and Chalikov (1996), as adjusted by Tolman (2002c). The more recent
parameterizations generally perform better, but there can be local advantages, like the lower rms errors
of TC off West Africa. The differences between the BJA and TEST451 parameterizations are mainly in

Figure 12.7: Bias and normalized root mean square (RMS) error against altimeter data for the year
2007, using the same forcings but 4 different parameterizations of the wind input and dissipation: WAM
Cycle 3 (WAMDI 1988), TC (Tolman and Chalikov 1996, including the Tolman 2002c adjustment), BJA
(Bidlot et al. 2005) and TEST451 (Rascle and Ardhuin 2013). Solid lines in the right column correspond
to contours at the 7.5, 10, 12.5, 15 and 20% levels.

the dissipation source function.

12.3.3 Parameterization of the dissipation Sds

The dissipation used by Bidlot et al. (2005) and later adjustments (Bidlot, 2012), are based on the initial
proposition by Hasselmann (1974), later adjusted by Komen et al. (1984), that the dissipation rate is
134 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

13 9
12 bouée 62163 bouée 51001
8 (Nord-Ouest
11 (au large de
10 la Bretagne) de Hawaii)
7

Hs model (m)
Hs model (m)
9
8 6
7
5
6
5 4
4
3
3
2
2
1
0 1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 1 2 3 4 5 6 7 8 9
observations observations
NRMSE (%): RMSE: Bias (%): Corr.(r): S. I.(%): NRMSE (%): RMSE: Bias (%): Corr.(r): S. I.(%):
8.8 0.28 1.52 0.985 8.7 11.3 0.27 2.98 0.950 10.9
10.6 0.34 4.40 0.981 9.6 13.5 0.33 1.81 0.925 13.3
0.24
bouée 51001 bouée 51001
0.22
(Nord-Ouest (Nord-Ouest
fm02 model (Hz)

de Hawaii) 0.20 de Hawaii)

0.18

0.16
fp (Hz)

0.14

0.12

0.10

0.08

0.06
0.04
0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24 0.04 0.06 0.08 0.10 0.12 0.14 0.16 0.18 0.20 0.22 0.24
observations
observations
NRMSE (%): RMSE: Bias (%): Corr.(r): S. I.(%): NRMSE (%): RMSE: Bias (%): Corr.(r): S. I.(%):
5.8 0.009 0.71 0.9356 5.8 19.1 0.020 0.81 0.686 19.0
7.4 0.012 3.36 0.9197 6.6 26.4 0.028 7.58 0.475 25.3

Figure 12.8: Example of scatter plot comparing two model runs, with the Bidlot et al. (2005) parame-
terization and the Ardhuin et al. (2010) parameterization, described below. Data for the full year 2007
is used at two locations: the Meteo-France - UK Met Office buoy 62163, located off the French Atlantic
coast, and the NOAA buoy 51001, located North-West of Hawaii. all observed data has been averaged
over 3 hours in order to reduce statistical uncertainties.

defined for the entire spectrum by single mean steepness defined from the full spectrum,
"  2 #
WAM k k
Sds (k, θ) = Cds α2 σ δ1 + δ2 N (k, θ) (12.7)
k k

where Cds is a non-dimensional constant, δ1 and δ2 are constant weights. This expression uses a mean
wavenumber R p 1/p
k N (k, θ) dkdθ
k= R (12.8)
N (k, θ) dkdθ
where p = 0.5 in the version used by Bidlot et al. (2005). The corresponding mean frequency is
1/p
σ p N (k, θ) dkdθ
R
σ= R , (12.9)
N (k, θ) dkdθ
This spectral average gives a mean steepness
2
α = Ek (12.10)

This parameterization gives unrealistic variations of the wind sea dissipation in the presence of swell
(Ardhuin et al., 2007): the windsea dissipation can be much reduced by the addition of swell. This
12.3. PARAMETERIZATIONS OF PHYSICAL PROCESSES 135

spurious effect contributes to the larger scatter in the western part of ocean basins which are dominated
by wind seas, with the occasional presence of swells.
As explained in chapter 8, today’s understanding of wave breaking and swell dissipation processes,
although not complete, have led to parameterizations in which the steepness is more local in spectral
space. For example Ardhuin et al. (2010) have proposed to use
(   2   ′ 2 )
sat B (f ) B (f, θ)
Sds (f, θ) = σCds 0.3 max − 1, 0 + 0.7 max − 1, 0 F (f, θ) + Sds,c (f, θ).
Br Br
(12.11)
with two different steepnesses, one that is integrated in directions,
Z θ−∆θ
B ′ (f, θ) = 2π k 3 cos2 (θ + θ′ ) F (f, θ′ )/Cg dθ′ , (12.12)
θ−∆θ

and the other that is the same for all directions

B (f, ) = max {B ′ (f, θ), θ ∈ [0, 2π[} , (12.13)

with a threshold Br = 0.0009 so that waves are expected to break if B > Br (Banner et al., 2000, 2002).
Because the first term of eq. (12.11), in curly brackets, is unable to give an energy balance at high
frequency, a ‘cumulative term’ Sds,c (f, θ) is added that produces dissipation of the short waves induced
by long waves. Hence, starting from a global steepness, we have come to a local steepness, and we are now
slowly trying to determine the mutual interactions of the different components through the dissipation
term. This will likely keep us busy for many years to come.
Compared to BJA, a ‘sheltering term’ was also added by Banner and Morison (2006, 2010), reducing
the input for high-frequency waves. Ardhuin et al. (2008a) found that this term was necessary to
reproduce parameters associated to the high frequency part of the spectrum, with further adjustments
discussed by Rascle and Ardhuin (2013).
The other dominant factor at global scales, is the parameterization of swell dissipation as first realized
by Tolman (2002c). Indeed, swells account for the majority of wave energy over most of the globe (Chen
et al., 2002), and given that dissipation is the only source term for swells away from the storm, a small
change of the dissipation rate can lead to very large biases. The swell wave height also depends on the
wave heights in the storm.

12.3.4 Beyond Hs , Tp ...

Besides the common sea state parameters, Hs , Tm0,2 ..., a wave model can be used to compute many
other parameters, estimated from the spectrum or from the fluxes associated to source terms. How good
are all these?
Some parameters like the mean square slope and the surface Stokes drift are strongly influenced by
the high frequency part of the spectrum. For example, the Stokes drift vector at the sea surface is given
by, Z ∞ Z 2π
cosh(2kD)
Uss = kσ E(f, θ)dθdf (12.14)
0 0 sinh2 (kD)
in which cosh(2kD)/sinh2 (kD) → 2 as kD → ∞. In deep water, if all waves go in the same direction,
Uss is thus proportional to the third moment of the frequency spectrum, m3 as defined by eq. (3.16).
For Uss or the mean square slope, the dissipation parameterizations that are based on a mean steep-
ness can introduce spurious effects of long waves on short waves that generally give a poor representation
of Stokes drift variability. This is illustrated in figure 12.9 (see also figure 8 in Ardhuin et al., 2010).
Hence, an empirical approximation that gives Uss as a function of Hs and wind speed can be a better
choice than some model results (see eq. (C3) in Ardhuin et al., 2009b).
136 CHAPTER 12. NUMERICAL MODELING IN DEEP WATER

Observations 0.4 Model 0.4

Ussnd(0.36 Hz) (m/s)

(Ardhuin & al. 2010)

Hs (m):
Model
11 0.3 0.3 (Bidlot & al. 2010)
10
9
8 0.2
7 0.2
6
5
4 0.1 0.1
321

0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25
U10 (m/s) U10 (m/s) U10 (m/s)

Figure 12.9: Surface Stokes drift for all wave components with a frequency below 0.36 Hz, at the location
of NOAA buoy 46005, off the U.S. West Coast. In this calculation we have assumed that all waves travel
in the same direction, giving a non-directional Stokes drift Ussnd . The true directional spreading typically
gives a true Stokes drift that is 15% less. Observed and modeled values were binned as a function of
wind speed U10 and wave height. For each bin, the mean value is plotted and the black bars represent
half the standard deviation of the bin. We find that over 90% of the variance of Ussnd is explained by U10
and Hs . When using the parameterization by Bidlot et al. (2005), the variability is not well reproduced.
Chapter 13

Extreme waves and historical storms

13.1 Extreme significant wave heights

The maximum measured and modeled wave heights are associated with storms. However, because it
takes both a high wind speed and a large fetch and duration to produce large waves, the largest waves
are generally caused by storms with high winds that move at a speed close to the group speed of the
dominant waves. Figure 13.1 illustrates that Hs generally exceeds 16 m once every 10 years in a good
fraction of the North Atlantic: the biggest waves on Earth are found between Ireland and Iceland.

16

14

10 year Hs (m)
12

10

Figure 13.1: Estimate of the 10-year significant wave height (the expected maximum value of Hs that
occurs every 10 years) based on a 30-year hindcast, from 1985 to 2015, using WAVEWATCH III forced by
CFSR winds Saha et al. (2010), and fitting yearly maxima with a Generalized Extreme Value distribution.
Image courtesy of J. Stopa.

Wind speeds in tropical storms can be much faster, probably exceeding 70 m/s in the Typhoon Haiyan
that hit the Philippines in 2014, but the motion of these storms does not generally lead to the largest
wave heights (see also Quilfen et al., 2010).
The pattern on maxima is thus very different in the tropics where they are related to individual storm
tracks, and in the higher latitudes where the broad structures of extra-tropical storm give large values
of Hs over a wide region. For example the maximum in the Gulf of Mexico (around 11 m) in figure 13.1
is the result of the passage of a single storm: Hurricane Katrina, which led to widespread damage and
coastal flooding (e.g. Resio and Westerink, 2008). Using a different time frame, say around 1900 instead
of 2000, would probably have highlighted a different maximum, around Galveston, Texas, associated to
the 1900 hurricane that hit that part of the coast. Because it is not possible to know where the next
tropical storm tracks will be, an assessment of coastal hazards usually uses empirical storm tracks with
randomly shifted positions to investigate the local impact of a track displacement.
The highest-ever measured value of Hs is 20.1 m, using 1-Hz altimeter data from Jason 2. This
measurement was made over the North-Atlantic storm Quirin, in February 2011. Although the wind

137
138 CHAPTER 13. EXTREME WAVES AND HISTORICAL STORMS

Figure 13.2: Modeled winds (top panel, every 12 hours from 13 February at 00h00 to 14 February at
12h00) and observed (bottom) during the development of the storm Quirin in February 2011. The light
blue and magenta contours give the limit of tropical storm (U10 > 24.5 m/s) and hurricane force (U10 >
32.7 m/s) wind speeds.

speed probably approached 40 m/s, these waves were associated to a storm travelling at the right speed
across the Atlantic, amplifying the dominant waves along its path, which resulted in an effective fetch
exceeding 2000 km and a duration larger than 48 hours.
For different regions of the world, different storms have been recorded as particularly severe. In
the United States, these are usually associated to hurricanes such as the 1969 Hurrican Camille, or
combination of North-Easter storms and hurricanes (e.g. the 1991 Perfect Storm that inspired the
movie, see Bromirski, 2001).
13.2. PHYSICAL PROCESSES AT HIGH WIND SPEEDS 139

Figure 13.3: Measured Hs from altimeters on 13 and 14 February 2012. Bottom: measured (black) and
modeled (color) wave heights using different wind forcings: red is ECMWF operational analyses, green
is NOAA/NCEP analyses, and blue is the wind speed from NOAA/NCEP enhanced by 10%.

13.2 Physical processes at high wind speeds

Parameterizations in wave models have been developed for wind speeds ranging from 5 m/s (Snyder
et al., 1981) to 20 m/s or so. At higher wind speeds, the physical properties of the sea surface and the
airflow above it is expected to be significantly different. The sea surface is characterized by the presence
of foam and spray (e.g. Holthuijsen et al., 2012). In particular, Kelvin-Helmholz or other instabilities can
play a leading role Soloviev et al. (2014). It may be surprising that wave models actually work pretty
well for Hs up to 18 m (Rascle and Ardhuin, 2013), but this is probably because these large waves mostly
occur in regions when the wind is not so extreme, but rather where storms move with the waves.

13.3 Variability and trends of sea state parameters

As waves are related to ocean winds, so does the variability in wind speeds strongly impacts the waves.
The El-Nino Southern Oscillation has a very big impact on waves across the Pacific (Bromirski et al.,
2005; Stopa and Cheung, 2014), and North Atlantic storms are strongly affected by the North Atlantic
Oscillations (e.g. Dodet et al., 2010; Charles et al., 2012). Given this very large interannual variability,
it is very difficult to estimate long term trends, such as associated to global change. The studies by
Hemer et al. (2013), Wang et al. (2014), and Shimura et al. (2016) show that the shift of high winds
speed regimes towards high latitude generally leads to a reduction in wave height at latitudes under 50
degrees, and an increase at higher latitudes. However, the trends for extremes are a complex combination
of the number and intensity of tropical storms with the extra tropical storms, so that the extreme wave
heights may rise in some regions and decrease in others (Shimura et al., 2016). The case of the Arctic is
particular, with a general increase in wave height that is associated to a reduced extent of sea ice (Stopa
et al., 2016b).
140 CHAPTER 13. EXTREME WAVES AND HISTORICAL STORMS
Chapter 14

Air-sea interactions:
wind stress and mixing

Fluxes of any quantity (momentum, heat, mass, carbon dioxide or other gases ...) between the ocean
and the atmosphere is very often parameterized as the difference in the volumic density of this quantity
between the air and atmosphere, multiplied by an exchange coefficient. This exchange coefficient itself is
generally parameterized as a mean velocity difference across the air-sea interface, for practical reason the
norm of the vector difference of the wind at 10 m height U10 minus the quasi-Eulerian current at some
depth ub multiplied by a non-dimensional exchange coefficient Ce . That coefficient generally represents
the full complexity of the ocean surface, the presence of bubble in the water, spray in the air, and the
geometry of the interface. Because turbulent transport can be much more efficient, the proper scale for
trhe velocity is rather the friction velocity u⋆ , which means that generally Ce ≪ 1. The magnitude of
this exchange has profound impact on both the ocean and atmosphere, and natural phenomena such as
hurricanes are very sensitive to exchanges of heat and momentum. Here we will particularly focus on
the fluxes of momentum and energy, looking only at their impact ont the ocean. In general the feedback
on the atmosphere cannot be neglected.
A self-similarity theory of turbulence in which big eddies feed smaller eddies following Kolmogorov,
gives the Monin-Obukhov theory for the turbulence and associated fluxes in the atmospheric and marine
boundary layers (Monin, 1962; Zilitinkevitch and Chalikov, 1968). This was well verified over land
(Businger et al., 1971), and in particular it is usual to correct the wind speed U10 to a neutral wind
speed U10N to take into account the source of turbulence coming from buoyancy, in the unstable case of
warm water under cold air. A detailed account can be found in Schlichting (1979).
The ocean mixed layer in its top few meters is strikingly different from the atmospheric side, due to
the extra source of turbulence coming from ocean waves, mostly due to wave breaking but also associated
to the strectching of turbulence by the Stokes drift. This was only revealed in the 1990s and the full
details are still being explored, including the interaction of boundary layer turbulence with ocean fronts,
internal waves and other features (e.g. Suzuki et al., 2016).

14.1 Sea state influence on air-sea fluxes

14.1.1 Wind stress and drag coefficient
For the horizontal momentum, which is a two-component horizontal vector (U, V ), the flux is a also
vector (τx , τy ) and the exchange coefficient that relates these two quantities Ce should generally be a
tensor. However, it is most often replaced by a scalar CDN parameterized as follows, which makes the
big assumption that the flux vector, which is usually called ”wind stress”, is aligned with the wind vector
at 10 m,
τ = ρa CDN U10N U10N , (14.1)
where the neutral wind speed is the equivalent wind speeds that gives the same stress in the atmospheric
stratification is neutral. It was found in most experiment that C10N tends to be higher for relatively
young waves as shown in figure 14.1.
However, in general the wave age is correlated to the wind speed, so that it is very difficult to isolate
wave age effects. In their recent review of wind stress parameterization for wind speeds 0 to 25 m/s,

141
142 CHAPTER 14. AIR-SEA INTERACTIONS: WIND STRESS AND MIXING

Figure 14.1: Effect of sea state on the wind stress.

Left: measured neutral drag coefficient C10N , during the FETCH experiment in the Mediterranean.
Right: variation of the roughness length z0 , normalized by a typical wave amplitude σ = Hs /2 for
various experiments. Figure reproduced from Drennan et al. (2003).

Edson et al. (2013) concluded that the updated COARE 3.5 parameterization gives a good reproduction
of the wind stress as a function of the neutral wind speed alone.
For higher wind speeds, several estimates of the stress suggest that the value of Cd may decrease,
possibly associated to a full separation of the boundary layer. Figure 14.2 shows several drag coefficient
from observations and as used in the ECMWF atmospheric model, for which the drag takes into account
the waves (Janssen, 2004).

3
Cd (x 1000)

0
0 5 10 15 20 25 30 35 40
Wind speed (m/s)

Figure 14.2: Left: Comparison of drag coefficient for ECMWF (CY41R1) parameterization, empirically-
adjusted Charnock parameterization and observations (Donelan et al., 2004, ’R’ or ’M’ corresponds to
different measurements techniques Reynolds or Momentum Budget. Adapted from Pineau-Guillou et al.
(2018).

14.1.2 Swell and stress direction

Besides the magnitude of the wind stress, waves also modify the stress direction. This is particularly
noticeable at low wind speeds in the presence of swell. Indeed, the wind stress may be opposite to the
wind direction for wind speeds below 3 m/s, as the wave are loosing momentum to the atmosphere and
14.2. DRIFT AND MIXING 143

generating a low-level jet of wave-driven wind (Semedo et al., 2009; Hogstrom et al., 2009). It is more
frequent to observe systematic deviations of the wind stress and wind speed directions when waves and
wind are not aligned (Potter et al., 2015).

14.1.3 Other effects

Waves generally have an impact on all air-sea exchanges, not just momentum. This includes mass fluxe
(see Veron, 2015, for a recent review of spray generation) and gas transfer (e.g. Brumer et al., 2017). Also
in the presence of an ice layer, wave-ice interactions should be taken into account. These are discussed
in chapter 11.

14.2 Drift and mixing

14.2.1 Momentum flux for the Eulerian mean current
Knowing the wind stress is a first step in determining how the ocean is forced, but this is not enough to
tell how much of that momentum goes into currents as most of it generally goes through the wave field.
It is only when the wave momentum is constant that the momentum flux from the wind goes entirely
to the ocean circulation. In general, we can use the wave energy balance with source terms Sin and Sdis
presented in Chapter 8, to compute the wave momentum balance. This was done at the end of Chapter
10, when we introduced the momentum flux from wind to waves τ aw and the flux from waves to ocean
τ wo .

14.2.2 Quasi-Eulerian currents

The interaction of waves and currents is the topic of onging research, in particular the effects of vertically
sheared currents on the waves. A general description of these interaction in three dimensions is given
in Part 3. Here we will consider the much more simple case of a horizontally homogeneous ocean. The
wave-averaged momentum equation for the mean current u b then takes the following form (Hasselmann,
1970; Xu and Bowen, 1994),
 
∂b
u ∂ ∂b
u
= −f ez × (b
u + Us ) + Kz − Two (14.2)
∂t ∂z ∂z

The solution is determined by the surface boundary conditions, and the profiles of the mixing coef-
ficient Kz , and the total momentum injected by the waves Two which is a depth-distributed force such
that the momentum flux is
Z ζ Z
ρw g
ταwo = Tαwo dz = [Stot (f, θ) − Satm (f, θ)] df dθ (14.3)
−h C

with the source terms S defined in Chapter 8.

The current can be obtained by solving (14.2) with a turbulent closure that gives an estimate of
Kz . Following Prandtl (1904), the usual turbulence closure has an addy viscosity that increases with the
distance from the boundary (Schlichting, 1979) and involves a velocity scale q associated to the turbulent
motion,
Kz = qSM l, (14.4)

where the turbulent kinetic energy per unit mass is q 2 = u′i u′i , SM ≃ 0.39 is a constant. With this, the
most simple model uses a prescribed mixing length l that is the maximum distance from the surface or
the base of the mixed layer.

l = max {−κDm ς, κz0− } for (−h + ζ)/2 < z < ζ

l = max {κDm (1 + ς), κz0b } for − h < z < (−h + ζ)/2, (14.5)

where Dm is the thickness of the mixed layer. The roughness length z0− yields a non-zero value of Kz
at the surface, which is consistent with measurements (e.g. Kitaigorodskii, 1994). Thorpe et al. (2003b)
144 CHAPTER 14. AIR-SEA INTERACTIONS: WIND STRESS AND MIXING

10

ûx

depth (m)
20
ûx ûy
ûy 30 q
q Usx
z= -z 0
40 ûx+Usx
z= -z 0
a) b)
50
-0.05 0 005
. 0.1 0.05 0 0.05 0.1
Velocity (m/s) Velocity (m/s)

Figure 14.3: Upper ocean current profiles for a wind speed U10 = 10 m/s
(a) Profiles with a very low value of z0 , representing an unrealistic situation without waves (b) profiles
with a realistic value of z0 . In terms of drift velocity, the much lower value of the surface current is
compensated by the Stokes drift. The velocity scale q is the square root of the turbulent kinetic enenergy
(TKE). A realistic surface flux of q is needed to get the realistic TKE profile in (b). (Figure courtesy of
Nicolas Rascle).

and other authors have suggested that z0 is of the order of Hs , the significant wave height, and should
be more precisely related to the height of the breaking waves1 .
More complex models consider also an equation of evolution for q, and one must also define the
surface boundary condition for q. For example Mellor and Yamada (1982) use the following equations

∂q 2 ρ0
lqSq = αCB 0a u3⋆ at ς = 0, (14.6)
∂z ρw
2
∂q
lqSq = 0 at ς = −1, (14.7)
∂z
where Sq = 0.2. The mixing coefficient for q 2 is lqSq , and αCB is the ratio of the energy flux coming
from ocean waves (presumably due to wave breaking), and the friction velocity cubed. This coefficient
was particulary discussed by Craig and Banner (1994) and many following authors (Terray et al., 2000;
Mellor and Blumberg, 2004; Rascle and Ardhuin, 2009, 2013). Typically αCB is of the order of 100.
For a simple estimation we consider a fully developed sea state, and we assume that Two is concen-
trated near the surface so that we can replace these terms in (14.2) by a local source of momentum in
the surface boundary condition. Namely the net momentum flux coming from the wind and waves is
balanced by vertical mixing,

∂b
uα
τa − ταaw + ταwo = ρw Kz on ς = 0. (14.8)
∂ς
In fact, in the presence of a strong surface mixing the numerical solution is not very sensitive to the
forcing in the boundary condition or as a body force concentrated near the surface (Rascle et al., 2013).
Assuming a fully delveloped sea state the momentum flux ταaw that goes to the wave growth is canceled
by the dissipation terms and we have

∂b
uα
τa = ρw u⋆ u⋆ = ρw Kz on ς = 0, (14.9)
∂ς
with u⋆ is the wind friction velocity.
Until the 1990s, all ocean circulation models used very small values of mixing at the surface (e.g.
Large et al., 1994), corresponding here to low values of z0− , such as in Figure 14.3.a. This can give
very high values of the surface current,s depending on the vertical resolution, up to the usally observed
3% of the wind speed (Huang, 1979) or more. But Agrawal et al. (1992) found that the disspation of
1 Ocean circulation models before the year 2000 used to ignore these effects, and could have very small values of K at
z
the surface, e.g. Large et al. (1994), which could give highly unrealistic values of the surface drift when using very high
vertical resolutions.
14.2. DRIFT AND MIXING 145

0.5

-z/H s SASS
VMCM
10

0 2 4 6 8
U/u w
*
Figure 14.4: Quasi-Eulerian velocities near the sea surface.
Two types of current-meters (SASS and VMCM) provide mean current velocities that have been corrected
for wave motion. u⋆w = (ρa /ρw )1/2 u⋆ is the friction velocity in the water (Figure from Terray et al.
2000). The difference in current velocity between the surface and the thermocline is of the order of 0.5%
of the wind speed U10 .

TKE was much higher in reality, by at least one order of magnitude, so that the mixing must have been
strongly underestimated. Recent models have clearly shown that higher surface mixing values are more
realistic. For example Mellor and Blumberg (2004) use z0 = 0.8Hs to get a better fit to measure sea
surface temperatures in the Gulf of Alaska. This gives a much weaker surface quasi-Eulerian velocity.
This mixing induced by wave breaking is particularly important for relatively shallow mixed layers such
as found in the Arabian Sea in summer (Janssen, 2011), or caused by the diurnal cycle of heating (Noh,
1996; Noh and Kim, 1999).
There are very few measurements of velocity profiles of Eulerian or Lagrangian velocity within the
upper few meters of the surface. Observations by Santala and Terray (1992) show an Eulerian current
that does not exceed 0.5% of the wind speed, see figure 14.4. How general is that? Is it still true if we
can measure within a few centimeters of the surface, or right at the surface? This is not known yet but
several techniques using thermal imagery or polarimetry and wave dispersion should be able to answer
these questions.

14.2.3 Stokes drift

The difference between the Lagragian mean velocity, which is the speed or water particles and the speed
that advects tracers (temperature, salinity ...), and the quasi-Eulerian velocity is the Stokes drift. To
lowest order, this Stokes drift can be computed from the wave spectrum (Chapter 9). For random wave
field in deep water, eq. (9.5) gives (Kenyon, 1969),
Z Z
Us (z) = 2σk cos(θ)E(f, θ) exp(2kz)df dθ, (14.10)
Z Z
Vs (z) = 2σk sin(θ)E(f, θ) exp(2kz)df dθ. (14.11)

Clearly this expression has a strong contribution from short waves, and thus should be largely influ-
enced by the local wind. Still, for any given wind speed, the surface Stokes drift value USs has a root
mean square variability of the order of 40%, particularly for relatively low wind speeds, below 7 m/s, as
shown in figure 14.5.
Based on direction spectra measured by surface-following buoys, Ardhuin et al. (2009b) found that
the surface Stokes drift USs could be estimated fairly accurately, with a root-mean-square error of the
 146 CHAPTER 14. AIR-SEA INTERACTIONS: WIND STRESS AND MIXING

0.35 mean(US)
0.3 mean(US)−std 0.3
mean(US)+std
0.25 0.25

0.2
U (m/s)

0.2

U (m/s)
0.15 0.15
S

S
0.1 0.1

0.05 (a) Station PAPA 0.05 (b) Pierres Noires

(WMO 46246) (WMO 62069)
0 0
0 5 10 15 20 25 0 5 10 15 20 25
U10 (m/s) U (m/s)
10

Figure 14.5: Example of mean value (in red) of the surface Stokes drift vector norm Uss = |(Uss , Vss )|
as a function of wind speed for two locations: station PAPA in the North-East Pacific, and buoy 62069
off the French Atlantic coast. These are obtained by integrating the wave spectrum up to 0.5 Hz. The
black symbols show the mean plus or minus one standard deviation for each wind speed. The dashed
grey line is US = 0.01U10 .

order of 20%, by an expression as a function of the wind speed and wave height,
"  1.3 #
−4 0.5
USs (fc ) ≃ 3.7 × 10 1.25 − 0.25 U10 min {U10 , 14.5} + 0.025 (Hs − 0.4) , (14.12)
fc

in which fc is the frequency up to which the Stokes drift is taken into account. This expression was
validated for 0.3 < fc < 0.6 Hz.
When subtracting this Stokes drift from HF radar data they found that the quasi-Eulerian current was
of the order of 0.4 to 0.8% of the wind speed, with important variations due to inertial oscillations, and, in
the Northern hemisphere, a mean direction 60 degrees to the right of the wind. The model proposed here
adds the Stokes drift to the quasi-Eulerian velocity and matches relatively well observations of surface
drift and mixing (Rascle et al., 2006). The effect of stratification is particularly discussed by Rascle and
Ardhuin (2009).
The present model gives both a strong vertical shear of the drift velocity (mostly due to the Stokes
drift) and a strong mixing (caused by wave breaking). Still the modeled drift is low by 0.5 to 1.5%
of U10 compared to the typical 3% surface drift. One possible reason is that surface drift objects are
trapped in convergence zones where the mean velocity is faster than that of the surrounding water. These
convergence zones are associated to Langmuir cells.

14.3 Langmuir circulation

Indeed the velocities in the upper ocean are not homogeneous horizontally (Weller et al., 1985). Many
observations, starting with Langmuir (1938) have revealed lines of convergence where foam, flotsam,
sargassum and any buoyant material gathers at the surface (Thorpe et al., 2003a). These lines are
roughly aligned with the wind direction, and correspond to the surface convergence of the rolls that form
the Langmuir circulation (Figure 14.6). These rolls emerge due to an instability of the mean vertical
shear ∂b u/∂z that is stretched by the Stokes drift shear ∂Us /∂z and thus get their energy from the wave
field via the turbulent kinetic energy production term u′ w′ ∂Us /∂z, as detailed in Part 3. The momentum
balance that gives these rolls was further analyzed by Suzuki and Fox-Kemper (2016). These rolls have
been observed in most experiments in the ocean and in the laboratory (e.g. Thorpe, 1992; Melville et al.,
1998; Smith, 1999), including in shallow water (Marmorino et al., 2005). They are well reproduced in
Large Eddy Simulations (e.g. Noh et al., 2004; Harcourt and D’Asaro, 2006; Sullivan and McWilliams,
2010) and can interact with mixed layer fronts (Suzuki et al., 2016).
The parameterization of Langmuir circulation in models that do not resolve them is the topic of active
research (e.g. Li and Fox-Kemper, 2017).
14.3. LANGMUIR CIRCULATION 147

u ~ (10 to 15)u*

Drift path of
Physalia 5 to 15
e
nc

ce
ge

en
er

erg
d
nv

Win

es
nv
Co

av
Co

v' asymmetric? v' ~ ?? Ekman

w' ~ 5u* Spiral
Plankton
bubbles

Thermocline
(Form or kinematics of bottom part still not well known)

Figure 14.6: Langmuir cells (figure by J. A. Smith)

148 CHAPTER 14. AIR-SEA INTERACTIONS: WIND STRESS AND MIXING
Chapter 15

Waves and ocean remote sensing

Waves are usually the first type of motion that you see at the sea surface, and they appear in all
remotely sensed data. In some cases, the direct wave influence can be averaged out of the signal. In
other cases there is a residual bias due to the presence of waves. This is the case in range measurements
with altimeters (e.g. Minster et al., 1991), velocity measurement with Doppler systems (Chapron et al.,
2005; Nouguier et al., 2018), surface brighness temperature measurements used to infer sea surface
temperature or salinity (Reul and Chapron, 2003). Wave shapes and motion also introduce a variance
in the measured quantity which can be useful in the case of sea level measurement with altimetry, or
can blur the signal beyond recognition in SAR imagery or interferometry (Peral et al., 2015). All these
effects are opportunities for measuring wave parameters, or measuring other processes thanks to their
influence on waves.

15.1 Back-scatter from a rough surface

The patterns observed on the sea surface depend on the surface properties (slopes, velocities, whitecaps
...) but also on the viewing geometry and on instrument parameters (electromagnetic wavelength po-
larization) and on processing details (real aperture, synthetic aperture ...), it is thus difficult to give a
comprehensive accounts of the many different looks that the same piece of ocean can have.

active radar passive optical

Sun
monostatic case bistatic case
rix r
ct so
se en
pl

bi n-s
at city

θsun
ve

fo
lo

Su
rm

θv
sensor sensor
Up Up
β
North θ North
VLOS
φ East 5 m pixel φ'
dR East
footprint reflecting facet
containing with slopes sx,sy
NRCS gradient nadir nadir

Figure 15.1: Two typical examples of observation geometry of the ocean. Left: observation with a
radar looking sideways, the radar is both the source and the receiver of the signal, this is a monostatic
measuremnt with ocean apparent properties determined by the incidence angle θv and azimuth of the
line of sight, as well as radar wavelength and polarization. Right: observation of the sun reflection of
the sea surface. Here, for this bistatic geometry, the important direction is not the line of sight but the
direction of the sun-sensor bisectrix.

For electromagnetic wavelengths much shorter than the shortest waves (a few millimeters), the ocean
surface is locally smooth and the reflection can be completely described by geometrical optics: we can
decompose the sea surface in elementary ”facets” that are locally flat and reflect the light in the specular
direction.
When looking in the sun glitter, the apparent brightness of the the ocean depends on the probability
that the piece of ocean considered (a ”pixel”) contains the slope that corresponds to a bistatic reflector

149
150 CHAPTER 15. WAVES AND OCEAN REMOTE SENSING

of the sun. For pixels smaller than a few hundreds of meters, the brightness of these pixels fluctuates
due to the finite number of ”specular points” (Longuet-Higgins, 1960).
A first good approximation for the slope PDF is that it follows a Guassian distribution for the two
dimensions that are the slopes sx = ∂ζ/∂x and sy = ∂ζ/∂y along the x and y directions. In this section
we will now choose to have the x direction in the wind direction. Under this Gaussian approximation
the slope PDF is completely determined by slopes variances mssx = s2x and mssy = s2y as,
" !#
1 1 s2x s2y
p(sx , sy ) ≃ pG (sx , sy ) = √ exp − + . (15.1)
2π mssx mssy 2 mssx mssy

For moderate wind speeds, mssx ≃ mssy ≃ mss/2 and the slope distribution further simplifies as an
isotropic function that is completely defined by the total mean square slope mss=mssx + mssy ,
" #
1 s2x + s2y
p(sx , sy ) ≃ pG,iso (sx , sy ) = exp − . (15.2)
πmss mss

All measurements, pioneered by Cox and Munk (1954) and confirmed by Bréon and Henriot (2006),
have shown that the variances of crosswind and downwind slopes grow linearly with the wind speed, as
shown in figure 15.2.a, with a faster growth of the downwind slopes. The total mean square slope (mss)
is the sum of these two components, with a value around 0.029 for a wind speed of 5 m/s, corresponding
to a root mean square slope of 0.17, or an angle of the sea surface of 10 degrees relative to the horizontal.
In practice the mss also varies with the state of development of the wind sea, by up to 20% (±1 dB in
Nouguier et al., 2016), and the presence of current gradients, by about 10% (e.g. Rascle et al., 2017).
The largest local changes in mss are associated to the presence of oily films at the sea surface that can
strongly damp the shortest wave components.

(b)
p(sx)
(a) 3
U = 5 m/s
U = 5 m/s,
Gaussian approx.
<sx²> or <sy²> x 0.01
5
4
Cox & Munk (1954)
Breon & Henriot (2006) ²>
2 U10
<s
x
²> 11 to 12 m s
<s y
3
2 1
1
00 1 0
5 10 15 -0.4 -0.2 0 0.2 0.4
U10(m/s) sx (downwind)
Figure 15.2: (a) variation of the downwind and crosswind slope varianace with wind speed. (b) example
of slope PDF for a given mean square slope, corresponding to a widn speed of 5 m/s or in the range 11
to 12 m/s. Adapted from Munk (2009)

For a given slope (or view direction), the PDF (or image brightness) changes if the mss changes.
In figure 15.2.b, the PDF increases for small slopes when the mss drops from 0.06 (for 11.5 m/s wind)
to 0.03 (for 5 m/s wind, or for higher winds but in the presence of surface slicks), and it decreases for
larger slopes. Hence the contrast in the sun reflection in an optical or radar image depends on viewing
geometry: near the center of the sun glint (or near the vertical for a monostatic radar) a slick flat surface
will appear bright, but it will be dark away from the center (at higher incidences for a monostatic radar)
as shown in the photograph in figure 15.3 for a bistatic viewing geometry.
The Gaussian approximation can be corrected with a Gram-Charlier expansion based on the mea-
surement of various statistical moments of the slope PDF, including the mean, which is not zero, as well
as the asymmetry and skewness of the waves (Cox and Munk, 1954; Munk, 2009). Fig. 15.2.b shows
that for low wind speeds, non-Gaussian effects mostly introduce a larger probability of near-zero slopes
and high slopes, but little change of the probability around sx = mssx . These can be associated to non-
linear Stokes-type corrections to the linear wave profile. At higher wind speeds, the asymmetry becomes
15.1. BACK-SCATTER FROM A ROUGH SURFACE 151

Figure 15.3: A view of the ocean from an aircraft, looking at the sun reflection: locally the ocean surface
is made smooth by the presence of surface material that makes the sea smooth. When this happens near
the center of the sun glint, the image is brighter, when this is away from the center of the sun glint, the
ocean appears darker. Image taken in the Californial bight by Luc Lenain.

significant, and the most likely surface slope is not at zero, but at slightly positive value of ∂ζ/∂x that
corresponds to the larger area of the (positively sloping: ∂ζ/∂x > 0) rear face of the waves compared to
the (negatively sloping) front face of the waves, the latter being steeper and shorter.
In the case of visible light, the diffuse light source from the blue sky and clouds must be taken into
account for large values of β, and at night the moon can be used instead of the sun.
For longer eletromagnetic wavelenghts, i.e. all main active radar bands used in ocean remote sensing
from Ka band (8 mm wavelength) to L band (20 cm or so), the active radar can generally be considered
to be the sole source of radiation, although that could change with 5G mobile phones infringing on the
Ka-band. The surface is rough for microwave radiation, although specular reflection is still relevant at
near-nadir viewing angles, particularly strong reflection occurs when the phases of the surface topography
combine constructively with the phases of the incident radar waves: this is Bragg scattering and it
typically explains most of the reflections at large incidence angles θv .

15.1.1 near-nadir incidences

For near-vertical angles the power recorded on a radar is well described by the theory for a nearly
Gaussian distribution of surface slopes, completely determined by the mss. In general, the mss is largely
determined by the wind speed, but it is also modified by the stage of development of the wave field,
increasing for more mature waves that generally correspond to higher wave heights, as shown in figure
15.4.b.
A flat surface gives a strong return near zero incidence (vertical sounding) and the return power
152 CHAPTER 15. WAVES AND OCEAN REMOTE SENSING

12
(b)
10

2
(a) 0
incidence angle (degrees)
-2
0 2 4 6 8 10 12 14 16
incidence angle (degrees)

Figure 15.4: (a) Average backscatter power from the TRMM radar (from Freilich and Vanhoff, 2003) (b)
Same variation for a given wind speed, as a function of wave height (from Nouguier et al., 2016).

decreases as 1/mss. For incience angles larger than about 10◦ in Ku-band, the return increases with the
roughness.

15.1.2 higher incidence angle

For large incidence angles the reflection is proportional to the amplitude of waves in the radar look
direction and with a wavelength equal to λe /(2 sin θi ) where λe is the electromagnetic wavelength. These
waves are called ’Bragg waves’. A similar scattering of waves by a periodic medium was described by
Bragg (1913) in the case of X-ray diffraction, and is thus known as Bragg scattering, although it was first
studied by Rayleigh (1896) for the reflection of sound waves at a wavy surface. This generally applies
to incidence angles above 30◦ , and is used in the remote sensing of winds with “scatterometers” as well
as the measurement of currents with HF-radars having wavelength of several meters and using grazing
incidence angles (close to 90◦ ).

15.2 Various applications

15.2.1 Roughness and surface current gradients
Waves with wavelengths under 3 m are the main contribution to the surface slope variance (e.g. Cox and
Munk, 1954; Vandemark et al., 2002), and these short waves are strongly modified by current gradients
(e.g. Phillips, 1984). As a result, current gradients have a clear signature in the mean square slopes and
the measured back-scatter intensity for incidences 0 to 20◦ , as shown in figure 15.5. This property has
been particularly investigated by Kudryavtsev et al. (2012); Rascle et al. (2014, 2017), with the objective
of estimated current gradients from optical imagery in the ’sun glint’, i.e. for incidence angles close to
the specular reflection direction of the sun. The quantitative analysis of glint data is generally based on
the wave action equation as given in chapter 10.
15.2. VARIOUS APPLICATIONS 153

Figure 15.5: Surface roughness observed by the MERIS instrument on board ENVISAT.
154 CHAPTER 15. WAVES AND OCEAN REMOTE SENSING
 Part II

Waves in coastal and nearshore

environments

155
Chapter 16

Linear shoaling, refraction and

reflection

In chapter 2, we assumed that the wave field had spatially homogeneous amplitudes, propagating over a
flat bottom. We will now allow the water depth to vary, which, like variations in the current velocity in
chapter 10, causes important wave modifications.

16.1 Wave shoaling

In the absence of current, wave rays are orthogonal to wave crests are streamlines for the energy. Con-
sidering two rays of the same monochromatic wave train, parallel in a region where C is uniform (for
instance in deep water, KD >> 1, and without current) and spaced from ∆l, the energy flux between
those two rays Cg E∆l is conserved along this ’tube’. All these properties can be demonstrated from
Laplace equation and from the boundary conditions at the surface and at the bottom, assuming that
amplitudes and phase velocities vary slowly.
Without current and for a beach with an uniform topography along the y-axis (longshore), wave
propagation along the x-axis have therefore rectilinear rays et ∆l does not change along the tube defined
par two of its rays. Hence, the spectral density of energy E matches the group velocity Cg changes
(see figure 2.51) so that the energy flux Cg E remains constant. In particular, in the coastal area,
1/2
Cg = (gD) , therefore, the spectral density of energy increases as D−1/2 . Yet, E is proportional to the
1/2
mean or significant wave height squared Hrms = 2−1/2 Hs and Hs = 4 (E/ρg) . Therefore, the wave
−1/4
height increases as D . In practice the wave height is limited for wave breaking when D approaches
zero.
This modification of the wave height due to variations of the group velocity is called shoaling. But
on a real shoal, which has a finite length, the non-uniformity of the depth along the crests also causes
refraction.

16.2 Refraction
We showed in chapter 2 that the wave phase velocity was a function of frequency and water depth. In
presence of an horizontal current U(x), vertically uniform, we observe a Doppler shift as well. The wave
angular frequency, in a fixed referential becomes then, ω = σ + k · U, and the phase speed is :
ω hg i1/2 1
C= = tanh (kD) + k·U (16.1)
k k k
The phase speed C changes induce the refraction phenomenon, discovered by Snel and Descartes in
optics.
Without current, this effect is perceptible from the moment that the water depth is less than half
the wave length (kD < π). Considering two areas with uniform water depths D1 and D2 for x < 0 and
x > 0, then the Snel law (also attributed to Descartes1 ) applies and expresses the wavenumber ky at the
boundary,
1 The Dutch mathematician Willebrord Snel discovered the refraction law in 1621, but it was only published in 1703

157
158 CHAPTER 16. LINEAR SHOALING, REFRACTION AND REFLECTION

sin θ1 sin θ2
= (16.2)
C1 C2
Where θ1 et θ2 are the angles between the propagation direction and the x-axis. This results applies to
a beach with a topography uniform along the y-axis. In this situation, sin θ/C is conserved by refraction
(figure 16.1).
All the results of geometric optics apply, replacing light velocity by C, in particular Fermat’s principle:
the integral of C along a trajectory is minimum in the variational sense. Hence, a bump at the bottom
acts like a optical converging lens while a trough will be divergent. This explains why waves converge
towards capes, increasing their height. Waves propagating against a localized current vein (the current
speed is zero outside the current vein) are deviated toward its center. In addition to being shortened due
to the Doppler effect, waves are higher and steeper, hence more dangerous, as in the Aguhlas current.
We can obtain a differential equation for the trajectories followed by waves from that same principle.
These trajectories are also called rays of characteristics. Let (x, y, θ) be the position and direction of
waves in a single point of the ray, and s the curvilinear coordinate along the ray, without current,

dx
= cos (θ) , (16.3)
ds
dy
= sin (θ) , (16.4)
ds  
dθ 1 dC dh dh
= · sin (θ) − · cos (θ) . (16.5)
ds C dh dx dy

Without current, the motion of a wave packet is given by his speed ds/dt = Cg in the direction θ.

C1 C2 C3

qi1
qi2 qi3
qr1
qr2

Figure 16.1: Propagation directions for incident or reflected waves, for each part of two phase speed
C discontinuities. For a natural topography, refraction is steady and the intensity of reflected waves is
generally weak.

With a current ds/dt = |Cg + U| and the wave propagation direction is different from the direction
perpendicular to wave crests. These ”rays” have long been calculated by hand (Munk and Traylor, 1947)
before numerical methods took over Dobson (1967).
It is very insightful to look at wave plans, the diagrams showing the ray spacing from directions
parallel offshore. (see example in figure 16.2). From the energy flux conservation between two rays, a
ray spacing l narrower than the offshore spacing l0 means that the local wave height increases. This is
the case in front
p of Long Beach pier for the 20 s south swell. This augmentation
p only due to refraction,
is of a factor l0 /l, than needs to be multiplied by the shoaling coefficient Cg0 /Cg . This combination
gives a seven-fold increase in wave height that explains the destruction of the Long Beach Pier in April
1930 due to an unusual south swell (Lacombe, 1950). Indeed, only the very long swells can be refracted
by the bottom topography at 200 m depth. The pier has been rebuilt since its destruction but with a
in Christian Huygens’s book Dioptrica where Snel is named in latin (Snellius) which leads to the frequent errors of the
English speakers that write his name Snell instead of Snel. The french philosoph René Descartes gives the Snel law in his
treaty La dioptrique, an appendix of his famous Discours de la méthode pour bien conduire sa raison et chercher la vérité
dans les sciences publish in Leiden in 1637 and apparently inspired from Snel’s work, although Descartes repeated Snel’s
experiments in 1626 and 1627.(Source: the MacTutor history of mathematics archive, University of St Andrews, Scotland,
http://www-groups.dcs.st-andrews.ac.uk/∼history)
16.2. REFRACTION 159

North

Wave rays

Advance of crest in 7.3 s

Wave direction
in deep water (260°)

Nautical miles (1 mile is 1.852 km)

Figure 16.2: Refraction diagram for waves of 20 s period from 160◦ at Long Beach harbor, California
From Lacombe (1950). Note the units on the dashed depth contours: 1 fathom is 1.83 m.
160 CHAPTER 16. LINEAR SHOALING, REFRACTION AND REFLECTION

Figure 16.3: Refraction diagram for 12 s swell over the Gouf de Cap Breton
From Lacombe (1950).
16.2. REFRACTION 161

Figure 16.4: Examples of two wave rays methods

a. ”Traditional” computation for a west swell with a 15-s period, near La Jolla, California. The location
of some the wave buoys deployed during the NCEX experiment is indicated by numbers 32 to 37 and the
10, 50, 100 et 200 m isobaths are shown by the dashed lines. From Magne et al. (2007). b. computation
of back-trajectories from point 34, for the same period T = 15 s and arrival directions evenly spaced
with a 1◦ interval. Only the wave rays reaching offshore, and able to bring energy at point 34 are shown
(this is true only within the optical geometry approximation, and by neglecting the wave reflection at
the shore. Regular marks are visible along the ray and correspond to the distance covered by the group
velocity within time intervals of 60 seconds.)
162 CHAPTER 16. LINEAR SHOALING, REFRACTION AND REFLECTION

slightly different configuration. This computation for Long Beach can be compared to the one for 12-s
waves at Cap Breton, France (figure 16.3).
When using a numerical calculation, back-trajectories from a fixed point are more reliable (see for
instance in figure 16.4.b). That latter method avoids the occurrence of singularities such as caustics
that occur when forward-propagated rays cross. Indeed, geometrical optics predicts that the energy of
a monochromatic wave train becomes infinite at the caustic. In fact, even under geometrical optics the
wave height remains finite because rela waves are not monochromatic and the the caustics wave rays of
the different spectral components are not at the same location. In the case of for monochromatic wave,
the wave height is in fine limited by wave breaking or diffraction.
When using backward ray tracing, we use the equality of the of the spectral densities in the coordinates
(kx , ky ), between two points A et B of the same ray. Hence for a zero source term, and in the stationary
case without current, using (3.10), equation (8.44) gives

Cg,B kA
EA (f, θA ) = EB (f, θB ) , (16.6)
Cg,A kB

where θA et θB the ray directions when they cross A and B, respectively. One may hence transform an
offshore spectrum to the coast to account for refraction and shoaling. In this situation, we shall often
assume that the offshore spectrum is relatively uniform. Reciprocally, we may also estimate an offshore
sea state from coastal measurements. This technique can easily be extended to non-stationary situations
by adding a time shift between A and B corresponding to the propagation duration, i.e., the integral of
1/Cg along the ray. The difficulty of these calculations are limited to the ray tracing that can be done
once for all for stationary media (when the tide effect is negligible). This transformation method of a sea
state from offshore to the coastline is often very precise, especially for situations dominated by shoaling
and refraction (O’Reilly and Guza, 1993; Ardhuin et al., 2003b; Ardhuin, 2006b; Magne et al., 2007).

16.3 Diffraction
So far, we have considered that the wave amplitude and the properties of the medium in which they
propagate vary slowly in comparison to the wave period on wave length (WKB approximation).
It has been noticed earlier that for caustics due to refraction of a monochromatic wave, this WKB
approximation is not valid. This assumption is not verified in the vicinity of obstacles such as breakwaters.
For small amplitude waves, neglecting the wind effect and the bottom friction, one may use the linearized
equations. Over a flat bottom and with such obstacles, a solution for ϕ can be found with the following
equation
cosh (k0 z + k0 D) −iωt
ϕ = ϕ̂(x) e + c. c., (16.7)
cosh (k0 D)
where ω 2 = gk0 tanh (k0 D). ϕ verifies the cinematic boundary conditions at the bottom and surface.
The Laplace equation, simplifies into the Helmoltz equation,

∇2 ϕ̂ + k 2 ϕ̂ = 0. (16.8)

In general, the elliptic nature of that equation imposes a solution method with boundary conditions
specified along a closed contour.
We can look at different approximations of the wave equation by taking a wave train with amplitude
Φ̂ that varies slowly on the scale x e = εx, giving a complex amplitude ϕ̂ = Φ̂ (ex) eiS(x) , with a local
wavenumber k = ∇S, that also varies slowly.
Using that form in the Laplace equation gives
 
∇2 ϕ̂ = ∇ · ∇Φ̂eiS (16.9)
h  i
= ∇ · ε∇Φ̂ + ikΦ̂ eiS (16.10)
h i
= ε2 ∇2 Φ̂ + 2iεk · ∇Φ̂ + iε∇ · k − k 2 Φ̂ eiS


(16.11)
(16.12)

At zero order in ε, the variations of the amplitude Φ̂ are neglected to keep only the phase variations,
S (x), and we obtain |k| = k0 meaning that wave trains propagate exactly like plane waves.
16.3. DIFFRACTION 163

Figure 16.5: Example of propagation of a monochromatic 16 s period, 1 m amplitude offshore swell.

Computation made with the finite differences elliptical extended Berkhoff model, from Athanassoulis et
Belibassakis (1999)

At first order, we neglect the spatial derivatives of Φ̂. One may show (Mei, 1989, chapter 3, see also
Ardhuin et Herbers 2002) that if the bottom slope is of order ε as well, other terms, in addition to ϕ̂ are
required to satisfy the kinematic boundary condition at the surface and we get the action conservation
equation,    
∂ E E
+∇· Cg = 0. (16.13)
∂t σ σ
Diffraction appears with the second-order terms, and waves tend to turn towards regions with lowest
wave amplitudes.
Berkhoff (1972) used the approximation, now known as the mild slope approximation, that ϕ verifies
the polarization and dispersion relations of linear waves over a flat bottom. After some calculations, we
obtain (see for instance Mei, 1989, chapter 3) the so-called mild-slope equation (or Berkhoff’s equation).

Cg
∇ · (CCg ∇ζ) + ω 2 ζ = 0. (16.14)
C
The mild slope equation (16.14) is an extension of the Helmoltz equation to a mild bottom slope.
This equation is widely used in coastal engineering for determining harbor agitation, using finite elements
numerical models. Results of this type of model are shown in figure 16.5.
Radder (1979) proposed a parabolic approximation of the elliptic equation, by neglecting the ζ gradi-
ents in the propagation direction, which conserves the diffraction effects. Such a model (called refraction-
diffraction model) has been used for swell forecasting over the Californian Coastline (http://cdip.ucsd.edu),
from offshore prediction. The Californian continental shelf is indeed narrow enough to neglected the local
generation of waves.
However, contrary to what happens in the vicinity of coastal structures, the diffraction can generally
be neglected in this region O’Reilly and Guza (1991); Peak (2004). For example, during the 3-month
long Near Canyon EXperiment (NCEX) offshore La Jolla, California, only one swell had been measured
with a frequency low enough to justify the use of a model solving Berkhoff’s equation. In this situation
the Berkhoff’s model provided slightly more accurate results than the ray tracing method and only at
a limited number of locations. These are the locations where the wave field varies strongly at the scale
of the wavelength, with a variation of a factor 5 in the wave amplitude across the Scripps canyon wall,
over a distance shorter that the wavelength (figure 16.6).
One may often just use the refraction computation given by eq. (16.6). This kind of approach take
into account the details of the bathymetry and of the spectral shape, which is very important form
complex coastline.
164 CHAPTER 16. LINEAR SHOALING, REFRACTION AND REFLECTION

1.1
Observations
1 Refraction
0.9 MSE
MMSE
0.8 NTUA5
0.7 Ref-dif

Hs (m)
0.6
0.5
0.4
0.3
0.2
0.1
0
32 33 34 35 36 37
Buoy ID number

Figure 16.6: Wave observations and modeling during NCEX.

The significant wave heights (Hs ) for the frequency band 0.04-0.08 Hz are computed from the spectra
measured from 1:30 to 4:30 pm UTC, on November 30 2003. The buoy location are shown on figure 16.4.
The models used are: a ray back-tracing (”refraction”, O’Reilly & Guza 1991) three elliptical models
and a parabolic approximation of the Berkhoff equation (”Ref-dif” Kirby 1986). The three models use
the same numerical code and solve Berkhoff’s equations (”Mild Slope Equation” or MSE, 1972), their
version modified by Massel (MMSE, 1993) and their extension by the coupling of three evanescent modes
and a local mode (NTUA5, Athanassoulis et Belibassakis 1999). The difference that appears between
the refraction model and the elliptical models at buoys 32, 36 and 37 is related to the wave transmission
by tunnel effect across the canyon, despite its large depth (see also Thomson et al. 2005). This tunnel
effect can not be represented in the geometric optics approximation over which relies the ray tracing of
the refraction model. Figure from Magne et al. (2007).

16.4 Reflection
Any variations of the water depth or current velocity - the wave guide parameters - result in partial
reflections. These reflections are significant only if the above cited variations are large over the wave
length scale. This is the case when the waves approach the coastline.

16.4.1 Reflection at the shoreline

Hence, the magnitude of the reflection, that can be characterized by the ratio R = ar /ai of the reflected
and incident wave amplitudes, increases strongly with the bottom slope β and with the wave period
T . Miche (1951) studied monochromatic waves over a constant bottom slope and reported that R2 is

Figure 16.7: Energy Reflection coefficient R2 as a function of the Miche number M , From Elgar et al.
(1994), copyright American Meteorological Society.
16.4. REFLECTION 165

roughly equal to the coefficient

0.0016g 2 tan5 βT 4
M= 2
, (16.15)
H∞
where H∞ is the offshore wave height, in the case M < 1 and R = 1 for M > 1. The only precise study
published on the topic of in situ random waves shows that, using Hs instead of H, Miche parameter M
effectively gives an order of magnitude of the reflection coefficient for swells and wind seas (figure 16.7).
The study of Elgar et al. (1994) suggests that R2 > M for M < 0.05 and R2 < M for M > 0.1. A
better approximation, R2 = min{0.007(log10 (M ) + 5) + 0.2M, 1}, has been implemented by Ardhuin and
Roland (2012) in the WAVEWATCH III model. Besides, for a given sea state, the reflection decreases
toward high frequencies.
Casual observation of wave reflection from a beach suggests that, in addition to the reflection there is
also a generation of short wave components, that are radiated toward the open ocean. To our knowledge,
this has not yet been discussed in scientific publications.

16.4.2 Reflection by underwater topography

The variations h′ of the water depth at the wave length scale also impact waves. One way to represent
this effect in the spectral evolution equation consists in decomposing the topography variation h′ in sine
waves of wavenumber l, X
h′ (x) = Bl (ex) eil·x , (16.16)
l

and to calculate the interaction of the wave spectrum with each sine wave. Strictly speaking, we obtain
a wave forcing by the topography. At fist order in bottom slope (ε = lh), we obtain a resonance between
two waves with wave numbers k et k′ , such as k = k ′ , that exchange energy thanks to bottom ripples
with wave number l = k − k′ . This leads to an energy transfer in other directions (figure 16.9).

Figure 16.8: Bathymetry of a region with underwater dunes in the south of the North Sea.
On the spectrum (b) of this bathymetry, the contours indicate the value log10 [4πFB (l)] with FB the
bottom spectrum. The circles indicate the components l that interact with waves of 12.5 s period coming
from the North-West (wave number k) for 3 current speeds, -2, 0 et 2 m/s. .

The local effect of this Bragg diffusion is a modification of the directional wave spectrum. If the
bottom topography is dominated by large scales (l << k), the result is an increase of the waves directional
spreading, that corresponds to a shortening of the wave crest length, rendering the wave field appearance
messier.
For a topography with a large spectral density at l = 2k, the wave are back-scattered (Heathershaw,
1982) and the wave height decreases in the initial direction of wave propagation. For random waves
166 CHAPTER 16. LINEAR SHOALING, REFRACTION AND REFLECTION

Figure 16.9: Example of spectral evolution caused by bottom reflection for the wave spectrum shown in
figure 16.8 and applied in 20 m depth
(a) computation domain, (b) incident spectrum imposed at point F, (c) and (d) evolution term of the
spectrum with and without current, (e) spectrum at point O, 40 km inside the domain after 5 hours of
propagation. The frequency is here the relative frequency σ/(2π).

this effect takes the form of a bottom induced scattering term Sbscat (Ardhuin and Magne, 2007), which
typically produces a broadening of the directional spectrum when important depth variations are present
at the scale of the wavelength.

16.4.3 Summary
In Shallow water (D < 0, 5L), waves are influenced by the bottom in addition to wave-wave interaction
and interaction with atmosphere. Leaving apart wave breaking that occurs in the direct vicinity of the
beach, the bottom effect depends on the relative amplitude of the waves and of the bottom topography.
All these effects are supposed independent of each other, and independent of the wave-wave interaction
or of the interaction between waves and atmosphere (e.g. Komen et al., 1994). This leads to a spectral
evolution equation that takes into account all the processes as source terms. This is thus an extension
of the deep-water action balance given by eq. (10.42) with additional source terms,

S = Sin + Snl + Sdis + Sfric + Sbscat + · · · (16.17)

Chapter 17

Nonlinear wave shoaling

17.1 Introduction
Chapter 3 has shown how random waves can be usually treated as a sum of linear waves with random
phases. Chapter 8 presented how it evolves under the influence of weak non-linearities as described by a
wave action equation. Non-linearities actually become stronger in shallow water when the waves are less
dispersive (kD << 1), and we have to reconsider the non-linear effects. In chapter 2 we have introduced
the two small parameters ε = ka and δ = a/D. The nonlinear effects that we will discuss here occur
when the parameter Ur = δ 3 /ε2 = a/k 2 D3 is of order 1. This parameter was introduced by Ursell (1953).
The regime Ur ≪ 1 corresponds to the weak nonlinearity already discussed in chapter 8.
The stronger non-linearity that occurs for Ur = O(1) correspond to a near-resonance at second order,
i.e. k3 ≃ k1 + k2 and f3 = f1 ± f2 , with the plus sign giving super-harmonics of higher frequency, and
the minus sign giving sub-harmonics of very low frequency or infragravity waves. These interactions can
also be seen as a non-linear wave scattering by the bottom topography with k3 = k1 + k2 + kb where kb
is the wavenumber of the topography (Liu and Yue, 1998; Groeneweg et al., 2015).
Instead of going back to the full Euler equations, several approximations have been proposed to
simplify the problem but keep the nonlinearity. In particular, assuming weak dispersion (kD << 1)
allows to derive a simplified set of equations.

17.2 Boussinesq and KdV equations

Both the Boussinesq and Korteweg and de Vries (1895, KdV for short) equations were first derived by
Boussinesq (1872) for waves propagating in one direction. We will here use the form given by Peregrine
(1967),
D2 ∂ 3 ζ
 
∂ζ p ∂ζ 3 ∂ζ
+ gD + ζ + = 0. (17.1)
∂t ∂x 2D ∂x 6 ∂x3
This is asymptotically valid for kD << 1 and ka << 1. For a constant depth D, solutions to the KdV
equation include quasi-periodic recurrent amplitudes (Fermi et al., 1955) caused by the near resonances
k3 ≃ k1 + k2 which exist when kD << 1.
Interestingly, the KdV equation can be solved exactly using the inverse scattering transform, which
expresses the solution as a superposition of interacting waves trains and solitary waves also known as
cnoidal waves or solitons (Osborne et al., 1996). Among these cnoidal waves, the classical solitary wave
solution is,
a
ζ= hp i (17.2)
2
cosh 3a/D3 (x − Ct)

with a propagation speed

p  a 
C= gD 1 + . (17.3)
2D
In this solitary wave, the dispersive effect of the finite water depth is exactly canceled by the dispersive
effect of the finite amplitude, so that the wave propagates without changing form.
The two-dimensional version of the KdV equation is the Boussinesq equation with many forms de-
veloped to improve on its dispersive properties, i.e. trying to extend it beyond kD ≪ 1 (Nwogu, 1993;

167
168 CHAPTER 17. NONLINEAR WAVE SHOALING

Elevation (m) 0.5

-0.5

-1

-1.5

-2
0 50 100 150 200
x (m)

Figure 17.1: Solitary wave given by eq. (17.2) that is a solution of the KdV equation (17.1) in the case
D = 10 m

Nadaoka et al., 1997). Other extensions for finite amplitude have been performed by Serre and others
(Lannes and Bonneton, 2009; Dias and Milewski, 2010). Some of these equations have been called fully
nonlinear Boussinesq equations but, although they are indeed valid for finite amplitude, they are not
correct for waves that are nearly breaking and always have round crests, not sharp like a nearly breaking
wave.

17.3 Wave evolution

As waves shoal, the reduction in group speed Cg comes with an increase in the significant wave height
Hs , in particular for a shore-normal incidence θ = 0. That effect is less severe for oblique propagation

05
. (a) 0.5 (b)
0.0 0

-0.5 -0.5
z (m)

-1.0 -1

-1.5 -1.5

0 0.5 1 1.5 2 2.5 3 3.5 0 1 2 3 4

x (m) x (m)

1
0
(c)
z (m)

-1
-2
-3
0 5 10 15 20 25 30 35 40 45
x (m)
0 5 10 15 20 25 30 35 40 45
2
(d)
z (m)

0
-2

Figure 17.2: Wave profiles computed using the numerical solution method byDean (1965) and Dalrymple
(1974) at 60th order for a 3 m water depth, in the case of deep water waves in (a) and (b) with a period
T = 1.5s (kD ≃ 5), and shallow water in (c) and (d) with T = 8s, kD ≃ 0.45. The arrows represent
orbital velocities. Waves in (a) and (c) are moderately nonlinear (Uc /C ≃ 0.3), whereas those in (b) and
(d) are nearly breaking (Uc /C = 0.97).

as the conserved energy flux is Cg Hs2 cos θ/16 and cos θ increases due to refraction. The wavelength L
also reduces as C = L/T slows down and the period remains constant. This increase in height H and
 17.3. WAVE EVOLUTION 169

decrease in wavelength L make the steepness H/L increase.

17.3.1 Waves over a flat bottom

For a monochromatic wave over a flat bottom, Miche (1944a) studied the shape of the waves of maximum
steepness. In the frame of reference moving with the phase speed C, it appears that a singularity
develops when the orbital velocity Uc approaches C. Figure 17.2 shows the difference between waves
with Uc /C ≃ 0.3 and Uc /C = 0.97. When Uc /C ≃ 0.3, waves in deep water look linear whereas waves in
shallow water look like cnoidal waves. As Uc /C approaches 1, the crest become triangular with an angle
of 120◦ , as was already found by Stokes for the deep water case.
Hence, using Uc /C = 1 as a sufficient condition for wave breaking, Miche used an analytic stream-
function expansion from the wave crest to find that the maximum wave height Hmax is

Hmax /L ≃ 0.14 tanh(kD) (17.4)

In the limit kD → 0, this Miche limit is Hmax /D ≃ 0.28π = 0.87. The deep water limit kD → ∞ gives
the previously known kH/2 = 0.44. Miche’s approximation is very accurate. Figure 17.3.a shows that it
can be improved a little by using

Hmax /L ≃ 0.10 tanh(kD) + 0.0298 tanh2 (kD). (17.5)

It should be noted that, so far, we have used H = 2a and E = a2 /2. In shallow water this does not

1.0
Streamfunction 80th order
=kHmax / tanh (kD)

Miche (1944): Y=0.14 X

0.8
Y=Hmax / L

0.6

Streamfunction 80th order

0.4 Ruessink et al. (2003)
E from 80th order streamfunction
E

(a) (b)
0.2
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
X=tanh(kD) X=tanh(kD)

Figure 17.3: Steepness of nearly breaking waves, Y = Hmax /L. The original breaking criterion of
Miche (1944) is recalled. A new criterion, using a second order polynomial fit for H/L as a function of
X = tanh(kD) is given. The bottom panel shows the alternative steepness parameters β defined from the
wave height, or βE defined from the wave energy. In the case of Ruessink et al. (2003), his γ parameter is
interpreted as Hmax /D, and transformed to βE , using the peak wavenumber kp to estimate k̄ (adapted
from Filipot et al., 2010a).

hold anymore, and the error on the wave height can be a factor 3 or more. Even if it is probably less
important on a sloping bottom, as we will see below, if one wants to apply a limit on the wave height
when using wave energy, we can first define a generalized maximum wave steepness from the height

β = kHmax tanh(kD) (17.6)

or from the wave energy

p
βE = k Emax /8 tanh(kD). (17.7)

This parameter varies with kD in a way that is very close to the estimated variation of the model
parameter γ used in the wave breaking parameterization by Ruessink et al. (2003). This γ parameter
is often interpreted
p as Hmax /D, the fact that it follows βE and not β suggests that it should rather be
interpreted as Emax /8/D.
170 CHAPTER 17. NONLINEAR WAVE SHOALING

Figure 17.4: Examples of (a) spilling, (b) plunging breakers, taken from Coastal Engineering Manual -
Part II (2002). (c et d) are not so nice pictures of a surging breaker on a steep beach and on a rocky
point (Ruscumunoc beach, Plouarzel, France).

17.3.2 Sloping bottom and the Iribarren number

Besides the water depth, the local bottom slope also has an influence on the shape of breaking waves,
as shown in figure 17.4. These different wave breaking shapes can be predicted using the surf similarity
parameter, proposed by Iribarren and Nogales (1949) and further discussed by Battjes (1974). This is
also called the Iribarren number,
tan β tan β
ξ0 = p =p (17.8)
H0 /L0 2πH0 /(gT 2 )

where tan β is the slope of the shoreline, H0 and L0 are the height and wavelength of the waves in deep
water (before shoaling), which is not always easy to define since shoaling is not the only process involved
in transforming the waves from deep water to the shoreline.
Depending on the value of ξ0 the breaking of waves takes three forms
ˆ spilling for ξ0 < 0.5. Foam is generated at the crest of the wave and spills over the front face.
Apart for this foam, the crest keeps its front-back symmetry, and can later evolve in an asymmetric
saw-tooth shape.
ˆ plunging for 0.5 < ξ0 < 3.3. This is characterized by a ballistic jet of water ahead of the crest,
creating a tube of air trapped by the water. The jet produces further splashes as it partially
bounces on the sea surface.
ˆ surging for ξ0 > 3.3: in that case the wave breaks right at the shoreface.
For the steepest slopes, waves can be reflected without breaking (Carrier and Greenspan, 1958).
On sandy beaches, which generally have moderate slopes, one can distinguish the surf zone where
the waves break, from the swash zone which is the region of the beach which gets wet and dry during a
wave cycle.

17.3.3 Wave shapes

The shape of a waves evolves from nearly symmetric front-back shapes outside of the surf zone with a
strongly asymmetric shape in the inner surf zone that resembles a hydraulic jump with a rapid rise of
 17.3. WAVE EVOLUTION 171

16 16
x=-2.4 m x=0 m
x=1.2 m
11 11

6 6

1 1
Surface elevation (cm)

-4 -4
0 1 2 3
t/T t/T t/T
16 16
x=2.4 m x=3.6 m
x=4.8 m
11 11

6 6

1 1

-4 -4

Figure 17.5: Time series of surface elevation before and after wave breaking, for monochromatic waves
with T = 2.2 s and H0 = 0.115 m over a laboratory beach of constant slope β = 1/35 (taken from Cox,
1995). These are spilling breakers with an Iribarren number ξ0 = 0.23. The distance is measured from
the point x = 0 of incipient breaking.

the surface elevation on the forward face of the waves, as shown on figure 17.5 for laboratory spilling
breakers. These wave shapes are fairly different from the flat bottom solutions shown in the previous
section. This shock-like behavior has been particularly well studied in the shallow water limit(Bonneton
et al., 2004). On the contrary top-bottom asymmetry decreases. The front-back asymmetry is associated
to a strong acceleration of the flow under the crest, which is very important for sediment transport. In
particular, Hoefel and Elgar (2003) have shown how the asymmetry could explain the onshore migration
of sand bars.

17.3.4 Wave spectra and bi-spectra

In the case of random waves, the wave evolution is more complex. Looking at the energy only, it is
possible to have a good approximation of the evolution of the wave energy with a wave dissipation
parameterized following the hydraulic jump model presented in chapter 8. This was first developed in
models for the total wave energy by Battjes and Janssen (1978) and Thornton and Guza (1983). An
example result from the model by Thornton and Guza, is shown in figure 17.6.
Although the wave height is a key parameter, the important application to sediment transport and
beach morphodynamics requires more than information on energy only. First of all, orbital velocities at
the bottom, which determine the re-suspension of sediments, require a knowledge of the wave spectrum, or
at least wave periods. Hence these energy models have been extended in the form of wave action equations
by defining suitable wave breaking parameterizations (e.g. Filipot and Ardhuin, 2012). However, the
spectral evolution also exhibits the development of harmonics and their progressive release over varying
bathymetry. These processes and the reproduction of the important asymmetry of the waves requires
an investigation of the relative phase of the wave components. The relative phases of the different
components are given by the bi-spectrum, which is defined as a generalization of the spectrum (our
equation 3.3),
B(ωi , ωj ) = ⟨am,i am,j am,i+j ⟩. (17.9)
This bi-spectrum is a complex number. Just like the spectrum is the Fourier transform of the 2-point
correlation function, the bi-spectrum is the 2D Fourier transform of the 3-point correlation function of a
time series,

= ⟨ζ(t)ζ(t + τ1 )ζ(t + τ2 )⟩
S(τ1 , τ2 ) (17.10)
Z ∞Z ∞
1
B(ω1 , ω2 ) = S(τ1 , τ2 )e−iω1 τ1 −iω2 τ2 dτ1 dτ2 . (17.11)
(2π)2 −∞ −∞
172 CHAPTER 17. NONLINEAR WAVE SHOALING

Depth (m)
10

20
0 100 200 300 400 500
3

Hs (m) 2

0
0 100 200 300 400 500
CgE/Cg0 E0(%)

100

50

0
0 100 200 300 400 500
cross-shore distance (m)

Figure 17.6: Evolution of significant wave height during breaking

Results of the model by Thornton and Guza (1986) for a wave period T = 12 s and 20◦ incidence angle
in 20 m depth. Waves propagate over a schematic beach with a mean slope β = 0.04 and a bar at
x = 300 m.

More properties of the bi-spectrum in the context of ocean waves are discussed in Hasselmann et al.
(1963). We will particularly note that B is zero for a Gaussian wave field. The skewness and asymmetry
of the time series are given by integrals of the bi-spectrum,
XX X
⟨ζ 3 (t)⟩ = 12 R [B(ωn , ωl )] + 6 R [B(ωn )] (17.12)
n l n
* 3 +
∂ζ XX X
3/2
= 12 ℑ [B(ωn , ωl )] + 6 ℑ [B(ωn )] / ⟨ζ 2 (t)⟩ , (17.13)
∂t n n
l

where R and ℑ stand for the real and imaginary parts respectively.
For practical purposed the bi-spectrum is generally normalized to obtain a bi-coherence b(ω1 , ω2 ),
with values between 0 and 1, and a bi-phase β(ω1 , ω2 ),
|B(ωi , ωj )|
b(ωi , ωj ) = rD ED E, (17.14)
2 2
|am,i am,j | |am,i+j |
 
ℑ[B(ωi , ωj )]
β(ωi , ωj ) = arctan . (17.15)
R[B(ωi , ωj )]
In the absence of phase-correlations between different components, b = 0 and all waves are free and
propagate at the linear phase speed. The other extreme are the harmonics of a monochromatic waves for
which b = 1. In that case the harmonics are bound to the underlying wave and propagate at its velocity,
different from the linear phase speed.
Real waves in shallow water are intermediate between these two extremes and harmonics are partially
locked and can be released as free waves by changes in the bottom topography (e.g. Sénéchal et al., 2003).
Elgar and Guza (1985) have made one of the first analyses of the bi-spectral evolution on a gently
sloping beach as part of the Nearshore Sediment Transport Study (NSTS) experiment. Figure 17.7
shows the evolution of time series of surface elevation, estimated from bottom-mounted pressure gauges,
and the statistical representation of phase-relationships give by the bi-coherence. From a superposition
of uncorrelated waves in 9 m depth, the spectrum evolves with the generation of harmonics that, on
this particular beach, remain locked in phase and exchange energy. In 3.9 m depth, the bi-coherence
already has 3 peaks at 0.06, 0.12 and 0.18 Hz that show the non-linear interaction of the main peak and
its harmonics. In other cases with a more complex topography, Sénéchal et al. (2003) showed the the
harmonics can become free and the bi-coherence can be reduced after wave propagation over a bar.
A representation of this effect as a ’triad’ interactions source term in the wave action equation can
be expressed theoretically from the bi-spectrum (Herbers and Burton, 1997; Becq, 1998). However, the
17.4. INFRAGRAVITY WAVES 173

D=9m
E(m 2 /Hz)

D=1.1 m 0

Depth (m)
2
D=3.9m
4
f2 (Hz)

6
D=2.7m
8
0.06 0.14 0.22 0.3 0.38
0 100 200 300 400
Cross-shore distance (m)
D=2.4m
E(m 2 /Hz)

D=2.7 m D=3.9 m D=9.0 m

D=2.7m
f2 (Hz)

D=1.0 m
0.06 0.14 0.22 0.3 0.38 0.06 0.14 0.22 0.3 0.38 0.06 0.14 0.22 0.3 0.38
frequency f1 (Hz)

Figure 17.7: Spectral and bi-spectral evolution recorded on February 2, 1980.

The position of bottom-mounted pressure gauges is shown in the upper-middle panel, from 9 m depth
to 1.1 m depth. Short pieces of the time series of surface elevation at the different sensors is shown in
the right panel. Left and bottom panels show spectra E(f1 ) and and associated bi-coherence b(f1 , f2 ) at
selected locations. Adapted from Elgar and Guza (1985).

integration of an evolution equation for the bi-spectrum comes at a considerable cost, although this is
still much less than the cost of a phase-resolving models. As a result, most spectral wave models have
adopted relatively crude parameterizations of the triad interactions (e.g. Eldeberky and Battjes, 1995).

17.4 Infragravity waves

17.4.1 Observations
In the same way that sum interactions f = f1 +f2 give rise to harmonics, difference interactions f = f1 −f2
also produce very low frequency components that are called infragravity waves, that we will often shorten
as ’IG waves’. This beating pattern was first investigated by Munk (1949). Infragravity water level
oscillations are often dominant in the swash, with a typical range of periods between 30 s (0.033 Hz) and
5 minutes (0.003 Hz). Figure 17.8 shows spectra of surface elevation from off-shore of the surf zone to
the swash zone, with a typical wind-sea spectrum transforming into a spectrum dominated by motions
at frequencies below 0.04 Hz. These long period oscillations of the water level are readily observed by

D=8.8 m D=1.7 m
E(f) (cm 2 /Hz)

D=3.7 m D=0 m

0.04 0.12 0.20 0.28 0.36 0.04 0.12 0.20 0.28 0.36
Frequency (Hz)

Figure 17.8: Spectra of surface elevation recorded in Santa Barbara on February 4, 1980.
Measurements at D = 0 m were performed with a run-up wire: the elevation is not at a fixed position
in x but along the beach profile. Adapted from Elgar and Guza (1985).
 174 CHAPTER 17. NONLINEAR WAVE SHOALING

video systems, as shown in figure 17.9. Their analysis is very important for the understanding of coastal
hazards.

ocean

cross-shore position (m)

dry beach

time (s)
Figure 17.9: Measurement of run-up using a video system
Left: Snapshot of the beach at Duck, North Carolina, with the white line marking the position of a
transect. Right: time-stack of pixel grey values along this transect. The red line marks the detected edge
of the water: using the beach profile h(x) this gives a time series of run-up. Adapted from Stockdon
et al. (2006).

The contribution of IG waves to the run-up is relatively larger for small beach slopes (Stockdon et al.,
2006), with some exceptions. The highest measured IG wave heights, around 2.5 m, was recorded on the
cliff of the small island of Banneg, France, as shown in figure 17.8.
Time: Feb. 10, 2009, (hour:min)
04:00 04:01 04:03 04:04 04:06 04:07 04:09 (b) 13
maximum level
11
(a) 12
recorded at P2: 11.8 m
10 11
Elevation (m above chart datum)

9 10
9
8
msl at P3: 7.3 m 8
7
7
P2 6
P2
z (m)

11 5
(b) 4
10
3
9
2
8

7
P3
Chart datum: 0
1
0
P2
05:00 06:00 07:00
−1
Time: Feb 10, 2009, 3h (min:s)
50 40 30 20 10 0
x (m)

Figure 17.10: Large infragravity wave heights measured on Banneg Island, France.
Left: (a) Time series of pressure converted to surface elevation (assuming hydrostatic pressure) from the
sensors at the top of the cliff (P2, red) and the bottom of the cliff (P3, blue). A typical burst of high
water levels, lasting 90 s, is enlarged in (b). Center: cliff profile and schematic of water level at the time
of highest recorded pressure. Right, picture of the cliff from P3, at low tide. Adapted from Sheremet
et al. (2014).

17.4.2 Theories for IG waves generation

A first theoretical explanation for the formation of infra-gravity waves was given by Whitham (1962) and
Longuet-Higgins and Stewart (1962), looking at the flow response to the passage of wave groups. They
actually gave two derivations, the first follows the perturbation method used in chapter 2. The second
method, which we will use here, is only applicable when wave groups are long compared to the water
depth, and uses the wave-averaged mass and momentum equation, (10.33) and (10.35). Neglecting the
Coriolis force, advection, and surface and bottom friction, these become, for waves propagating in the x
17.4. INFRAGRAVITY WAVES 175

direction only
rad
∂M ∂ζ ∂Sxx
= −ρw gD − , (17.16)
∂t ∂x ∂x
∂M ∂ζ
= −ρw . (17.17)
∂x ∂t
rad
These are exactly the equation of long waves (i.e. in the shallow water limit),forced by Sxx . For a
rad
narrow wave spectrum, we note that Sxx is attached to the wave groups and travels at the speed Cg ,
the solution of such a forced wave is bound to travel with the forcing, so that the time derivative is equal
to −Cg times the horizontal gradient.
rad
We recall eq. (9.25) that gives Sxx = ρw gE(2Cg /C − 0.5), which is always positive. Taking a
rad
particular case of modulation of energy E = E0 +E1 cos[Kx(1−Cg t)], we have Sxx = S0 +S1 cos[Kx(1−
Cg t)] we look for solutions of the form, M = M1 cos[Kx(1−Cg t)] and ζ = A1 cos[Kx(1−Cg t). Replacing
in our mass and momentum equation, this gives,

Cg M1 = −ρw gDA1 − S1 (17.18)

M1 = −Cg ρw A1 , (17.19)

which gives the transport and elevation amplitudes of the bound long wave,

CgS1
M1 = − (17.20)
gD − Cg2
S1
A1 = − . (17.21)
ρw (gD − Cg2 )

This solution thus gives a total transport M1 by the long and short waves that is partly compensating
the modulation of the wave-induced transport ρw Cg E1 /C. Because the compensation is not exact, the
net divergence of the flow is driving a change in mean sea level, with a lower level where the wave energy
E1 is larger. It should
√ be noted that this bound wave response has a singularity in the limit of shallow
water when Cg → gD, this is the limit into which the interaction of two short waves and the long wave
becomes resonant.

wave group break point

z =0

bound long wave mean sea level

propagation speed = group speed z=

Nota bene: vertical scale is strongly exaggerated

Figure 17.11: Schematic of bound infragravity waves associated to wave groups.

This 180◦ phase shift between the wave energy and the long waves was indeed verified in some
experiments, typically just outside the surf zone on gently sloping beaches, but it is typically less in the
surf zone (Elgar and Guza, 1985). This may be due to the difference between the flat bottom solution
where the long wave is exactly bound, and the real case of a varying depth in which the long wave is
partially free. For a general wave spectrum with waves in all directions, the theoretical bound response
is integral over all possible pairs of waves, and the theory is well verified over a flat bottom (Herbers
et al., 1994). The free and bound IG wave energy can be separated using a bi-spectral analysis (Herbers
et al., 1994). Free IG energy usually dominates even in 8 m depth, and even more so in deeper water
(Herbers et al., 1995).
It is also possible that other mechanisms generate long period oscillations. In particular, over coral
reefs, the modulation of the position of the initial wave breaking with wave groups can produce mean
sea level oscillations that are rather in phase with the wave groups (Symonds et al., 1982).
176 CHAPTER 17. NONLINEAR WAVE SHOALING

17.4.3 Free wave radiation from the coast

Whatever the details of the generating process, it is clear that bound waves traveling with the incoming
wave groups are released and partially dissipated in the surf zone where the groups are dissipated.
Because the phase speed of the released waves is much larger than that of the incident waves, most of
the energy is strongly trapped by refraction along the coast.
Observations compiled by Ardhuin et al. (2014) suggest that the free IG wave spectrum near the
coast can be parameterized as,
2
AIG = Hs Tm0,−2 (17.22)
2 2
kg (AIG /4) 1.5
EIG (f ) = 1.2α12 [min(1., 0.015Hz/f )] (17.23)
Cg 2πf ∆f
EIG (f, θ) = EIG (f )/(2π), (17.24)

where α1 = 6 × 10−4 s−1 might vary with bottom topography, and ∆f = 0.0279 Hz. The k/Cg factor
accounts for the shoaling of a broad directional spectrum, while the frequency shape of the spectrum is
given by the other terms. In the shallow water limit, i.e. kD going to zero, this spectrum is constant up
to f = 15 mHz and decreases like f −1.5 for higher frequencies. In that frequency range, this asymptote
is identical to the form tanh(kD)−1.5 given by Godin et al. (2013). The differences at lower frequencies
may be due to the fact that, in particular for f < 2 mHz, the measured wave field in the open ocean is
mostly driven by atmospheric pressure and not IG waves radiated from shorelines (Filloux, 1980; de Jong
et al., 2003).
Eq. (17.24) gives an estimate H
b IG of the infragravity wave height,
s
Z 30 mHz
HIG = 4 E
bIG (f )df . (17.25)
0.05 mHz
Chapter 18

Bottom boundary layer: processes

and parameterizations

Bottom friction is a complex process because the bottom is generally a complex medium that can combine
sand, mud, rocks, plants and animals. Also the bottom topography can strongly vary in the case of mobile
sediments, with wave-generated ripples. Finally, flow in the bottom boundary layer may be turbulent,
which requires a turbulence closure of the flow equations.
The wave bottom boundary layer connects the region of potential flow where the oscillatory wave-
induced velocity is given by eq. (2.28) to the sea bed where the velocity is zero. This boundary layer is
very thin, typically 10 cm or less. This is much less than the boundary layer of the mean current. As a
result, when a current is present, the wave boundary layer also affects the friction for the current.
We will consider here the case of a linear monochromatic waves. The velocity above the boundary
layer is given by eq. (2.28),
σa
u+ (x, t) = cos(kx − ωt + Θ0 ), (18.1)
sinh(kD)

Because the wave propagates at the phase velocity C, the horizontal advection of any quantity X by the
velocity u is u∂X/∂x, which can be neglected compared to ∂X/∂t because the first term is u/C smaller
than the second, and u/C is typically less than 0.2. Let us define uδ (x, z, t) = ⟨u(x, z, t)⟩ − u+ (x, t), in
which the brackets ⟨.⟩ represent a Reynolds average, over the realizations of the turbulent flow.
The conservation of the horizontal momentum component reads

∂e
u 1 ∂p ∂u+
=− − +G (18.2)
∂t ρw ∂x ∂t

with G the divergence of the vertical momentum fluxes due to viscosity or turbulence,

e ∂ ⟨u′ w′ ⟩
∂2u
G=ν + . (18.3)
∂z 2 ∂z

Because the thickness of the boundary layer δ is much less than the wavelength, the pressure gradient
in the boundary layer equals the pressure gradient outside the boundary layer, which is balanced by the
acceleration,
1 ∂p σ2 a
− = sin(kx − σt) = ∂u+ /∂t. (18.4)
ρw ∂x sinh(kD)

This is another way to write the Bernoulli equation (see also Mei, 1989).
Replacing (18.4) in (18.2) one obtains
∂uδ
= G, (18.5)
∂t
with the matching condition
uδ → 0 for z ≫ δ. (18.6)

177
178 CHAPTER 18. BOTTOM BOUNDARY LAYER: PROCESSES AND PARAMETERIZATIONS

18.1 Viscous solution

The laminar situation is very interesting because it has an analytical solution that allows to show general
properties of the boundary layer. In the laminar case G = ν∂ 2 ue/∂z 2 and the solution is
σa
uδ (x, z, t) = e−z+ cos (kx − σt − z+ ) (18.7)
sinh(kD)
with p
z+ = (z + h)/ 2ν/σ = (z + h)/δ, (18.8)
in which we have used the definition p
δ≡ 2ν/σ (18.9)
that gives an order of magnitude of the boundary layer thickness.
The full velocity profile is thus u = u+ + uδ . Because of the −z+ term in the phase, the velocity u
in the boundary layer has a phase ahead of the the free stream velocity u+ . This phase advance grows
with z+ but it only impacts a smaller fraction of velocity for larger values of z+ . This advance is due to
the importance of the friction term in the momentum balance, which gets added to the pressure gradient
and inertia. A linear friction for laminar flow is proportional to −u and thus in phase with the pressure
gradient ∂p/∂x, whereas inertia ρw ∂u/∂t has a phase lags of 90◦ .
From the velocity solution, we obtain the dissipation rate of wave energy per unit surface, as given
by the work of the viscous stress. Once divided by the wave energy Et per unit surface, one gets the
attenuation coefficient
⟨ρw νu∂u/∂z⟩
βν = (18.10)
ρw ga2 /2
Using the solution in eq. (18.7) we get

∂uδ σa e−z+
=− p [cos (kx − σt − z+ ) − sin (kx − σt − z+ )] (18.11)
∂z sinh(kD) 2ν/σ

in which only the first term correlates with u, and gives

σ3 δ
βν = − . (18.12)
2g sinh2 (kD)
The spectral dissipation rate is thus of the following form

σ3 δ
Sbfric (k) = − E(k). (18.13)
2g sinh2 (kD)
As we have seen in chapter 9, the dissipation of wave energy also comes with a loss of momentum at
the rate βν E/C for a monochromatic wave train. So what happens to that momentum?

18.2 Streaming
Because momentum is conserved, we have to investigate where it goes, and thus look into the mean
current. For this we also need the vertical velocity wδ associated to uδ . It is simply given by the
conservation of mass ∂wδ /∂z = −∂uδ /∂x, with the result
2kδσa −z+
wδ = − e [cos (kx − σt − z+ ) + sin (kx − σt − z+ )] . (18.14)
sinh(kD)
The phase shift between the boundary layer and the free stream has interesting consequences. Indeed,
the conservation of mean flow momentum is
∂uw ∂2U
=ν 2. (18.15)
∂z ∂z
Following Phillips (1977), one easily gets

σ 2 a2 kδ
 
2 −z+ −z+ −2z+
(u + uδ ) (w + wδ ) = e sin(z+ ) − 1 + 2e cos z+ − e . (18.16)
4 sinh2 (kD) δ
18.2. STREAMING 179

6
U
5
UL
4
z + = (z+h)/δ

0
0 1 2 3 4 5 6
U/ [k σ a / 4 sinh2(kD)]
2 2

Figure 18.1: Mean current known as ‘streaming’, induced by a monochromatic wave train.
Profiles of the Eulerian
p mean and Lagrangian mean velocities in the boundary layer for a constant
viscosity ν, with δ = 2ν/σ.

As z+ goes to very large values (z + D ≫ δ), the momentum flux goes to

ρw σ 2 a2 kδ
lim ρw uw = − . (18.17)
z+ →∞ 4 sinh2 (kD)
This is exactly the momentum lost by the waves per unit time and unit horizontal surface. Hence, the
momentum lost by the waves is taken up by the mean flow, and the waves accelerate a near boundary
‘streaming current’ that was first reported by de Caliginy (1878). The steady state corresponds to a
situation in which the bottom friction for the mean current passes on this momentum to the sea floor
(Longuet-Higgins, 2005).
The integration of eq. (18.15) gives

σa2 k
3 − 2 (z+ + 2) e−z+ cos z+ − 2 (z+ + 2) e−z+ sin z+ + e−2z+ .
 
U (z) = 2 (18.18)
4 sinh (kD)
We can also update our estimation of the Stokes drift from chapter 9, now using u
e and w.
e This gives
L
a mass transport velocity U = U + Us (Longuet-Higgins, 1953),

L σa2 k
5 − 8e−z+ cos z+ + 3e−2z+ .
 
U = (18.19)
4 sinh2 (kD)
We note that this mass transport velocity at the top of the boundary layer is independent of the value
of ν, because the same viscosity appears in the wave dissipation and in the friction for the current. An
infinitely small viscosity gives the same current at the top of the boundary layer as strong viscosity, only
the thickness of the boundary layer changes, whereas there is no mean current for a zero viscosity. The
mass transport velocity reaches 2.5 times the Stokes drift we had estimated in chapter 2. The effect of
bottom friction on near-bed currents is thus considerable, which is very important for sediment transport.
The velocity profiles are shown in figure 18.1. The boundary layer introduces a significant change in the
velocity profile between the bottom and z = −h + 2.5δ, where the velocity amplitude is maximum. The
effective thickness of the boundary layer is thus close to 2.5δ. We note that the velocity profile is linear
near the bed, and its vertical derivative gives the bottom stress, fond ρw ν∂U/∂z,

∂U σ 2 a2 kδ
ρw ν = = βν E/C. (18.20)
∂z 4 sinh2 (kD)
The loss of wave momentum to the bottom means that it is not the full radiation stress that is relevant
for the change in sea level (Longuet-Higgins, 2005; Ardhuin, 2006a).
180 CHAPTER 18. BOTTOM BOUNDARY LAYER: PROCESSES AND PARAMETERIZATIONS

In the presence of partially standing waves, the situation is a little more complicated, and the stream-
ing is directed towards the surface elevation nodes, which is probably the cause of the formation of
multiple sand bars (Heathershaw, 1982).

18.3 Turbulent boundary layer

In practice the boundary layer is often turbulent, as very well observed in the laboratory by Jensen et al.
(1989). In the presence of a mean current, the wave boundary layer is thinner than the current boundary
layer so that the waves define the roughness felt by the current. As a result a change in wave height can
lead to a change in tidal currents and tidal range (Wolf and Prandle, 1999). As the effect of currents on
the boundary layer is generally weaker, we will firs assume that the mean current is zero.
Defining a Reynolds number from the free stream wave displacement amplitude aorb and velocity
amplitude uorb ,
aorb uorb
Re = (18.21)
ν
a transition to turbulence is observed for a smooth bottom for Re > 105 . For a wave period of 10 s and
monochromatic waves, this corresponds to a aorb ≃ 0.5 m. In practice the bottom roughness, even with
fin sand, is enough to make the boundary layer turbulent for much lower amplitudes (Jonsson, 1967).
The bottom roughness is defined by all the elements of topography of horizontal scales less than the
typical orbital amplitude ab,rms .
Following Prandtl, turbulence is characterized by eddies with a typical diameter κδ at a distance δ
from the bed. These eddies turn around in time that is of the order of κδ/u⋆ . The equivalent viscosity
(or eddy viscosity) Kz is defined by a linear relation between the momentum flux and the velocity shear

u2⋆ = u′ w′ = Kz ∂u/∂z. (18.22)

When the flux u2⋆ is constant, Prandtl’s mixing length l = κ(z + D) gives Kz = κ(z + D)u⋆ , leading to
a log profile. Namely the solution of ∂u/∂z = u⋆ /(κz) is
 
u⋆ z
u(z) = log . (18.23)
κ z0
The eddies mix the flow on a time scale that is the u⋆ /l. When this is is faster than the wave period
we are in the boundary layer. Thus, an order of magnitude for the boundary layer thickness is given by
p
δ = u⋆ /(κf ) ≈ Kz /f /κ. (18.24)

The friction velocity u⋆ is often related to the velocity u∞ outside of the boundary layer, with a
friction coefficient fw , this gives u2⋆ = 0.5fw u2∞ . Experiments typically give a highly variable fw , with
values under 0.3. Typically fw is a function of the ratio of the size of roughness elements ks and the
orbital displacement A, and of the Reynolds number Re. In particular fw decreases when A/ks increases,
and the wave boundary layer thickness is typically less than 10 cm.
Using eq. (18.2) and (18.3) a parameterization of the turbulent term G following Prandtl mixing
length ideas gives (u′ w′ = Kz ∂u/∂z) with Kz = κ(z + D)u⋆ . This leads to the following equation
(Kajiura, 1968; Grant and Madsen, 1979),
 
∂ ⋆ ∂uδ
z + ie
u = 0, (18.25)
∂z ⋆ ∂z ⋆

p+ p-

Figure 18.2: Schematic of the flow in the boundary layer, showing the importance of flow detachment
from the boundary that induces a drag.
Detachments can occur at all scales, from the grain size to ripples or coral elements (e.g. Monismith
et al., 2015).
18.4. BOTTOM ROUGHNESS 181

Figure 18.3: Comparison of velocity profiles in the bottom boundary layer observed by Jensen et al.
(1989) at high Reynolds numbers and the model of Wiberg (1995). This case has a wave period of
10 s, with an amplitude of velocity oscillations of 1 m s−1 . The solid line represents model results using
⋆
Kz = κu⋆ (z + D)e−z that is time-independent, and the dashed line is the model with a time-varying
Kz . Left panel: acceleration phase, right panel: deceleration. This figure is taken from Wiberg (1995).
We note that the laboratory experiment of Jensen et al. (1989) uses a U-shaped tube and thus does not
include wave propagation effects such as the bottom streaming.

with z ⋆ = ω(z + D)/(κu⋆ ). This equation has an analytical solution. The boundary conditions of uδ
going to zero when z ⋆ is large gives
" √ √ #
ker(2 z ⋆ ) + ikei(2 z ⋆ )
u= 1− p √ u+ , (18.26)
ker(2 z0⋆ ) + ikei(2 z ⋆ )

where z0⋆ is the non-dimensional roughness z0⋆ = ωz0 /(κu⋆ ). For a smooth sandy bottom with well sorted
grain sizes, Nikuradse (1933) gives z0 = D50 /30 where D50 is the median sand grain size.
The Kelvin functions ker and kei are similar to a logarithm for z ⋆ → 0, and oscillate for z ⋆ → ∞.
The general result is that the velocity amplitude is largest near the top of the boundary layer, and has
a phase that leads the free stream oscillation (figure 18.3).
Using this solution and taking the limit z ⋆ → 0, one gets the stress at the bottom and the wave
energy dissipation rate, in the form of eq. (18.33) below. But we will first have to determine the bottom
roughness.
This linear eddy viscosity model was extended to more realistic profiles, including a decrease of Kz
outside of the boundary layer and its variation with the wave phase. These modifications have a very
limited impact on the results (Trowbridge and Madsen, 1984; Jensen et al., 1989; Wiberg, 1995; Davies
and Villaret, 1999; Marin, 2004). In particular the velocity profile is well reproduced by the linear and
time-independent eddy viscosity (figure 18.3). Still, there can be an impact on the shear stress at the
bed, with consequences for the resuspension of sediments due, among other things, to the asymmetry
between the acceleration and deceleration phases of the flow. Such flows are generally better captured
by a k − ω turbulence closure (e.g. Marieu et al., 2008).

18.4 Bottom roughness

For a rocky seafloor, the bottom geometry may be complex, but at least it is fixed. In the case of sand,
silt and mixtures the shape of the bottom and the associated roughness is modified by the flow and can
also be strongly impacted by animals living in the sediment. Here we will discuss the case of a solid
bottom, and leave the possibility of mud liquefaction for later (e.g. Jaramillo et al., 2009).
182 CHAPTER 18. BOTTOM BOUNDARY LAYER: PROCESSES AND PARAMETERIZATIONS

18.4.1 Sand and ripples

For a sandy seafloor, ripples are generally formed by the oscillatory wave motion, and these ripples
determine the roughness and thus the dissipation rate of wave energy (Zhukovets, 1963; Nielsen, 1992).
Starting from a flat seafloor, we can assume that the roughness of the flow over sand grains is the same as
measured by Nikuradse (1933) in pipes in which sand was glued with lacquer. He found that z0 = 30kN
where kN is the grain diameter. This scale kN is also called the Nikuradse roughness. In practice the
sand grains may not have all the same diameter and it is customary to use the median grain size D50 .
Another issue is that the bottom is not flat and thus it is not just the grains that contribute to the
roughness (that part is called “skin friction”) but there is also a “form drag” associated to the bottom
geometry. This form drag generally varies with the amplitude of the flow characterized by the orbital
diameter which is the length of orbits of water particles just outside the wave bottom boundary layer.
When the effects of a mean current are neglected, the bottom boundary layer can have three flow
regimes, depending on the ratio of friction and buoyancy forces acting on sediment grains. This ratio is
the Shields number, and we generally consider the maximum value over a wave period defined as

fw′ u2max
ψmax = , (18.27)
(s − 1)gD

where fw′ is a skin friction factor, s is the sediment grain density normalized by the density of seawater,
e.g. s = 2.65 for quartz, D is the grain size (Shields, 1936).
When ψmax is low, the skin friction is not enough to set grains in motion and the bottom shape is
fixed, possibly due to previous motions when waves were bigger. This is also called “relict roughness”.
This does not mean that the roughness z0 is constant, because z0 is a function of both the bottom shape
and the flow. In that regime the dissipation of wave energy is weak.
In a turbulent regime, the dissipation source term can be put in the following quasi-linear form, as
proposed by Madsen et al. (1990),

Sfric (f, θ) = λ (f ) × E (f, θ) (18.28)

2
(2πf )
λ (f ) = −fe ub,rms , (18.29)
2g sinh2 (kD)

where the friction factor fe is the ratio of the average dissipation rate and the cube of the root mean
square velocity.
As the orbital velocity increases, ψmax may exceed the threshold ψc and sand grains start moving.
At that point, it only takes a few wave periods to form sand ripples. For quartz sand, the threshold ψc
ranges from 0.03 to 0.1, depending on the grain size D (e.g. Soulsby, 1997). The formation of these ripples
enhances the form drag, and thus the dissipation rate of wave energy, corresponding to larger values of
fe . Random wave experiments by Madsen et al. (1990) show that fe is maximum for ψrms ≃ 1.2ψc ,
where ψrms is estimated from the rms orbital velocity amplitude in (18.27) instead of umax . For example,
fine sand (D = 0.15 mm) in 25 m depth and a wave period Tp =12 s give this maximum dissipation for
Hs = 1.5 m. Beyond this threshold for sediment motion, any increase of the orbital velocity leads to
smoother ripples and a reduction of fe .
For very large values of ψmax , of the order of 10ψc (Li and Amos, 1999), a layer of sediment called the
‘sheet flow’ is fluidized and oscillates with the water column. At that stage, ripples are completely wiped
out. Grain collisions become an important factor in the sediment layer. In that regime, the dissipation
factor fe grows with ψ. The three regimes of relict roughness, actively formed ripples and sheet flow are
illustrated in figure 18.4.
Using dimensional analysis and numerical modeling, Andersen (1999) showed that ripples self-organize
to reach wavelengths of the order of λ = 0.63d where d is the orbital diameter of water parcels above the
boundary layer, and their slopes are of the order of 15%. This is typically observed for random waves in
depths larger than 10 m, replacing d by 21/2 drms (Traykovski et al., 1999; Ardhuin et al., 2002). In very
shallow water, it seems that ripple wavelengths are shorter and scale with the grain size (Dingler, 1974;
Wiberg and Harris, 1994).
The determination of wave energy dissipation reduces to a parameterization of fe that must take into
account ripples (Graber and Madsen, 1988; Tolman, 1994; Ardhuin et al., 2003b).
It is common to use a representative grain size for the sediment, usually the median diameter D50
for which we may estimate a critical Shields number ψc , above which grains start to move. It is also
18.4. BOTTOM ROUGHNESS 183

(a) (b) (c)

Figure 18.4: Three regimes of the boundary layer over a sandy bottom
(a) Relict roughness with possible contribution from benthic fauna. (b) ripple formation. (c) sheet flow.
The ocean wave wavelength L and the depth are not to scale L ≈ 100 m, compared to λ ≃ 1 m for the
ripple wavelength. Horizontal arrows represent the velocity profile under wave crests. Curvy arrows in
(b) represent the eddies in the wake of ripples.

common to use root mean square values for the wave forcing at the top of the bottom boundary layer,
in particular the orbital velocity ub,rms and displacement ab,rms ,

8π 2 f 2
Z
2
ub,rms = 2 E (k) dk, (18.30)
k sinh (kD)
Z
2
a2b,rms = 2 E (k) dk. (18.31)
k sinh (kD)

0.2
Tolman (1994)
e

23
Jonswap Γ =0.038 m s
Dissipation factor f

Ardhuin et al. (2003)

0.15

0.1

0.05

0 1 0 1 2
10 10 10 10
ψ ψc

Figure 18.5: Example of dissipation factors fe as a function of the r.m.s. Shields number ψrms
The solid line is the parameterization by Tolman (1994), for fine sand (D50 = 0.15 mm), and a wave
period T = 14 s. This parameterization was adjusted to SHOWEX field data by Ardhuin et al. (2003a).
The dashed line correspond to an equivalent value of fe given by the JONSWAP parameterization, which
does not take into account the varying bottom roughness.

The boundary layer model gives a skin friction factor fw′ , a Shields number ψrms = fw′ u2b,rms / [g (s − 1) D50 ],
and a total friction factor that includes skin and form drag fw , that is the ratio of the shear stress τ and
u2b,rms ,
s
z0 2 D50 ou kN
= , (18.32)
l fw′ ou fw 30κ ab,rms
κ2
fw′ or fw = h  p   p i . (18.33)
2 ker2 2 z0 /l + kei2 2 z0 /l

where z0 /l is a non-dimensional roughness length, κ is von Karman’s constant (κ = 0, 4), ker and kei are
the Kelvin functions of order 0 and 1.
184 CHAPTER 18. BOTTOM BOUNDARY LAYER: PROCESSES AND PARAMETERIZATIONS

When ψrms /ψc < 1, 2, kN is given by the relic roughness and may weakly increase with the orbital
diameter ab,rms because larger horizontal scales contribute to the roughness when the flow amplitude
increases. When ψrms /ψc > 1, 2 kN can be taken as the sum of a ripple roughness kr and a sheet flow
roughness ‘sheet flow’ ks . For example, Madsen et al. (1990) and Wilson (1989) give
 −2.5
ψrms
kr = ab,rms × 1.5 , (18.34)
ψc
−0.4
u2.8
b,rms ab,rms
ks = 0.57 1.4 2 . (18.35)
[g (s − 1)] (2π)

Another approach to bottom friction that is completely empirical, goes back to the analysis of the
JONSWAP experiment by Hasselmann et al. (1973). In his analysis K. Hasselmann had initially thought
that the bottom stress would be quadratic with a relation given by Hasselmann and Collins (1968), so
that the dissipation rate may be cubic in the wave amplitude, with a theoretical formulation given for a
Gaussian velocity distribution. The JONSWAP data contradicted this view, and K. Hasselmann ended
up fitting a dissipation rate that is proportional to the spectrum of the velocity variance at the bottom,
 2
2πf
Sfric,JONSWAP (f, θ) = −Γ E (f, θ) . (18.36)
g sinh (kD)

The only justification for this expression was that the near-bottom tidal current u b could have played a
role and would give Γ = gcb u b where cb is the drag coefficient of Hasselmann and Collins (1968). The
JONSWAP data gives a mean value Γ = 0,038 m2 s−3 , but it ranges from 0.0019 to 0.160 m2 s−3 ). It may
be surprising that this parameterization have been widely used, even in the absence of mean current. It
is probably because, by chance, it follows the general trend over a sandy bottom shown in figure 18.5.
However, it fails in many cases (figure 18.6).
As for the effect of a mean current, what is happening in the wave bottom boundary layer will be
discussed in the next section. Other wave-turbulence interaction effects are probably similar to what
is happening near the surface, with a production of turbulence kinetic energy due to the Stokes drift
stretching of the turbulence, at the rate u′ w′ ∂Us /∂z, which can be large in the presence of tidal currents.

18.4.2 Rocks
The strong dissipation over coral reefs can be attributed, to a large extent, to the very large roughness
of corals. Lowe et al. (2007) and Monismith et al. (2015) gave detailed analyses of this situation.
18.5. JOINT WAVE AND CURRENT BOUNDARY LAYERS 185

Figure 18.6: Sand ripples and wave dissipation

Top: three ripple fields observed with a sidescan sonar, each square is 30 m across (Ardhuin et al.,
2002). Bottom: comparison of modeled and measured wave heights, normalized by the measured wave
height at the location of the Chesapeake Lighthouse in 18 m depth, during swell-dominated conditions
for the SHOWEX experiment (Ardhuin et al., 2003a). The left panel shows the case in which the model
has no dissipation, and the modeled Hs can be 4 times the measured value. The center panel uses
the JONSWAP parameterization, with Γ = 0,038 m2 s−3 , and the right panels are based on Grant and
Madsen (1979) as modified by Tolman (1994) and Ardhuin et al. (2003b) to include ripple formation and
relic ripple roughness. This last parameterization is now called the SHOWEX parameterization.

18.5 Joint wave and current boundary layers

The joint interaction of waves, current and turbulence can be relatively complex and Madsen (1994) has
extensively studied this problem, showin that the roughness for random waves could be given by a single
parameter. A recent review and analysis is given by Zou (2004).
The theory distinguishes a wave boundary layer z + H < δwc , in which the wave stress τw = ρw u2⋆w
and the current stress τc = ρw u2⋆c add up to give a total stress τwc = ρw u2⋆wc , that is the net momentum
flux towards the bottom, and a thicker current-only boundary layer z + H > δwc in which the stress is
given by the current stress only τ = ρw u′ w′ = τc .
Following the analysis by Grant and Madsen (1979), the Eulerian mean horizontal momentum equa-
tion is
∂b
u
κ |u⋆c | z = u2⋆c where (z + H) > δw c, (18.37)
∂z
with the momentum mixed by eddies of sizes controled by the distance to the bottom, following Prantl.
Close to the bottom, turbulence comes from shear production by both the current and the wave-induced
velocities so that the current velocity is mixed by this combined turbulence

∂b
u
κ |u⋆c | z = u2⋆wc where (z + H) < δw c. (18.38)
∂z
These two equations can be solved separately giving
u⋆c z
u
b= ln where (z + H) > δw c, (18.39)
κ z0c

and,
u⋆c u⋆c z
u
b= ln where (z + H) < δw c. (18.40)
κ u⋆wc z1c
186 CHAPTER 18. BOTTOM BOUNDARY LAYER: PROCESSES AND PARAMETERIZATIONS

Because u
b is continuous, we get
ϵ 1−ϵ
z0c = z0wc δwc . (18.41)

where z0wc is the roughness for the wave motion, and ϵ = u⋆c /u⋆m with u⋆m the maximum combined
friction velocity, defined by
u2⋆m = u2⋆wm + u2⋆c (18.42)
in which u⋆wm = max(u⋆w ) for monochromatic waves. In the case of random waves, it is logical to use
the r.m.s. value.
This theory thus gives a wave-induced modification of the bottom roughness applicable to the current.
Mellor (2002) showed that a wave-averaged numerical model could use an enhancement of the turbulent
kinetic energy due to wave dissipation, and the resulting mean current profile is similar to the result
of a k − l turbulence closure with a modified bottom roughness. Such an approach can be extended to
add other sources of turbulence like wave breaking, which may influence bottom friction in the surf zone
(Feddersen et al., 2003) .
The equation (18.38) for z0c misses the source of momentum due to wave dissipation, which is re-
sponsible for streaming. Mathisen and Madsen (1996) have thus modified the equivalent roughness to
shift the velocity profile by −b
u(δwc ), namely,

z0a = z0c exp [κb

u(δwc )/u⋆c )] . (18.43)

b(z) = u⋆c /κ ln(z/z0 a) = u⋆c /κ ln(z/z0 c) − u

Near the bottom this gives u b(δwc ).
This adjustment of the roughness gives a mean velocity u b at the elevation −H + δwc . In practice this
may change the effective z0 from 1 to 3 cm (Mathisen and Madsen, 1996).
Chapter 19

Wave-driven nearshore flows: water

levels and currents

The storm surge is generally defined as the difference between the observed mean sea level and the
astronomical tide. To be more precise, one should specify the duration over which this average is
performed. It is customary to take a long enough average so that the crest-to-trough variations due to
wind waves (on a time scale under 30 s) are filtered out. In general, the longer infragravity with periods
20 to 300 s are also filtered. It is thus customary to write the instantaneous sea level at the coast as a
sum of the tide, surge, and run-up, with the run-up containing all the oscillatory motions shorter than
about 300 s.
However, even if the water level oscillations caused by waves are filtered out, the storm surge does
contain some direct effect of the waves: this is called the ”wave set up”. This wave set-up can be large
part of the storm surge. For example, it was estimated to be about 1 m out of the 8 m of storm surge
that flooded New Orleans during the passage of Hurricane Katrina in 2005 (Resio and Westerink, 2008).
The proportion of set-up in the storm surge increases. On the French Atlantic coast, the highest storm
surge recorded is around 4 m on the steep cliffs of Bannec Island near Molne (Ardhuin and Magne,
2010; Sheremet et al., 2014). In that case the wave set-up exceeds 80% of the storm surge. In the
context of rising global sea level exacerbated by land subsidence in many coastal cities, understanding
and managing the wave-induced extra meters of sea level will thus become more and more important.
This, together with wave-induced currents that are the main source of nearshore sediment transport, are
the main motivations for the present chapter.

19.1 Mean flow equations

We have established in chapter 9 that waves introduce and extra flux of momentum, the radiation stress
S rad (which is a 2 by 2 tensor) and a mass flux, the Stokes transport Mw (10.35) which is a horizontal
vector. We have given in chapter 10 the form of the wave-averaged momentum and mass conservation
equations. We are now going to look at the consequences of these wave-induced fluxes in the surf zone.
We recall that the mean water depth is D = ζ + h and the total mass flux caused by a mean current
and the waves is M = ρb uD + M. We also define the mean mass flux velocity to be U = M/ (ρD) =
b + Mw / (ρD).
u
Given the small spatial scales of the surf zone we can neglect the Coriolis force. Considering a
stationary solution, the momentum equations (10.33) becomes

∂ rad

∂ rad

∂ζ
Ux Mx + Sxx + Ux My + Syx = −ρgD + τx,s + τx,b (19.1)
∂x ∂y ∂x

∂ rad

∂ rad

∂ζ
Uy Mx + Sxy + Uy My + Syy = −ρgD + τy,s + τy,b , (19.2)
∂x ∂y ∂y

where (τx,s , τy,s ) and (τy,b , τy,b ) are the surface and bottom stresses respectively.

187
188 CHAPTER 19. WAVE-DRIVEN NEARSHORE FLOWS: WATER LEVELS AND CURRENTS

19.2 Mean sea level: set-down and set-up

We now take the coast to be straight and aligned in the y direction. Assuming that everything is uniform
alongshore, all derivatives with respect to y vanish. The mass conservation given by eq. (10.35) thus
becomes,
∂Mx
= 0. (19.3)
∂x
In other words Mx is a constant, and since we expect no flow through the coast this constant is zero,
Mx = 0. We note that it means that the transport due to the mean current must be opposite to the
Stokes transport. Since the Stokes drift is directed shoreward and concentrated near the surface this
explains that the mean current is directed offshore: this current is also called the ”undertow”.
Momentum conservation in the cross-shore x direction reads

∂ rad ζ
S = −ρgD∂ + τa,x + τb,x (19.4)
∂x xx ∂x
For waves with crests parallel to the shore θ = 0. If we first neglect the surface and bottom stress,
τa,x = τb,x = 0, we get a sloping sea level
  
∂ζ 1 ∂ Cg E sinh 2kD
=− 2− . (19.5)
∂x ρgD ∂x C sinh 2kD + 2kD

Even without computing the solution, we note that for waves which do not break the energy flux Cg E is
rad
constant, and given that the phase speed C becomes smaller in shallower water, the flux Sxx increases
in shallower water, which gives a negative ∂ζ/∂x, meaning that the sea level decreases towards the shore:
this is the ”set-down”. For a first approximation one could assume ζ << h and then replace D with h,
to obtain
a2 k
ζ=− (19.6)
2 sinh (2kD)
where a is the local wave amplitude. To obtain this relation you can consider that the wave frequency is
constant and take the derivative of σ 2 = gk tanh kD to obtain a relationship between ∂k/∂x and ∂D/∂x.
Besides, using Snel’s law one finds that this relation also holds for any wave direction offshore.
At the location where waves start to break, with depth hd , we can assume that waves are in shallow
water, i.e. kD << 1, giving a wave height Hb = 2a = γhd and thus a lowest elevation of the mean sea
level
γ
ζ = − Hb . (19.7)
16
Using values in Fig. 17.6 for a 2 m 12 s swell with an incidence angle of 20◦ the breaking wave height is
2.3 m and taking γ = 0.4 we get a set-down of 6 cm.
Now, when waves break the flux generally decreases and the sea level goes up: this is the ”set-up”.
In the surf zone, the wave height is controled by the local water depth, given roughly by 2a = Hb = γD
so that the energy density is
E = ρgγ 2 D2 /8. (19.8)
rad
Assuming shallow water kD << 1 gives Sxx = 1.5E and

rad 3
2
Sxx = ρgγ 2 h + ζ . (19.9)
16
Now replacing that in the momentum equation

3 2
∂

∂ζ
γ 2 h+ζ h+ζ =− h+ζ (19.10)
16 ∂x ∂x
we get the sea level slope
∂ζ ∂h
= −B (19.11)
∂x ∂x
where  −1
1
B = 1+ (19.12)
3γ 2 /8
19.2. MEAN SEA LEVEL: SET-DOWN AND SET-UP 189

model without "wave roller"

modeled

model without bottom friction

measured

Figure 19.1: Wave set up on the beach: measurements and model

Data were collected from September to November at locations 10 to 300 m from the shoreline at Duck,
North Carolina. The data was sliced into 8.5 minute-long records, and the figure shows the observed
mean sea level compared to the prediction using a model that integrates gradients of radiation stresses,
with radiation stresses estimated from local wave measurements. One flavor of the model includes a
parameterization for a ”wave roller” which is represents the wave roller as an intermediate buffer of
momentum between the waves and the mean flow. Bottom friction is parameterized here with an eddy
viscosity that is a function of he wave height, giving bottom drag coefficient valies close to those of eq.
(19.21) avec un coefficient 0.018 < Cf < 0.028 that is typically higher than values used for the longshore
current, but which could be justified by the orientation of surfzone ”megaripples” (Gallagher et al., 1998).
Picture from Apotsos et al. (2007), copyright American Geophysical Union.

which integrates to
ζ = −Bh + A0 . (19.13)
where A0 is given by matching the set-down value at the depth hd ,

ζ = B (hd − h) + ζ d . (19.14)

The depth difference (hd − h) is positive in the surf zone and the wave set-up increases towards the shore.
Using the same numerical values, (swell of offshore wave height 2 m), at the coast where h = 0, we get
ζ = 26 cm.

19.2.1 Wave set-up in practice

Measurements of mean sea level in the surf zone haev generaly confirmed the cross-shore momentum
balance between the convergence of the radiation stresses and the slope in the mean sea level. However,
in reality, bottom friction can be an important term for shallow water, in particular D < 1 m in the
examples shown in Fig. 19.1.
Because the set-up (mean level) and run-up (which includes wave-by-wave oscillations) are a function
of the details of the beach profile, which is often not known well enough, many alternative empirical
methods have been proposed using a simplification of the beach shape and offshore wave parameters.
The values of set-up and run-up are particularly important for understanding how far the sea can reach
inland, and designing shoreline protection structures. Hunt (1959) considered a height parameter that
190 CHAPTER 19. WAVE-DRIVEN NEARSHORE FLOWS: WATER LEVELS AND CURRENTS

Figure 19.2: Left: map of Bannec island, Center: topography profile along the A-A’ transect, Right:
Maximum surface elevation for every 10-minute record from October 2008 to March 2009, as a function
of the offshore wave height and period estimated from the wave model of Ardhuin and Magne (2010).
Blue squares correspond to an astronomical tide below the mean sea level, and red squares are for tide
levels above 2.5 m.

combines the offshore wave period and offshore wave height,

p
HH = Tm0,−1 gHs , (19.15)

Many observations show that extreme water levels generally scale with this height HH . For a given
depth profile, Fig. 19.2 shows how HH gives a useful scale of extreme water levels measured at Bannec
island (to be precise, the measurement is not a water level but a pressure converted to height above the
sensor assuming hydrostatic equilibrium). This height parameter is also related to the Irribaren number
introduced in chapter 17, indeed the set-up and run-up are also a function of the foreshore slope βf (the
slope of the part of the beach that goes from dry to wet with the passage of the waves), so that the
run-up can be written as proportional to the offshore wave height times the Irribaren number ξ0 .
For example, Stockdon et al. (2006) found an empirical relation of the form

ζ ≃ 0.14βf HH (19.16)

This formula is semi-quantitative: it only explains half of the variance in the data collected by these
authors, and up to 70% of the variance for the most dissipative cases. In fact, for ξ0 < 0.3, the slope
appears secondary and this expression fits these cases better,

ζ ≃ 0.0064HH . (19.17)

Similar expression have been derived for the run-up (Holman, 1986), and recent analysis of pebble
beaches (Poate et al., 2016) and cliffs (Dodet et al., 2018) show that they probably need to be adjusted
for these environments. Further studies have also investigated how much water goes over a dune or
coastal structure, which is really the quantity of interest for flooding where the coast is not erodible
(Pullen et al., 2007).

19.3 The longshore current

Now looking that the momentum balance in the along-shore y direction, we get
rad
∂Sxy
= τy,s + τy,b (19.18)
∂x
and for monochromatic waves with incidence angle θ the radiation stress is

rad Cg
Sxy =E sin θ cos θ. (19.19)
C
19.3. THE LONGSHORE CURRENT 191

Since sin θ/C is a constant according to Snel’s law, the force exerted by waves along the y axis is zero
where ECg / cos θ is constant, which is outside the surf zone. However, inside the surf zone, the energy
flux ECg / cos θ is reduced in the case sin θ > 0 or augmented if sin θ < 0. As a result, the divergence
rad
of Sxy gives a force that pushes the water along the shore. If we neglect the wind stress, the bottom
stress τy,b is the only force that can balance the wave-induced force. Since the bottom stress is in the
opposite direction to the mean current, then there should be a current V flowing towards y > 0 in the
case sin θ > 0.
For a quantitative estimate, we may assume an instantatenous quadratic drag law,

T (t) = −Cf |(U + u)| (U + u) (19.20)

Longuet-Higgins (1970) showed that for |V | << |u| and small values of θ, the bottom stress becomes

τy,b = −ρCf |u| (V + v) (19.21)

where V is the mean current in the y direction and (u, v) is the orbital velocity at the sea floor. Using
linear wave theory that gives
gD
u = cos (kx − ωt) (19.22)
2C
gD
|u| = (19.23)
πC
gD
τy,f = −ρCf V (19.24)
πC
and the longshore current velocity is given by (Thornton and Guza, 1986),
πC ∂ (ECg cos θ) sin θ0
V =− . (19.25)
ρCf gD ∂x C0
One example of estimated surf zone current for a beach profile that includes a bar is shown in Fig.
19.3. The large magnitude of the current, of the order of 1 m/s is realistic. This current has also been
called the ”river of sand” as it is responsible for transporting sand along beaches in the direction of
the dominant waves. Sandy beaches are only in equilibrium when they are facing the waves, which
0
Depth (m)

10

20
0 100 200 300 400 500
3
Hs (m)

0
0 100 200 300 400 500
Direction (deg)

20

10

0
0 100 200 300 400 500
2
V (m/s)

0
0 100 200 300 400 500

Figure 19.3: Wave breaking and longshore current

Results of the model by Thornton and Guza (1986) for the same case shown in Fig. 17.6, namely a wave
period T = 12 s and 20◦ incidence angle in 20 m depth. Waves propagate over a schematic beach with
a mean slope β = 0.04 and a bar at x = 300 m.

qualitatively explains the usual concave shape of beaches (when seen from above) that naturally tends to
realign itself to face the waves: this also explains the formation of tombolos behind detached breakwaters,
and many other features.
192 CHAPTER 19. WAVE-DRIVEN NEARSHORE FLOWS: WATER LEVELS AND CURRENTS

In practice, however, the spatial variation of the current on barred beaches does not generally have the
two maxima in the longshore current (for example at 300 m and 430 m in Fig. 19.3) but rather one single
broad maximum (Reniers and Battjes, 1997). Many reasons have been used to explain this: horizontal
mixing by horizontal vortices (Church and Thornton, 1993; Brocchini et al., 2004), the ”buffering effect”
of the ”wave roller” (Lippmann et al., 1996), or the mixing effect of vertical shear (Putrevu and Svendsen,
1999). Also, the current is not a steady flow but instead is highly variable, due to the variability of the
forcing itself and also related to instabilities of the current (Oltman-Shay et al., 1989).

19.4 Cross-shore flows

Besides the along-shore transport of material, and in particular sediments, surf zones are also active
regions of cross-shore exchanges. Part of this exchange is associated to the vertical profile of the cross-
shore current, with an onshore flow in the wave bottom boundary layer, associated to streaming (Longuet-
Higgins, 1953), an offshore flow above the boundary layer and onshore transport related to the Stokes
drift near the surface.
The other part of the cross-shore flow is related to non-stationarity and non-uniformities of the surf
zone currents. Figure 19.4 shows two examples of the transport of water marked by the fluorescent tracer
rhodamine which appears in pink. The tracer was released at a single point in the surf zone or near the
mouth of a river during the ebb flow. The tracer is clearly transported along the coast by the longshore
current and also mixed across the surf zone. This has clear implications for the sediment transport and
the transport of contaminant around beaches which is important for water quality and human health
(Delpey et al., 2014).
Entrainment of material further offshore is of particular interest for fisheries (larvae recruitment),
the global carbon cycle, or plastic contamination of the ocean. Waves are again important right at the
surface or in the shallower parts of the continental shelf (Lentz et al., 2008; Onink et al., 2019).

Figure 19.4: Longshore and cross-shore transport in the surf zone. (A) shows a release in the surfzone
at Imperial Beach, south of San Diego, California, and (B) shows a release within an ebb tidal flow at
New River Inlet, North Carolina. Figure from Clark et al. (2014), copyright American Meteorological
Society.
Chapter 20

Numerical wave modelling at

regional to beach scales

Many different approaches have been used and refined for coastal waves. Looking at models that solve
for a single-valued field ζ(x, y, t) (i.e. the waves are not allowed to overturn) they can be phase resolving
(with the individual waves resolved), spectral with random phases (like the models described in Chapter
12), coupled spectral and bispectral (to keep some nonlinear properties specific of shallow water). There
are also many fluid dynamics models for multi-phase flow (air, water and sediment) that have been
applied to wave breaking and sediment transport.
The choice of model is often dictated by the complexity that can be afforded and the space and time
scales that must be resolved: a few minutes of a 100 by 100 m region can be studied with much more
detail than century-scale shoreline evolution under climate change scenarios, even if only for a small bay.
One very important aspect to keep in mind is that the shore is often a region of intense breaking and
high fluid accelerations and velocities: in this context the waves, currents and even bathymetry evolution
are tightly coupled. For this reason, many efforts have been made to develop and validate coupled wave-
flow-sediment models (e.g. Reniers et al., 2004). The first questions to ask are : What space and time
scales do I need to resolve? Do I need an explicit resolution of these scales or some parameterization?
What are the associated processes that need to be well represented? When in doubt, it is tempting to
go for the ”fullest” possible model. However, some very simple parameterizations (e.g. one-line models
of shoreline change) can often capture most of the dynamics we are interested in. In short, all models
are good, but some are better than others for one’s specific needs.
This chapter will thus not cover all possible modeling strategies but explore the limits and usefulness
of the same spectral approach that is used for marine weather forecasting offshore, and illustrate a few
other possibilities. The quality of coastal wave models solution depends on many factors, and is more
complex than open ocean situations. This is reviewed by Roland and Ardhuin (2014). In general, the
important aspects are,

ˆ the forcing: waves are generated by the wind, and are strongly influenced by currents, varying
water depth, and any obstacles (small islands, sea ice, icebergs ...). All these forcing fields should
thus be defined as accurately as possible, starting with the winds, but not forgetting currents that
are generally stronger in coastal regions.

ˆ parameterizations of physical processes: many errors, in particular in coastal areas are due to
deficiencies in source term parameterizations. In particular, waves in coastal regions can be often
fetch-limited, and the growth of waves when wind is oblique relative to the shoreline can reveal
errors in the magnitude of the wind generation and whitecapping source terms (Ardhuin et al.,
2007). Also, bottom friction is a dominant factor when the water depth is less than half the
dominant wavelength, and bottom properties can have a very large impact the dissipation rate of
wave energy (Roland and Ardhuin, 2014; Monismith, 2007).

ˆ Numerics: Although some numerical methods are clearly worse than others, there is no perfect
method that would be fast and accurate enough to deal with the very localized depth-induced
breaking. For resolutions coarser than 100 m, explicit triangle-based grids using residual distribu-
tion (RD) schemes (Csı́k et al., 2002; Ricchiuto et al., 2005) may well be the most efficient (Roland

193
194 CHAPTER 20. NUMERICAL WAVE MODELLING AT REGIONAL TO BEACH SCALES

and Ardhuin, 2014). At higher resolution, the CFL constraint of explicit schemes makes them
prohibitively expensive, and one may have to live with low order schemes and their diffusion.

20.1 Forcing fields

20.1.1 Winds, currents and water depth
Since it takes a long fetch to develop waves, the primary concern is generally on the offshore winds, as
discussed in chapter and coastal winds only have a moderate impact on the sea state. These issues are
discussed in chapter 12. Recent operational weather forecasts are now available at resolutions around
1 km and can be compared to SAR-derived winds at resolutions around 500 m, for example from Sentinel-
1, or meteorological Doppler radar data (e.g. Ahsbahs et al., 2020).
Getting good currents can be much more problematic, except possibly for tidal current. Still, one
should be careful that tidal models that are very good for sea level, thanks to satellite data assimilation,
may not be that good for currents. Some regions have coastal High-Frequency radars that can provide a
good validation dataset for surface currents, but the spatial resolution can be marginal for resolving some
sharp current gradients (Ardhuin et al., 2009b). With small-scale ( < 50 km) details of currents important
for the waves propagating to the coast (Marechal and Ardhuin, 2021), some regions are particularly prone
to large model errors given the lack of detailed measurements of surface currents.
Finally, water depth is the sum of the varying water level and of the (also varying but usually more
slowly) bathymetry. Recent efforts to obtain bathymetry from remote sensing using wave dispersion or
changes in wet-dry boundary associated with tides will certainly lead to a better knowledge in some of
the poorly charted regions of the world. One possible exception are the last few kilometers to the coast,
in particular for exposed sandy shorelines where wave-driven sand transport leads to large changes in
water depths on the scale of storms and seasons.

20.2 Numerics
Restricting ourselves to spectral methods, one of the first developed methods is the use ray-tracing to
transform global-scale forecasts (or offshore bouy measurements) into coastal wave predictions. Whereas
the early versions used forward ray tracing and considered a single or a few offshore wave periods and
directions, later implementations used eq. (16.6) with backward ray-tracing to transform the full wave
spectrum. This is still used in some regions where local wind generation can be neglected and where
time-varying currents do not require a frequency recomputation of wave rays (O’Reilly and Guza, 1993,
1998; Crosby et al., 2017). However, this method is not generally applicable due to wind generation and
dissipation source terms.
A more general approach is thus to apply the wave action equation in coastal areas, including shal-
low water effects such as bottom friction, and computing the evolution of the wave spectrum across a
discretized model grid. A straightforward implementation of the global model in a coastal setting leads
to a generally high cost due to the fine resolution required to resolve sharp gradients near the coast.
This cost is particularly high if the model uses an explicit numerical scheme, with a time step that is not
dictated by the true time scale of evolution of the wave field (driven by the forcing, i.e. one hour or so
when taking into account tidal currents), but by the stability of the numerical scheme: explicit schemes
requires time steps of a few minutes (dt around 50 s) at most because the wave energy cannot jump over
more than one grid cell (imagine dx=1000 m) in a time step and the speed for the longest wave periods
is still around 20 m/s. A possible solution is to use implicit schemes with issues related to non-linearity
or non-monotonicity or numerical diffusion, as discussed in chapter 12.
Specific issues in coastal areas are the large range of time scales of different processes: wave breaking
in the surf zone is a very ”stiff” source term. Also, in order to optimize the model cost, grids that allow
higher resolution where gradients are large (generally near the coast) can save a lot on the number of grid
points, and for this reason many coastal models use grids with variable resolution, such as triangle-based
meshes. Further details are discussed in Roland and Ardhuin (2014).
Numerical effects are most visible on the directional spread because numerical diffusion directly
increases the spread, as shown in Fig. 20.1. We note that a finer spatial resolution does not necessarily
reduce the error, as noted by Ardhuin and Herbers (2005): to converge to the true solution one needs
finer resolution in both space and directions.
20.2. NUMERICS 195

Analytical results using Snell's law SWAN Higher Order Scheme, DX =1000m
SWAN BSBT Scheme, DX =1000m
25 Depth[m] Ardhuin & Herbers, 4-step, DX =2000m
SWAN Higher Order Scheme, DX =2000m
SWAN BSBT Scheme, DX =2000m
[°]

Depth (m)
20 0
Directional spread

15 -50

10 -100

5 -150

0 10000 20000 30000 40000 50000 60000 70000 80000 90000 100000
Length [m]
25 WWM CRD-N1 scheme, DT =150, DX =1000m WWM CRD-FCT scheme, DT =150, DX =1000m
WWM CRD-N3 scheme, DT =150, DX =1000m Ardhuin & Herbers, 4-step, dx =2000m
WWM CRD-N scheme, DT =150, DX =1000m
[°]

Depth (m)
20 0
Directional spread

15 -50

10 -100

5 -150

Figure 20.1: Test case with a wave spectrum propagating obliquely over a broad and shallow continental
shelf, with a mean direction starting at 60 degrees from shore normal: the directional spread should de-
crease as the water depth decreases and all wave components are refracted to the shore-normal direction.
In numerical models, this refraction is not exact. Here the test involves three different models: one uses
ray-tracing (Ardhuin and Herbers, 2005) and the others are SWAN with 2 different advection schemes
and two different spatial resolutions in the top panel, and Contour Residual Distribution schemes (Csı́k
et al., 2002) of different orders in the Wind Wave Model of Roland (2008), using triangle-based grids,
in the bottom panel. Schemes N1 and N2 are implicit while N and FCT are explicit. Reproduced from
Roland (2008).
196 CHAPTER 20. NUMERICAL WAVE MODELLING AT REGIONAL TO BEACH SCALES

Although directional spread is not the first variable that people care about, it can have a significant
impact on radiation stresses and infragravity wave amplitudes, and thus surf zone dynamics.
At the same time, the numerical diffusion of some numerical schemes may compensate for the lack
of diffraction and correct part of the Garden Sprinkler Effect that was discussed in chapter 12 and that
is also present in coastal models. The trade-off between these issues and the cost of running the model
(CPU time) have led some of us to prefer the use of the ”N” scheme (that is equivalent to a first order
scheme for triangle-based grids) instead of the higher order ”Flux Corrected Transport” (FCT) scheme
(Cancouët, 2008; Ardhuin et al., 2009b).

20.3 Parameterizations of physical processes

20.3.1 Bottom friction
Many models still use an empirical ”JONSWAP” parameterization, presented in chapter 18, that has very
little connection with what happens in the bottom boundary layer, beyond the obvious transformation
of the surface elevation spectrum into a bottom velocity spectrum. Indeed, bottom friction is only
significant where bottom velocities are large enough. The reason for this default empirical choice is
that other choices require the use of some information about bottom sediments. In practice we found
that using a ripple parameterization and assuming a grain size of 0.2 mm (coarse sand) everywhere was
generally acceptable. Obviously, if information on the bottom is available, it should be used: friction
over rock or mud can indeed be many times stronger than over sand (Elgar and Raubenheimer, 2008;
Monismith, 2007).

20.3.2 Triad wave-wave interactions: sum interactions

Spectrum evolution around the surf zone is severely affected by the near-resonant triad interactions
which can produce a net transfer of energy from the peak of the spectrum to harmonics, but also a flow
of energy from the harmonics back to the peak. An accurate description of this process requires some
knowlege of the bispectrum that is not available in spectral phase-averaged models. Even if results are
not perfect, different parameterizations have been developed to mimic some of the energy transfers in
phase-averaged models (Eldeberky and Battjes, 1995; Becq-Girard et al., 1999).
Solving this issue is one of the main reasons why specific models such as the bi-spectral model of
(Herbers and Burton, 1997) or the time domain model SWASH (Zijlema et al., 2011) have been developed
specifically to model waves in surf zones. The other important benefit of time-domain models is that
they can also solve for the wave-driven currents and sediment transport.

20.3.3 Triad wave-wave interactions: difference interactions and infragravity

wave generation
Since infragravity (IG) motions can be very energetic in the surf zone, leading to changes in water level
of 1 m or more over a time scale around 100 s, it is impossible to ignore IG waves near the shoreline:
indeed the primary limitation of wave heigth due to the depth-induced breaking introduce a strong
coupling between the IG water levels and the wave heights. IG waves can also transport sediment in the
cross-shore. Resolving IG-waves while not resolving the shorter waves is the strategy that is used in the
X-Beach model (Reniers et al., 2004; Roelvink et al., 2009).

20.4 Examples of wave model results and validation

Our first example is a regional model that was designed for coastal applications and has been used,
among other things for marine energy studies (Boudière et al., 2013). The model mesh is shown in Fig.
20.2. Triangles near the coast have a resoltion around 300 m, rapidly increasing to a few kilometers when
moving offshore. The 300 m at the shoreline are not enough to resolve cross-shore gradients in the surf
zone, but is sufficient to represent many coastal effects including refraction caused by depth and tidal
currents and sheltering by headlands and islands that create ”wave shadows”.
For example Fig. 20.3 shows wave heights simulated for the severe storm ”Johanna” that hit France
in March 2008. The offshore wave orientation relative to the shoreline explains the lower values of wave
heights in coastal bays, or behind islands. One can also see the clear effect of the paleo-river in the
20.4. EXAMPLES OF WAVE MODEL RESULTS AND VALIDATION 197

middle of the Channel which causes a sharp refraction and local minimum in the wave height, because
these very large storm waves also have long periods that are refracted at depths larger than 100 m.
Such a regional model can use satellite altimetry for validation. Fig. 20.4 shows a statistical validation
over a full year. Recent altimeter retracking and denoising techniques (Passaro et al., 2014; Quilfen et al.,
2018; Dodet et al., 2020) allow a more detailed validation at higher resolution.
Thanks to their continuous time series it is often easier to use in situ measurements, such as performed
by moored buoys to analyze particular events. Fig. 20.6 highlights the different results obtained with
different parameterizations of bottom friction. However, the only location where models are significantly
different is buoy 62067 located inshore of Ile d’Yeu, across a shallow rock oucrop which is represented
by a high roughness when using the ”SHOWEX” parameterization. We also note that both models are
not very good at the Cherbourg buoy (62059) for these February 2011 swells. All three selected location
are strongly influenced by tidal currents. Unlike altimeter data, buoys say nothing about the spatial
variability of the wave field, and it is often difficult to know how far a particular buoy is representative
of a larger area.

62288

62072
62288
62072

62059

Ratelets
62061
62069 62069
62074 Four 62064
62067

62067
62064

62066

Figure 20.2: Map showing the North Sea - Channel - Biscay mesh used for our hindcasts and forecasts.
Magenta and green circles show location of permanent and temporary buoys used for calibration or
validation in addition to satellite altimeter data. Inset are zooms of four grid areas, showing typical
alongshore resolutions, with color bars displaying the elevation relative to mean sea level, in meters. The
full mesh contains 110,000 wet nodes. Reproduced from Roland and Ardhuin (2014).

We close this chapter with one example of modeled and observed time series at different locations
across a surf zone, here in Duck, North Carolina, with data collected during the DUCK’94 experiment
(Elgar et al., 1997; Gallagher et al., 1998). In this case the model used is also WAVEWATCH III with
a spatial resolution of 5 m, and a ridiculously small time step of 0.2 s because here we use an explicit
numerical scheme in a regular spatial grid. The two model runs have different breaking parameterizations:
the ”Model 1” corresponds to Filipot and Ardhuin (2012) and ”Model 2” is Battjes and Janssen (1978).
 198 CHAPTER 20. NUMERICAL WAVE MODELLING AT REGIONAL TO BEACH SCALES

Figure 20.3: Wave heights simulated around Brittany and in the Channel the March 10, 2008 at 15h00
UTC.

53 40 30

52
30
51 25
20
50
10
49 20
Latitude

normalized RMS
48 bias against 0 error against
altimeters (cm) altimeters (%)
47 15
−10
full year 2011 full year 2011
46
( n > 3) ( n > 5)
−20
45 10

44 (a) −30
(b)
43 −40 5
−8 −6 −4 −2 0 2 4
Longitude

Figure 20.4: Validation of the bay of Biscay model grid for year 2011 using all available altimeter data.
(a) bias in centimeters, and (b) normalized root mean square error (NRMSE). The satellite data was
taken from the calibrated Ifremer database (Queffeulou and Croizé-Fillon, 2010). The along-track time
series at 1 Hz sampling was averaged over 0.5 degree along the track. These ’super-observations’ (SO)
were then binned with latitude and longitude, with an average number of 35 SOs for one year in each
0.5 by 0.5 degree bin. Results are only shown for bins with at least 4 SOs for the bias and 6 SOs for the
NRMSE. Reproduced from Roland and Ardhuin (2014).
 20.4. EXAMPLES OF WAVE MODEL RESULTS AND VALIDATION 199

2.0 6 3.5
(a) 62059 (b) 62069 (c) 62067
5 3.0
1.5
2.5
Hs (m)

4
1.0 2.0
3
1.5
0.5 bottom friction
2
1.0 with variable roughness

empirical JONSWAP bottom friction

0.0 1 0.5
1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9
Days (February 2011) Days (February 2011) Days (February 2011)

Figure 20.5: Time series of observed and modelled significant wave height at several buoys using the
SHOWEX (blue diamonds) or JONSWAP (red triangles) parameterizations for bottom friction, compared
to hourly buoy measurements (solid line). Reproduced from Roland and Ardhuin (2014).

rms wave height (m)

data 2
Model 1 1
D1
Depth (m) Model 2
0
0 2
D2
1
−1 0
2
−2 D5 D3
D4 1
D3
−3 D2 0
2
D4
1
−4 D
1 0
2
−5 D5
1
400 450 500 550 600 650 700 750
Distance (m) 0
0 50 100 150 200 250
Time (hours)

Figure 20.6: Left: depth profile and location D1√to D5 where waves were measured. Right: time series
of the rms wave height (defined as Hrms = Hs / 2) at the 5 measurement locations. Reproduced from
Filipot and Ardhuin (2012).
200 CHAPTER 20. NUMERICAL WAVE MODELLING AT REGIONAL TO BEACH SCALES
Bibliography

Abdalla, S. and L. Cavaleri, 2002: Effect of wind variability and variable air density on wave modelling.
J. Geophys. Res., 107, 17.
Adler, R., 1981: The Geometry of Random Fields. John Wiley, New York.
Agnon, Y., A. V. Babanin, I. R. Young, and D. Chalikov, 2005: Fine scale inhomogeneity of wind-wave
energy input, skewness and asymetry. Geophys. Res. Lett., 32, L12603.
Agrawal, Y. C., E. A. Terray, M. A. Donelan, P. A. Hwang, A. J. Williams, W. Drennan, K. Kahma,
and S. Kitaigorodskii, 1992: Enhanced dissipation of kinetic energy beneath breaking waves. Nature,
359, 219–220.
Ahsbahs, T., N. G. Nygaard, A. Newcombe, and M. Badger, 2020: Wind farm wakes from SAR and
doppler radar. Remote Sensing, 12, 462. doi:10.3390/rs12030462.
Airy, G. B., 1841: Tides and waves. Encyclopedia metropolitana (1817–1845), H. J. R. et al., ed., London.
Alcock, R. K. and D. G. Morgan, 1978: Investigations of wind and sea state with respect to the beaufort
scale. Weather , 33, 271–277. doi:10.1002/j.1477-8696.1978.tb04683.x.
Alday, M., M. Accensi, F. Ardhuin, and G. Dodet, 2021: A global wave parameter database for geo-
physical applications. part 3: Improved forcing and spectral resolution. Ocean Modelling, 166, 101848.
doi:10.1016/j.ocemod.2021.101848.
Alpers, W. R. and C. L. Rufenach, 1979: The effect of orbital motions on synthetic aperture radar
imagery of ocean waves. IEEE Trans. Antennas Propagat., 27, 685–690.
Alves, J. H. G. M. and M. L. Banner, 2003: Performance of a saturation-based dissipation-rate source
term in modeling the fetch-limited evolution of wind waves. J. Phys. Oceanogr., 33, 1274–1298.URL
link.
Andersen, K. H., 1999: The dynamics of ripples beneath surface waves and topics in shell models of
turbulence. Ph.D. thesis, Det Naturvidenskabelige Fakultet Københavns Universitet.URL link.
Andreas, E. L., 2004: Spray stress revisited. J. Phys. Oceanogr., 34, 1429–1439.
Andréfouët, S., F. Ardhuin, P. Queffeulou, and R. L. Gendre, 2012: Island shadow effects and the wave
climate of the Western Tuamotu archipelago (french polynesia) inferred from altimetry and numerical
model data. Marine Pollution Bulletin, 66, 415–424.
Andrews, D. G. and M. E. McIntyre, 1978: On wave action and its relatives. J. Fluid Mech., 89, 647–664,
corrigendum: vol. 95, p. 796.
Annenkov, S. Y. and V. I. Shrira, 2006: A note on Hasselmann’s energy-transfer model. J. Fluid Mech.,
561, 181–207. doi:10.1017/S0022112006000632.
Apotsos, A., B. Raubenheimer, S. Elgar, R. T. Guza, and J. A. Smith, 2007: Effects of wave rollers and
bottom stress on wave setup. J. Geophys. Res., 112, C02003. doi:10.1029/2006JC003549.
Ardhuin, F., 2006a: On the momentum balance in shoaling gravity waves: a commentary of -shoaling
surface gravity waves cause a force and a torque on the bottom- by K. E. Kenyon. Journal of Oceanog-
raphy, 62, 917–922.URL link.
— 2006b: Quelles mesures pour la prévision des états de mer en zone côtière? Communications de
l’Atelier Experimentation et Instrumentation.URL link.
Ardhuin, F., A. Balanche, E. Stutzmann, and M. Obrebski, 2012a: From seismic noise to ocean wave pa-
rameters: general methods and validation. J. Geophys. Res., 117, C05002. doi:10.1029/2011JC007449.
Ardhuin, F., G. Boutin, J. Stopa, F. Girard-Ardhuin, C. Melsheimer, J. Thomson, A. Kohout, M. Doble,
and P. Wadhams, 2018: Wave attenuation through an Arctic marginal ice zone on 12 october, 2015:
2: numerical modeling of waves and associated ice break-up. J. Geophys. Res., 123, 5652–5668.
doi:10.1002/2018JC013784.
Ardhuin, F., P. Brandt, L. Gaultier, C. Donlon, A. Battaglia, F. Boy, T. Casal, B. Chapron, F. Collard,
S. Cravatte, J.-M. Delouis, E. D. Witte, G. Dibarboure, G. Engen, H. Johnsen, C. Lique, P. Lopez-
Dekker, C. Maes, A. Martin, L. Marié, D. Menemenlis, F. Nouguier, C. Peureux, P. Rampal, G. Ressler,

201
202 BIBLIOGRAPHY

M.-H. Rio, B. Rommen, J. D. Shutler, M. Suess, M. Tsamados, C. Ubelmann, E. van Sebille, M. van den
Oever, and D. Stammer, 2019: Skim, a candidate satellite mission exploring global ocean currents and
waves. Frontiers in Marine Sci., 6, 124. doi:10.3389/fmars.2019.00209.
Ardhuin, F., B. Chapron, and F. Collard, 2009a: Observation of swell dissipation across oceans. Geophys.
Res. Lett., 36, L06607. doi:10.1029/2008GL037030.
Ardhuin, F., B. Chapron, F. Collard, M. Smith, J. Stopa, J. Thomson, M. Doble, P. Wadhams,
B. Blomquist, O. Persson, and C. O. Collins, III, 2017a: Measuring ocean waves in sea ice using
SAR imagery: A quasi-deterministic approach evaluated with Sentinel-1 and in situ data. Remote
sensing of Environment, 189, 211–222. doi:10.1016/j.rse.2016.11.024.
Ardhuin, F., F. Collard, B. Chapron, F. Girard-Ardhuin, G. Guitton, A. Mouche, and J. Stopa, 2015:
Estimates of ocean wave heights and attenuation in sea ice using the sar wave mode on Sentinel-1A.
Geophys. Res. Lett., 42, 2317–2325. doi:10.1002/2014GL062940.
Ardhuin, F., F. Collard, B. Chapron, P. Queffeulou, J.-F. Filipot, and M. Hamon, 2008a: Spectral
wave dissipation based on observations: a global validation. Proceedings of Chinese-German Joint
Symposium on Hydraulics and Ocean Engineering, Darmstadt, Germany, 391–400, iSBN: 3-936146-
23-3.URL link.
Ardhuin, F., T. G. Drake, and T. H. C. Herbers, 2002: Observations of wave-generated vortex ripples
on the North Carolina continental shelf. J. Geophys. Res., 107. doi:10.1029/2001JC000986.
Ardhuin, F., F. Dumas, A.-C. Bennis, A. Roland, A. Sentchev, P. Forget, J. Wolf, F. Girard, P. Osuna,
and M. Benoit, 2012b: Numerical wave modeling in conditions with strong currents: dissipation,
refraction and relative wind. J. Phys. Oceanogr., 42, 2101–2120.
Ardhuin, F. and T. H. C. Herbers, 2002: Bragg scattering of random surface gravity waves by irregular
sea bed topography. J. Fluid Mech., 451, 1–33.
— 2005: Numerical and physical diffusion: Can wave prediction models resolve directional spread? J.
Atmos. Ocean Technol., 22, 886–895.URL link.
Ardhuin, F., T. H. C. Herbers, P. F. Jessen, and W. C. O’Reilly, 2003a: Swell transformation across
the continental shelf. part II: validation of a spectral energy balance equation. J. Phys. Oceanogr., 33,
1940–1953.URL link.
Ardhuin, F., T. H. C. Herbers, K. P. Watts, G. P. van Vledder, R. Jensen, and H. Graber, 2007: Swell and
slanting fetch effects on wind wave growth. J. Phys. Oceanogr., 37, 908–931. doi:10.1175/JPO3039.1.
Ardhuin, F. and A. D. Jenkins, 2006: On the interaction of surface waves and upper ocean turbulence.
J. Phys. Oceanogr., 36, 551–557. doi:10.1175/JPO2862.1.
Ardhuin, F., A. D. Jenkins, and K. Belibassakis, 2008b: Comments on ‘the three-dimensional
current and surface wave equations’ by George Mellor. J. Phys. Oceanogr., 38, 1340–1349.
doi:10.1175/2007JPO3670.1.
Ardhuin, F., T. Lavanant, M. Obrebski, L. Marié, J.-Y. Royer, J.-F. d’Eu, B. M. Howe, R. Lukas, and
J. Aucan, 2013: A numerical model for ocean ultra low frequency noise: wave-generated acoustic-
gravity and Rayleigh modes. J. Acoust. Soc. Amer., 134, 3242–3259.
Ardhuin, F. and R. Magne, 2007: Scattering of surface gravity waves by bottom topography with a
current. J. Fluid Mech., 576, 235–264.
— 2010: Modélisation des états de mer du globe à la plage : validation de nouveaux paramètres produits
par prévimer. Actes des Xèmes journées Génie côtier-Génie civil, Les Sables d’Olonne, Centre Français
du Littoral, 87–94.URL link. doi:10.5150/jngcgc.2010.002-A.
Ardhuin, F., L. Marié, N. Rascle, P. Forget, and A. Roland, 2009b: Observation and estimation of
Lagrangian, Stokes and Eulerian currents induced by wind and waves at the sea surface. J. Phys.
Oceanogr., 39, 2820–2838.URL link. doi:10.1175/2009JPO4169.1.
Ardhuin, F., F.-R. Martin-Lauzer, B. Chapron, P. Craneguy, F. Girard-Ardhuin, and T. Elfouhaily, 2004:
Dérive à la surface de l’océan sous l’effet des vagues. Comptes Rendus Géosciences, 336, 1121–1130.
doi:10.1016/j.crte.2004.04.007.
Ardhuin, F., B. Molero, A. Bohé, F. Nouguier, F. Collard, I. Houghton, A. Hay, and B. Legresy, 2024:
Phase-resolved swells across ocean basins in swot altimetry data: revealing centimeter-scale wave
heights including coastal reflection. Geophys. Res. Lett., in press. doi:10.1029/2024GL109658.
Ardhuin, F., W. C. O’Reilly, T. H. C. Herbers, and P. F. Jessen, 2003b: Swell transformation across
the continental shelf. part I: Attenuation and directional broadening. J. Phys. Oceanogr., 33, 1921–
1939.URL link.
Ardhuin, F., M. Otero, S. Merrifield, A. Grouazel, and E. Terrill, 2020: Ice breakup controls dis-
sipation of wind waves across southern ocean sea ice. Geophys. Res. Lett., 47, e2020GL087699.
doi:10.1029/2020GL087699.
BIBLIOGRAPHY 203

Ardhuin, F., N. Rascle, and K. A. Belibassakis, 2008c: Explicit wave-averaged primitive equations using
a generalized Lagrangian mean. Ocean Modelling, 20, 35–60. doi:10.1016/j.ocemod.2007.07.001.
Ardhuin, F., N. Rascle, B. Chapron, J. Gula, J. Molemaker, S. T. Gille, D. Menemenlis, and C. Rocha,
2017b: Small scale currents have large effects on wind wave heights. J. Geophys. Res., 122, 4500–4517.
doi:10.1002/2016JC012413.
Ardhuin, F., A. Rawat, and J. Aucan, 2014: A numerical model for free infragravity waves: Definition and
validation at regional and global scales. Ocean Modelling, 77, 20–32. doi:10.1016/j.ocemod.2014.02.006.
Ardhuin, F., E. Rogers, A. Babanin, J.-F. Filipot, R. Magne, A. Roland, A. van der Westhuysen,
P. Queffeulou, J.-M. Lefevre, L. Aouf, and F. Collard, 2010: Semi-empirical dissipation source functions
for wind-wave models: part I, definition, calibration and validation. J. Phys. Oceanogr., 40, 1917–1941.
doi:10.1175/2010JPO4324.1.
Ardhuin, F. and A. Roland, 2012: Coastal wave reflection, directional spreading, and seismo-acoustic
noise sources. J. Geophys. Res., 117, C00J20. doi:10.1029/2011JC007832.
Ardhuin, F., P. Sutherland, M. Doble, and P. Wadhams, 2016: Ocean waves across the Arctic: attenu-
ation due to dissipation dominates over scattering for periods longer than 19 s. Geophys. Res. Lett.,
43, 5775–5783. doi:10.1002/2016GL068204.
Ardhuin, F., J. Tournadre, P. Queffelou, and F. Girard-Ardhuin, 2011: Observation and parameteri-
zation of small icebergs: drifting breakwaters in the southern ocean. Ocean Modelling, 39, 405–410.
doi:10.1016/j.ocemod.2011.03.004.
Arhan, M. and R. Ezraty, 1978: Statistical relations between successive wave heights. Oceanol. Acta, 1,
151–158.
Athanassoulis, G. A. and K. A. Belibassakis, 1999: A consistent coupled-mode theory for the propagation
of small amplitude water waves over variable bathymetry regions. J. Fluid Mech., 389, 275–301.
Aubourg, Q., A. Campagne, C. Peureux, F. Ardhuin, J. Sommeria, S. Viboud, and N. Mordant,
2017: 3-wave and 4-wave interactions in gravity wave turbulence. Phys. Rev. Fluids, 2, 114802.
doi:10.1103/PhysRevFluids.2.114802.
Babanin, A., D. Chalikov, I. Young, and I. Savelyev, 2007: Predicting the breaking onset of surface water
waves. Geophys. Res. Lett., 34, L07605. doi:10.1029/2006GL029135.
Babanin, A., I. Young, and M. Banner, 2001: Breaking probabilities for dominant surface waves on water
of finite depth. J. Geophys. Res., 106, 11659–11676.
Backus, G. E., 1962: The effect of the earth rotation on the propagation of ocean waves over long
distances. Deep Sea Res., 9, 185–197.
Banner, M. L., A. V. Babanin, and I. R. Young, 2000: Breaking probability for dominant
waves on the sea surface. J. Phys. Oceanogr., 30, 3145–3160.URL link. doi:10.1175/1520-
0485(2000)030¡3145:BPFDWO¿2.0.CO;2.
Banner, M. L., X. Barthelemy, F. Fedele, M. Allis, A. Benetazzo, F. Dias, and W. L. Peirson, 2014:
Linking reduced breaking crest speeds to unsteady nonlinear water wave group behavior. Phys. Rev.
Lett., 112, 114502.
Banner, M. L., J. R. Gemmrich, and D. M. Farmer, 2002: Multiscale measurement of ocean wave breaking
probability. J. Phys. Oceanogr., 32, 3364–3374.URL link.
Banner, M. L., I. S. F. Jones, and J. C. Trinder, 1989: Wavenumber spectra of short gravity waves. J.
Fluid Mech., 198, 321–344.
Banner, M. L. and R. P. Morison, 2006: On modeling spectral dissipation due to wave breaking for
ocean wind waves. Proceedings of the 9th International workshop on wave hindcasting and forecasting,
Victoria, Canada.
— 2010: Refined source terms in wind wave models with explicit wave breaking prediction.
part I: Model framework and validation against field data. Ocean Modelling, 33, 177–189.
doi:10.1016/j.ocemod.2010.01.002.
Banner, M. L. and W. L. Peirson, 1998: Tangential stress beneath wind-driven air-water interfaces. J.
Fluid Mech., 364, 115–145.
Banner, M. L. and O. M. Phillips, 1974: On the incipient breaking of small scale waves. J. Fluid Mech.,
65, 647–656.
Barber, N. F., F. Ursell, J. Darbyshire, and M. J. Tucker, 1946: A frequency analyser used in the study
of ocean waves. Nature, 329–335.
Barnett, T. P. and A. J. Sutherland, 1968: A note on an overshoot effect in wind-generated waves. J.
Geophys. Res., 73, 6879–6885.
Barrick, D. E., J. M. Headrick, R. W. Bogle, and D. D. Crombie, 1974: Sea backscatter at HF: interpre-
tation and utilization of the echo. Proc. IEEE , 62, 673.
204 BIBLIOGRAPHY

Bates, C. C., 1949: Utilization of wave forecasting in the invasions of Normandy, Burma, and Japan.
Ann. N. Y. Acad. Sci., 545–572.
Battjes, J., 1974: Computations of set-up, longshore currents, run-up and overtopping due to wind-
generated waves. Ph.D. thesis, Delft University of Technology, The Netherlands.
Battjes, J. A. and J. P. F. M. Janssen, 1978: Energy loss and set-up due to breaking of random waves.
Proceedings of the 16th international conference on coastal engineering, ASCE, 569–587.
Becq, F., 1998: Extension de la modélisation spectrale des états de mer vers le domaine côtier . Ph.D.
thesis, Université de Toulon et du Var, France.
Becq-Girard, F., P. Forget, and M. Benoit, 1999: Non-linear propagation of unidirectional wave fields
over varying topography. Coastal Eng., 38, 91–113.
Belcher, S. E., 1999: Wave growth by non-separated sheltering. Eur. J. Mech. B/Fluids, 18, 447–462.
Belibassakis, K. A., G. A. Athanassoulis, and T. P. Gerostathis, 2001: A coupled-mode model for the
refraction-diffraction of linear waves over steep three-dimensional bathymetry. Appl. Ocean Res., 23,
319–336.
Benetazzo, A., 2006: Measurements of short water waves using stereo matched image sequences. Coastal
Eng., 53, 1013–1032.
Benetazzo, A., F. Ardhuin, F. Bergamasco, L. Cavaleri, P. V. Guimarães, M. Schwendeman, M. Sclavo,
J. Thomson, and A. Torsello, 2017: On the shape and likelihood of oceanic rogue waves. Scientific
Reports, 7, 8276. doi:10.1038/s41598-017-07704-9.
Benjamin, T. B. and J. E. Feir, 1967: The disintegration of wav trains on deep water. part 1. theory. J.
Fluid Mech., 27, 417–430.
Benoit, M., P. Frigaard, and H. A. Schäffer, 1997: Analysing multidirectional wave spectra: a tentative
classification of available methods. Proceedings of the 1997 IAHR conference, San Francisco, The Johns
Hopkins University Press, Baltimore, 131–158.
Benoit, M., F. Marcos, and F. Becq, 1996: Development of a third generation shallow-water wave
model with unstructured spatial meshing. Proceedings of the 25th International Conference on Coastal
Engineering, Orlando, ASCE, 465–478.
Bentamy, A., H.-L. Ayina, P. Queffeulou, D. Croize-Fillon, and V. Kerbaol, 2007: Improved near real
time surface wind resolution over the mediterranean sea. Ocean Sci., 112, 259–271.
Berkhoff, J. C. W., 1972: Computation of combined refraction-diffraction. Proceedings of the 13th Inter-
national Conference on Coastal Engineering, Vancouver , ASCE, New York, N. Y., 796–814.
Bidlot, J., P. Janssen, and S. Abdalla, 2005: A revised formulation for ocean wave dissipation in CY25R1.
Technical Report Memorandum R60.9/JB/0516, Research Department, ECMWF, Reading, U. K.
Bidlot, J.-R., 2008: Intercomparison of operational wave forecasting systems against buoys: data from
ECMWF, MetOffice, FNMOC, NCEP, DWD, BoM, SHOM and JMA, September 2008 to Novem-
ber 2008. Technical report, Joint WMO-IOC Technical Commission for Oceanography and Marine
Meteorology, available from http://preview.tinyurl.com/7bz6jj.URL link.
— 2012: Present status of wave forecasting at E.C.M.W.F. Proceedings of ECMWF workshop on ocean
wave forecasting, 25–27 June, 1–16.
Biesel, F., 1950: Etude théorique de la houle en eau courante. La houille blanche, Numéro spécial A,
279–285.
Blachman, N. M. and G. A. McAlpine, 1969: The spectrum of a high-index fm waveform: Woodward’s
theorem revisited. IEEE Trans. Comm. Tech., COM-17, 201–207.
Boccotti, P., 2000: Wave mechanics for ocean engineering. Elsevier, Amsterdam, 496 pp.
Bonmarin, P., 1989: Geometric properties of deep-water breaking waves. J. Fluid Mech., 209, 405–433.
Bonneton, P. and D. Lannes, 2017: Recovering water wave elevation from pressure measurements. J.
Fluid Mech., 833, 399–429. doi:10.1017/jfm.2017.666.
Bonneton, P., V. Marieu, H. Dupuis, N. Sénéchal, and B. Castelle, 2004: Wave transformation and
energy dissipation in the surf zone: comparison between a non-linear model and field data. Journal of
Coastal Research, 39, 329–333.
Booij, N., R. C. Ris, and L. H. Holthuijsen, 1999: A third-generation wave model for coastal regions. 1.
model description and validation. J. Geophys. Res., 104, 7,649–7,666.
Boudière, E., C. Maisondieu, F. Ardhuin, M. Accensi, L. Pineau-Guillou, and J. Lepesqueur, 2013: A
suitable metocean hindcast database for the design of marine energy converters. Int. J. Mar. Energy,
28, e40–e52.
Boussinesq, J., 1872: Théorie des ondes et des remous qui se propagent le long d’un canal rectangulaire
horizontal, en communiquant au liquide contenu dans ce canal des vitesses sensiblement pareilles de
la surface au fond. J. Math. Pures Appl., 17, 55–108.
BIBLIOGRAPHY 205

Boutin, G., F. Ardhuin, D. Dumont, C. Sévigny, and F. Girard-Ardhuin, 2018: Floe size effects on wave-
ice interactions: theoretical background, implementation and applications. J. Geophys. Res., 123,
4779–4805. doi:10.1029/2017JC013622.
Bragg, W. L., 1913: The structure of some crystals as indicated by their diffraction of x-rays. Proc. Roy.
Soc. Lond. A, 89, 248–277.
Breivik, Ø., J.-R. Bidlot, and P. A. Janssen, 2016: A Stokes drift approximation based on the phillips
spectrum. Ocean Modelling. doi:10.1016/j.ocemod.2016.01.005.
Bréon, F. M. and N. Henriot, 2006: Spaceborne observations of ocean glint reflectance and modeling of
wave slope distributions. J. Geophys. Res., 111, C0605. doi:10.1029/2005JC003343.
Bretherton, F. P. and C. J. R. Garrett, 1968: Wavetrains in inhomogeneous moving media. Proc. Roy.
Soc. of London, A302, 529–554.
Brocchini, M., A. B. Kennedy, L. Soldini, and A. Mancinelli, 2004: Topographically controlled, breaking-
wave-induced macrovortices. part 3. the mixing features. J. Fluid Mech., 507, 289–307.
Broche, P., J. C. de Maistre, and P. Forget, 1983: Mesure par radar décamétrique cohérent des courants
superficiels engendrés par le vent. Oceanol. Acta, 6, 43–53.
Bromirski, P. D., 2001: Vibration from the Perfect Storm. Geochemistry Geophysics Geosystems, 2,
2000GC000119. doi:10.1029/2004GC000768.
Bromirski, P. D., D. R. Cayan, and R. E. Flick, 2005: Wave spectral energy variability in the northeast
Pacific. J. Geophys. Res., 110, C03005. doi:10.1029/2004JC002398.
Brown, G. S., 1977: The average impulse response of a rough surface and its applications. IEEE J.
Oceanic Eng., 2, 67–63. doi:10.1109/JOE.1977.1145328.
Brumer, S. E., C. J. Zappa, B. W. Blomquist, C. W. Fairall, A. Cifuentes-Lorenzen, J. B. Edson, I. M.
Brooks, and B. J. Huebert, 2017: Wave-related reynolds number parameterizations of CO2 and DMS
transfer velocities. Geophys. Res. Lett., 44, 9865–9875.
Businger, J. A., J. C. Wyngaard, I. Izumi, and E. F. Bradley, 1971: Flux-profile relationships in the
atmospheric surface layer. J. Atmos. Sci., 28, 181–189.URL link.
Caires, S. and A. Sterl, 2005: A new nonparametric method to correct model data: Application to
significant wave height from the ERA-40 re-analysis. J. Atmos. Ocean Technol., 22, 443–459.
Cancouët, R., 2008: Évolution multi-échelles des états de mer en zone côtière (Multiscale evolution of
sea states in the coastal ocean). Master’s thesis, ENSIETA, in French.
Carrier, G. F. and H. P. Greenspan, 1958: Water waves of finite amplitude on a sloping beach. J. Fluid
Mech., 4, 97–109.
Cartwright, D. E. and M. S. Longuet-Higgins, 1956: The statistical distribution of the maxima of a
random function. Proc. Roy. Soc. Lond. A, 237, 212–232.
Cartwright, D. E. and N. D. Smith, 1964: Buoy techniques for obtaining directional wave spectra. Buoy
technology, transactions of the International Buoy technology symposium, Washington, D. C., National
Academy of Sciences, Marine Technology society, 111–136.
Caudal, G., D. Hauser, R. Valentin, and C. L. Gac, 2014: KuROS: A new airborne ku-band doppler
radar for observation of surfaces. J. Atmos. Ocean Technol., 31, 2023–2245. doi:10.1175/JTECH-D-
14-00013.1.
Cavaleri, L. and L. Bertotti, 2006: The improvement of modelled wind and wave fields with increasing
resolution. Tellus, 33, 553–565.
Cavaleri, L., S. Curiotto, G. D. Porta, and A. Mazzoldi, 1981: Directional wave recording in the northern
Adriatic sea. Nuovo Cimento, 4C, 519–534.
CERC, 1977: Shore protection manual , volume 3 volumes. U. S. Army Coastal Engineering Research
Center.
Chalikov, D., 1993: Comments on ”wave-induced stress and the drag of air flow over sea waves” and
”quasi-linear theory of wind wave generation applied to wave forecasting”. J. Phys. Oceanogr., 23,
1597–1600.
Chapron, B., F. Collard, and F. Ardhuin, 2005: Direct measurements of ocean surface velocity from
space: interpretation and validation. J. Geophys. Res., 110. doi:10.1029/2004JC002809.
Charles, E., D. Idier, J. Thiebot, G. L. Cozannet, R. Pedreros, F. Ardhuin, and S. Planton, 2012: Present
wave climate in the bay of biscay: Spatiotemporal variability and trends from 1958 to 2001. Journal
of Climate, 25, 2020–2039. doi:10.1175/JCLI-D-11-00086.1.
Charnock, H., 1955: Wind stress on a water surface. Quart. Journ. Roy. Meteorol. Soc., 81, 639–640.
Chawla, A. and J. T. Kirby, 2002: Monochromatic and random wave breaking at blocking points. J.
Geophys. Res., 107, 3067.
Chawla, A. and H. L. Tolman, 2008: Obstruction grids for spectral wave models. Ocean Modelling, 22,
206 BIBLIOGRAPHY

12–25.
Chelton, D. B., E. J. Walsh, and J. L. MacArthur, 1989: Pulse compression and sea
level tracking in satellite altimetry. J. Atmos. Ocean Technol., 6, 407–438. doi:10.1175/1520-
0426(1989)006¡0407:pcaslt¿2.0.co;2.
Chen, G., B. Chapron, R. Ezraty, and D. Vandemark, 2002: A global view of swell and wind sea climate
in the ocean by satellite altimeter and scatterometer. J. Atmos. Ocean Technol., 19, 1849–1859.
Church, J. C. and E. B. Thornton, 1993: Effects of breaking wave induced turbulence within a longshore
current model. Coastal Eng., 20, 1–28.
Clark, D. B., L. Lenain, F. Feddersen, E. Boss, and R. T. Guza, 2014: Aerial imaging of fluorescent dye
in the near shore. J. Atmos. Ocean Technol., 31, 1410–. doi:10.1175/JTECH-D-13-00230.1.
Cole, D. M., 1995: A model for the anelastic straining of saline ice subjected to cyclic loading. Phil. Mag.
A, 72, 231–248.
Cole, D. M. and G. D. Durell, 2001: A dislocation-based analysis of strain history effects in ice. Phil.
Mag. A, 81, 1849–1872. doi:10.1080/01418610108216640.
Cole, D. M., R. A. Johnson, and G. D. Durell, 1998: Cyclic loading and creep response of aligned
first-year sea ice. J. Geophys. Res., 103, 21,751–21,758.
Collard, F., F. Ardhuin, and B. Chapron, 2005: Extraction of coastal ocean wave fields from SAR images.
IEEE J. Oceanic Eng., 30, 526–533.
— 2009: Monitoring and analysis of ocean swell fields using a spaceborne SAR: a new method for routine
observations. J. Geophys. Res., 114, C07023.
Compo, G. P., J. S. Whitaker, P. D. Sardeshmukh, N. Matsui, R. J. Allan, X. Yin, J. B. E. Gleason,
B. E., R. S. Vose, G. Rutledge, P. Bessemoulin, S. Brönnimann, M. Brunet, R. I. Crouthamel, A. N.
Grant, P. Y. Groisman, P. D. Jones, M. C. Kruk, A. C. Kruger, G. J. Marshall, M. Maugeri, H. Y.
Mok, O. Nordli, T. F. Ross, R. M. Trigo, X. L. Wang, S. D. Woodruff, and S. J. Worley, 2011: The
twentieth century reanalysis. Quart. Journ. Roy. Meteorol. Soc., 137, 1–28. doi:10.1002/qj.776.
Cote, L. J., J. O. Davis, W. Marks, R. J. McGough, E. Mehr, W. J. Pierson, Jr., J. F. Ropek, G. Stephen-
son, and R. C. Vetter, 1960: The directional spectrum of a wind generated sea as determined from
data obtained by the Stereo Wave Observation Project. Technical Report 6, N. Y. U. Coll. of Eng.
Courant, R., K. Friedrichs, and H. Lewy, 1928: Über die partiellen differenzengleichungen der mathema-
tischen physik. Mathematische Annalen, 100, 32–74. doi:10.1007/BF01448839.
Cox, C. and W. Munk, 1954: Measurement of the roughness of the sea surface from photographs of the
sun’s glitter. J. Opt. Soc. Am., 44, 838–850. doi:10.1364/josa.44.000838.
Cox, D., 1995: Experimental and numerical modelling of surf zone hydrodynamics. Ph.D. thesis, Univ.
of Delaware, Newark.
Craig, P. D. and M. L. Banner, 1994: Modeling wave-enhanced turbulence in the ocean surface layer. J.
Phys. Oceanogr., 24, 2546–2559.URL link.
Craik, A. D. D., 2004: The origins of water wave theory. Annu. Rev. Fluid Mech., 36, 1–28.
Crosby, S. C., B. D. Cornuelle, W. C. O’Reilly, and R. T. Guza, 2017: Assimilating global wave model
predictions and deep-water wave observations in nearshore swell predictions. J. Atmos. Ocean Technol.,
34, 1823–1836. doi:10.1175/JTECH-D-17-0003.1.
Csı́k, Á., M. Ricchiuto, and H. Deconinck, 2002: A conservative formulation of the multidimensional
upwind residual distribution schemes for general nonlinear conservation laws. J. Comp. Phys., 172,
286–312.
Dalrymple, R. A., 1974: A finite amplitude wave on a linear shear current. J. Geophys. Res., 79, 4498–
4504. doi:10.1029/jc079i030p04498.
Darbyshire, J., 1957: Attenuation of swell in the north atlantic ocean. Quart. Journ. Roy. Meteorol.
Soc., 83, 351–359.
Darrigol, O., 2003: The spirited horse, the engineer, and the mathematician: Water waves in nineteenth-
century hydrodynamics. Arch. Hist. Exact Sci., 58, 21–95. doi:10.1007/s00407-003-0070-5.
Davies, A. G. and C. Villaret, 1999: Eulerian drift induced by progressive waves above rippled and very
rough beds. J. Geophys. Res., C1, 1465–1488.
Davis, R. E. and L. A. Regier, 1977: Methods for estimating directional wave spectra from multi-element
arrays. J. Mar. Res., 35, 453–478.
de Caliginy, A., 1878: Expériences sur les mouvements des molécules liquides des ondes courantes,
considérées dans leur mode d’action sur la marche des navires. C. R. Acad. Sci. Paris, 87, 1019–1023.
De Carlo, M. and F. Ardhuin, 2024: Along-track resolution and uncertainty of altimeter-derived wave
height and sea level: re-defining the significant wave height in extreme storms. J. Geophys. Res., in
press. doi:10.1029/2023JC020832.
BIBLIOGRAPHY 207

De Carlo, M., F. Ardhuin, A. Ollivier, and A. Nigou, 2023: Wave groups and small scale variability of wave
heights observed by altimeters. J. Geophys. Res., 133, e2023JC019740. doi:10.1029/2023JC019740.
de Carolis, G., P. Olla, and L. Pignagnoli, 2005: Effective viscosity of grease ice in linearized gravity
waves. J. Fluid Mech., 535, 369–381. doi:10.1017/S002211200500474X.
de Jong, M. P. C., L. H. Holthuijsen, and J. A. Battjes, 2003: Generation of seiches by cold fronts over
the southern north sea. J. Geophys. Res., 108, 3117. doi:10.1029/2002JC001422.
de Laplace, P. S., 1776: Suite des recherches sur plusieurs points du système du monde (XXV–XXVII).
Mém. Présentés Acad. R. Sci. Inst. France, 542–552.
Dean, R. G., 1965: Stream function representation of nonlinear ocean waves. J. Geophys. Res., 70,
4561–4572.
Dean, R. G. and R. A. Dalrymple, 1991: Water wave mechanics for engineers and scientists. World
Scientific, Singapore, second edition, 353 pp.
Dee, D. P., S. M. Uppala, A. J. Simmons, P. Berrisford, P. Poli, S. Kobayashi, U. Andrae, M. A. Bal-
maseda, G. Balsamo, P. Bauer, P. Bechtold, A. C. M. Beljaars, L. van de Berg, J. Bidlot, N. Bormann,
C. Delsol, R. Dragani, M. Fuentes, A. J. Geer, L. Haimbergere, S. B. Healy, H. Hersbach, E. V. Holm,
L. Isaksena, P. Køallberg, M. Köhler, M. Matricardi, A. P. McNally, B. M. Monge-Sanz, J.-J. Mor-
crette, B.-K. Park, C. Peubey, P. de Rosnay, C. Tavolato, J.-N. Thépaut, and F. Vitart, 2011: The
era-interim reanalysis: configuration and performance of the data assimilation system. Quart. Journ.
Roy. Meteorol. Soc., 137, 553–597. doi:10.1002/qj.828.
Delpey, M., F. Ardhuin, F. Collard, and B. Chapron, 2010: Space-time structure of long swell systems.
J. Geophys. Res., 115, C12037. doi:10.1029/2009JC005885.
Delpey, M. T., F. Ardhuin, P. Otheguy, A. Jouon, and F. Ardhuin, 2014: Effects of waves on coastal
water dispersion in a small estuarine bay. J. Geophys. Res., 119, 1–17. doi:10.1002/2013JC009466.
Dias, F. and P. Milewski, 2010: On the fully-nonlinear shallow-water generalized Serre equations. Phys.
Lett. A, 374, 1049–1053.
Dibarboure, G., F. Boy, J. D. Desjonqueres, S. Labroue, Y. Lasne, N. Picot, J. C. Poisson, and P. Thibaut,
2014: Investigating short-wavelength correlated errors on low-resolution mode altimetry. J. Atmos.
Ocean Technol., 31, 1337–1362. doi:10.1175/JTECH-D-13-00081.1.
Dingemans, M. W., 1997: Water wave propagation over uneven bottoms. Part 1 linear wave propagation.
World Scientific, Singapore, 471 p.
Dingler, J. R., 1974: Wave-formed ripples in nearshore sands. Ph.D. thesis, University of California, San
Diego.
Doble, M. J., M. D. Coon, and P. Wadhams, 2003: Pancake ice formation in the weddell sea. J. Geophys.
Res., 108, 3209. doi::10.1029/2002JC001373.
Doble, M. J., H. Skourup, P. Wadhams, and C. A. Geiger, 2011: The relation between arctic sea ice
surface elevation and draft: A case study using coincident auv sonar and airborne scanning laser. J.
Geophys. Res., 116, COOE03. doi:10.1029/2011JC007076.
Dobson, F. W., 1971: Measurements of atmospheric pressure on wind-generated sea waves. J. Fluid
Mech., 48, 91–127.
Dobson, R. S., 1967: Some applications of a digital computer to hydraulic engineering problems. Technical
Report 80, Department of Civil Engineering, Stanford University.
Dodet, G., S. Abdalla, M. Alday, M. Accensi, J. Bidlot, and F. Ardhuin, 2022: Error characteriza-
tion of significant wave heights in multidecadal satellite altimeter product, model hindcast, and in
situ measurements using the triple collocation technique. J. Atmos. Ocean Technol., 39, 887–901.
doi:10.1175/JTECH-D-21-0179.1.
Dodet, G., X. Bertin, and R. Taborda, 2010: Wave climate variability in the north-east Atlantic ocean
over the last six decades. Ocean Modelling, 31, 120–131.
Dodet, G., F. Leckler, D. Sous, F. Ardhuin, J. Filipot, and S. Suanez, 2018: Wave runup over steep
rocky cliffs. J. Geophys. Res., 123. doi:10.1029/2018JC013967.
Dodet, G., J.-F. Piolle, Y. Quilfen, S. Abdalla, M. Accensi, F. Ardhuin, E. Ash, J.-R. Bidlot, C. Gom-
menginger, G. Marechal, M. Passaro, G. Quartly, J. Stopa, B. Timmermans, I. Young, P. Cipollini,
and C. Donlon, 2020: The sea state cci dataset v1: towards a sea state climate data record based on
satellite observations. Earth System Sci. Data, 12, 1929–1951. doi:10.5194/essd-12-1929-2020.
Donelan, M., M. Skafel, H. Graber, P. Liu, D. Schwab, and S. Venkatesh, 1992: On the growth rate of
wind-generated waves. Atmosphere Ocean, 30, 457–478.
Donelan, M. A., A. V. Babanin, I. R. Young, and M. L. Banner, 2006: Wave-follower field measurements
of the wind-input spectral function. Part II: parameterization of the wind input. J. Phys. Oceanogr.,
36, 1672–1689.
208 BIBLIOGRAPHY

Donelan, M. A., F. W. Dobson, H. C. Graber, N. Madsen, and C. Mccormick, 2005: Measurement of

wind waves and wave-coherent air pressures on the open sea from a moving SWATH vessel. J. Atmos.
Ocean Technol., 22, 896–906.
Donelan, M. A., W. M. Drennan, and A. K. Magnusson, 1996: Nonstationary analysis of the directional
properties of propagating waves. J. Phys. Oceanogr., 26, 1901–1914.URL link.
Donelan, M. A., J. Hamilton, and W. H. Hui, 1985: Directional spectra of wind-generated waves. Phil.
Trans. Roy. Soc. London A, 315, 509–562.
Dore, B. D., 1978: Some effects of the air-water interface on gravity waves. Geophys. Astrophys. Fluid.
Dyn., 10, 215–230.
Drennan, W. M., H. C. Graber, D. Hauser, and C. Quentin, 2003: On the wave age dependence of wind
stress over pure wind seas. J. Geophys. Res., 108, 8062. doi:10.1029/2000JC00715.
Duennebier, F. K., R. Lukas, E.-M. Nosal, J. Aucan, and R. A. Weller, 2012: Wind, waves, and acoustic
background levels at station ALOHA. J. Geophys. Res., 117, C03017. doi:10.1029/2011JC007267.
Dulov, V., V. Kudryavtsev, and A. Bolshakov, 2002: A field study of white caps coverage and its
modulations by energy containing wave. Gas Transfer at the Water Surface, Geophysical Monographs
127 , M. A. D. et al., ed., American Geophysical Union, pages 187–192.
Dulov, V. A. and M. V. Kosnik, 2009: Effects of three-wave interactions in the gravity-capillary range
of wind waves. Izv. Atmos. Ocean. Phys., 45, 380–391.
Duncan, J. H., 1981: An experimental investigation of breaking waves produced by a towed hydrofoil.
Proc. Roy. Soc. Lond. A, 377, 331–348.
— 1983: The breaking and non-breaking wave resistance of a two-dimensional hydrofoil. J. Fluid Mech.,
126, 507–520.
Dysthe, K. B., 2001: Refraction of gravity waves by weak current gradients. J. Fluid Mech., 442, 157–159.
Edson, J. B., V. Jampana, R. A. Weller, S. P. Bigorre, A. J. Plueddemann, C. W. Fairall, S. D. Miller,
L. Mahrt, D. Vickers, and H. Hersbach, 2013: On the exchange of momentum over the open ocean. J.
Phys. Oceanogr., 43, 1589–1610. doi:10.1175/JPO-D-12-0173.1.
Ehrenmark, U. T., 2005: An alternative dispersion equation for water waves over an inclined bed. J.
Fluid Mech., 543, 249–266.
Eldeberky, Y. and J. Battjes, 1995: Parameterization of triad interactions in wave energy models. Coastal
Dynamics ’95 , W. Dally and R. Zeidler, eds., ASCE, 140–148.
Elfouhaily, T., B. Chapron, K. Katsaros, and D. Vandemark, 1997: A unified directional spectrum for
long and short wind-driven waves. J. Geophys. Res., 102, 15781–15796. doi:10.1029/97jc00467.
Elfouhaily, T., M. Joelson, S. Guignard, H. Branger, D. R. Thompson, B. Chapron, and D. Vandemark,
2003: Analysis of random nonlinear water waves: the Stokes-Woodward technique. C. R. Acad. Sci.
Paris, 331, 189–196.
Elfouhaily, T., D. R. Thompson, D. Vandemark, and B. Chapron, 2000: Truncated hamiltonian versus
surface perturbation in nonlinear wave theories. Waves in Random Media, 10, 103–116.
Elgar, S. and R. T. Guza, 1985: Observation of bispectra of shoaling surface gravity waves. J. Fluid
Mech., 161, 425–448.
Elgar, S., R. T. Guza, B. Raubenheimer, T. H. C. Herbers, and E. L. Gallagher, 1997: Spectral evolution
of shoaling and breaking waves on a barred beach. J. Geophys. Res., 102, 15797–15805.
Elgar, S., T. H. C. Herbers, and R. T. Guza, 1994: Reflection of ocean surface gravity waves from a
natural beach. J. Phys. Oceanogr., 24, 1,503–1,511.URL link.
Elgar, S. and B. Raubenheimer, 2008: Wave dissipation by muddy seafloors. Geophys. Res. Lett., 35,
L07611, 10.1029/2008GL033245.
Elliot, J. A., 1972: Microscale pressure fluctuations measured within the lower atmospheric boundary
layer. J. Fluid Mech., 53, 351–384.
Esau, I. N., 2007: Amplification of turbulent exchange over wide arctic leads: Large-eddy simulation
study. J. Geophys. Res., 112, D08109. doi:10.1029/2006JD007225.
Ewans, K. C., 1998: Observations of the directional spectrum of fetch-limited waves. J. Phys. Oceanogr.,
28, 495–512.URL link.
Fabrikant, A. L., 1976: Quasilinear theory of wind-wave generation. Izv. Atmos. Ocean. Phys., 12,
524–526.
Farquharson, G., S. J. Frasier, B. Raubenheimer, and S. Elgar, 2005: Microwave radar cross sections
and doppler velocities measured in the surf zone. IEEE Trans. on Geosci. and Remote Sensing, 110,
C12024. doi:10.1029/2005JC003022.
Farrell, W. E. and W. Munk, 2008: What do deep sea pressure fluctuations tell about short surface
waves? Geophys. Res. Lett., 35, L19605. doi:10.1029/2008GL035008.
BIBLIOGRAPHY 209

Feddersen, F., E. L. Gallagher, R. T. Guza, and S. E. Feddersen, 2003: The drag coefficient, bottom
roughness, and wave-breaking in the nearshore. Coastal Eng., 48, 189–195.
Fedele, F., A. Benetazzo, G. Gallego, P.-C. Shih, A. Yezzi, F. Barbariol, and F. Ardhuin, 2013: Space-time
measurements of oceanic sea states. Ocean Modelling, 70, 103–115. doi:10.1016/j.ocemod.2013.01.001.
Fedele, F., G. Gallego, A. Benetazzo, A. Yezzi, and M. A. Tayfun, 2009: Euler characteristics and
maxima of oceanic sea states. Proceedings of the Rogue waves workshop, October 2008, Brest, France,
Ifremer, 145–154.
Fermi, E., J. Pasta, and S. Ulam, 1955: Studies of nonlinear problems. i. Nonlinear wave motion. Lectures
in applied mathematics, vol. 15 , A. C. Newell, ed., Amer. Math. Soc., Providence, R.I., pages 143–156.
Filipot, J.-F. and F. Ardhuin, 2012: A unified spectral parameterization for wave breaking: from the
deep ocean to the surf zone. J. Geophys. Res., 117, C00J08. doi:10.1029/2011JC007784.
Filipot, J.-F., F. Ardhuin, and A. Babanin, 2010a: A unified deep-to-shallow-water wave-breaking prob-
ability parameterization. J. Geophys. Res., 115, C04022. doi:10.1029/2009JC005448.
Filipot, J.-F., F. Ardhuin, and R. Magne, 2010b: Validation d’un terme source de dissipation par
déferlement applicable du large à la côte. Actes des Xèmes journées Génie côtier-Génie civil, Les
Sables d’Olonne, Centre Français du Littoral, 87–94.URL link.
Filloux, J. H., 1980: Pressure fluctuations on the open ocean floor over a broad frequency range: new
program and early results. J. Phys. Oceanogr., 10, 1959–1971.
Fox, C. and T. G. Haskell, 2001: Ocean wave speed in the antarctic marginal ice zone. Annales Geophys-
icae, 33, 350–354. doi:10.3189/172756401781818239.
Freilich, M. H. and B. A. Vanhoff, 2003: The relationship between winds, surface roughness, and radar
backscatter at low incidence angles from TRMM precipitation radar measurements. J. Atmos. Ocean
Technol., 20, 549–562.
Gain, L., 1918: La prédiction des houles au Maroc. Annales Hydrographiques, 65–75.
Gallagher, E. L., S. Elgar, and E. B. Thornton, 1998: Megaripple migration in a natural surf zone.
Nature, 394, 165–168.
Gallego, G., A. Yezzi, F. Fedele, and A. Benetazzo, 2011: A variational stereo method for the three-
dimensional reconstruction of ocean waves. IEEE Trans. on Geosci. and Remote Sensing, 49, 4445–
4457. doi:10.1109/TGRS.2011.2150230.
Garcı́a-Nava, H., F. J. Ocampo-Torres, P. A. Hwang, and P. Osuna, 2012: Reduction of wind stress due
to swell at high wind conditions. J. Geophys. Res., 117, C00J11. doi:10.1029/2011JC007833.
Garrett, C., 1976: Generation of Langmuir circulations by surface waves - a feedback mechanism. J.
Mar. Res., 34, 117–130.
Gelci, R., H. Cazalé, and J. Vassal, 1957: Prévision de la houle. La méthode des densités spectroangu-
laires. Bulletin d’information du Comité d’Océanographie et d’Etude des Côtes, 9, 416–435.
Goda, Y., 1985: Random seas and design of marine structures. University of Tokyo Press, 323 p.
Godin, O. A., N. A. Zabotin, A. F. Sheehan, Z. Yang, and J. A. Collins, 2013: Power spectra of
infragravity waves in a deep ocean. Geophys. Res. Lett., 40, 2159–2165. doi:10.1002/grl.50321.
Gorman, R. M., 2003: The treatment of discontinuities in computing the nonlinear energy transfer for
finite-depth gravity wave spectra. J. Atmos. Ocean Technol., 20, 206–216.
Gow, A. J., S. F. Ackley, W. F. Weeks, and J. W. Govoni, 1982: physical and structural characteristics
of Antarctic sea ice. Annales Geophysicae, 3, 113–117.
Graber, H., E. Terray, M. Donelan, W. Drennan, J. V. Leer, and D. Peters, 2000: ASIS – a new air-sea
interaction spar buoy: design and performance at sea. J. Atmos. Ocean Technol., 17, 708–720.
Graber, H. C. and O. S. Madsen, 1988: A finite-depth wind-wave model. part 1: model description. J.
Phys. Oceanogr., 18, 1465–1483.URL link.
Grant, W. D. and O. S. Madsen, 1979: Combined wave and current interaction with a rough bottom. J.
Geophys. Res., 84, 1797–1808.
Greenslade, D. J. M., 2001: A wave modelling study of the 1998 Sydney to Hobart yacht race. Aust.
Met. Mag., 50, 53–63.
Groeneweg, J., M. van Gent ad Joana van Nieuwkoop, and Y. Toledo4, 2015: Wave propagation into
complex coastal systems and the role of nonlinear interactions. J. of Waterway, Port Coast. Ocean
Eng., 141, 04015003. doi:10.1061/(ASCE)WW.1943-5460.0000300.
Halimi, A., 2013: From conventional to delay/Doppler altimetry. Ph.D. thesis, INP Toulouse, Toulouse,
France.URL link.
Hanafin, J., Y. Quilfen, F. Ardhuin, J. Sienkiewicz, P. Queffeulou, M. Obrebski, B. Chapron, N. Reul,
F. Collard, D. Corman, E. B. de Azevedo, D. Vandemark, and E. Stutzmann, 2012: Phenomenal sea
states and swell radiation: a comprehensive analysis of the 12-16 February 2011 North Atlantic storms.
210 BIBLIOGRAPHY

Bull. Amer. Meterol. Soc., 93, 1825–1832. doi:10.1175/BAMS-D-11-00128.1.

Hara, T. and S. E. Belcher, 2002: Wind forcing in the equilibrium range of wind-wave spectra. J. Fluid
Mech., 470, 223–245.
Harcourt, R. R. and E. A. D’Asaro, 2006: Large Eddy Simulation of Langmuir Turbulence in Pure Wind
Seas. AGU Fall Meeting Abstracts, A562+.
Hargreaves, J. C. and J. D. Annan, 2000: Comments on –improvement of the short-fetch behavior in the
wave ocean model (WAM)–. J. Atmos. Ocean Technol., 18, 711–715.URL link.
Hasselmann, D. and J. Bösenberg, 1991: Field measurements of wave-induced pressure over wind-sea
and swell. J. Fluid Mech., 230, 391–428.
Hasselmann, K., 1960: Grundgleichugen der seegangsvorhersage. Schifftechnik , 7, 191–195.
— 1962: On the non-linear energy transfer in a gravity wave spectrum, part 1: general theory. J. Fluid
Mech., 12, 481–501.
— 1970: Wave-driven inertial oscillations. Geophys. Fluid Dyn., 1, 463–502.
doi:10.1080/03091927009365783.
— 1974: On the spectral dissipation of ocean waves due to white capping. Boundary-Layer Meteorol., 6,
107–127.
Hasselmann, K., T. P. Barnett, E. Bouws, H. Carlson, D. E. Cartwright, K. Enke, J. A. Ewing, H. Gien-
app, D. E. Hasselmann, P. Kruseman, A. Meerburg, P. Müller, D. J. Olbers, K. Richter, W. Sell, and
H. Walden, 1973: Measurements of wind-wave growth and swell decay during the Joint North Sea
Wave Project. Deut. Hydrogr. Z., 8, 1–95, suppl. A.
Hasselmann, K. and J. I. Collins, 1968: Spectral dissipation of finite depth gravity waves due to turbulent
bottom friction. J. Mar. Res., 26, 1–12.
Hasselmann, K. and S. Hasselmann, 1991: On the nonlinear mapping of an ocean wave spectrum into a
synthetic aperture radar image spectrum and its inversion. J. Geophys. Res., 96, 10713–10729.
Hasselmann, K., W. Munk, and G. MacDonald, 1963: Bispectra of ocean waves. Time series analysis,
M. Rosenblatt, ed., Wiley, 125–139.
Hasselmann, K., R. K. Raney, W. J. Plant, W. Alpers, R. A. Shuchman, D. R. Lyzenga, C. L. Rufenach,
and M. J. Tucker, 1985a: Theory of Synthetic Aperture Radar ocean imaging: a MARSEN view. J.
Geophys. Res., 90, 4659–4686.
Hasselmann, S., K. Hasselmann, J. Allender, and T. Barnett, 1985b: Computation and parameterizations
of the nonlinear energy transfer in a gravity-wave spectrum. Part II: Parameterizations of the nonlinear
energy transfer for application in wave models. J. Phys. Oceanogr., 15, 1378–1391.URL link.
Hauser, D., G. Caudal, G. J. Rijckenberg, D. Vidal-Majar, G. Laurent, and P. Lancelin, 1992: RESSAC:
a new airborne FM/CW radar ocean wave spectrometer. IEEE Trans. on Geosci. and Remote Sensing,
30, 981–995.
Hauser, D., C. Tourain, L. Hermozo, D. Alraddawi, L. Aouf, B. Chapron, A. Dalphinet, L. Delaye,
M. Dalila, E. Dormy, F. Gouillon, V. Gressani, A. Grouazel, G. Guitton, R. Husson, A. Mironov,
A. Mouche, A. Ollivier, L. Oruba, F. Piras, R. R. Suquet, P. Schippers, C. Tison, and N. Tran,
2021: New observations from the SWIM radar on-board CFOSAT: Instrument validation and
ocean wave measurement assessment. IEEE Trans. on Geosci. and Remote Sensing, 59, 5–26.
doi:10.1109/tgrs.2020.2994372.
Hay, A., C. Watson, B. Legresy, M. King, and J. Beardsley, 2023: In situ validation of altimetry and
CFOSAT SWIM measurements in a high wave environmen. J. Atmos. Ocean Technol., 40, 1137–1152.
doi:10.1175/JTECH-D-23-0031.1.
Heathershaw, A. D., 1982: Seabed-wave resonance and sand bar growth. Nature, 296, 343–345.
Hemer, M. A., J. Katzfey, and C. E. Trenham, 2013: Global dynamical projections of surface ocean wave
climate for a future high greenhouse gas emission scenario. Ocean Modelling, 70, 221–245.
Herbers, T. H. C. and M. C. Burton, 1997: Nonlinear shoaling of directionally spread waves on a beach.
J. Geophys. Res., 102, 21,101–21,114.
Herbers, T. H. C., S. Elgar, and R. T. Guza, 1994: Infragravity-frequency (0.005-0.05 Hz) motions on
the shelf, part I, forced waves. J. Phys. Oceanogr., 24, 917–927.URL link.
— 1995: Infragravity-frequency (0.005-0.05 Hz) motions on the shelf. part II: free waves. J. Phys.
Oceanogr., 25, 1063–1079.URL link.
Herbers, T. H. C. and R. T. Guza, 1990: Estimation of directional wave spectra from multicomponent
observations. J. Phys. Oceanogr., 20, 1703–1724.URL link.
Herbers, T. H. C., T. T. Janssen, D. B. Colbert, and J. H. MacMahan, 2012: Observing ocean surface
waves with gps-tracked buoys. J. Atmos. Ocean Technol., 29, 944–959.
Herbers, T. H. C. and S. J. Lentz, 2010: Observing directional properties of ocean swell with an acoustic
BIBLIOGRAPHY 211

doppler current profiler (ADCP). J. Atmos. Ocean Technol., 27, 210–225.URL link.
Herbers, T. H. C., R. L. Lowe, and R. T. Guza, 1992: Field observations of orbital velocities and pressure
in weakly nonlinear surface gravity waves. J. Fluid Mech., 245, 413–435.
Hersbach, H., B. Bell, P. Berrisford, S. Hirahara, A. Horányi, J. Muñoz-Sabater, J. Nicolas, C. Peubey,
R. Radu, D. Schepers, A. Simmons, C. Soci, S. Abdalla, X. Abellan, G. Balsamo, P. Bechtold, G. Bi-
avati, J. Bidlot, M. Bonavita, G. D. Chiara, P. Dahlgren, D. Dee, M. Diamantakis, R. Dragani,
J. Flemming, R. Forbes, M. Fuentes, A. Geer, L. Haimberger, S. Healy, R. J. Hogan, E. Hólm,
M. Janisková, S. Keeley, P. Laloyaux, P. Lopez, C. Lupu, G. Radnoti, P. de Rosnay, I. Rozum,
F. Vamborg, S. Villaume, and t. . T. E. g. r. Jean-Noël Thépaut”, 2020: Quart. Journ. Roy. Meteorol.
Soc., 146, 1999–2049. doi:10.1002/qj.3803.
Herterich, K. and K. Hasselmann, 1980: A similarity relation for the non-linear energy transfer in a
finite-depth gravity-wave spectrum. J. Fluid Mech., 97, 215–224.
Hoefel, F. and S. Elgar, 2003: Wave-induced sediment transport and sandbar migration. Science, 299,
1885–1887.
Hogstrom, U., A. Smedman, E. Sahlee, W. M. Drennan, K. K. Kahma, H. Pettersson, and F. Zhang,
2009: The atmospheric boundary layer during swell: A field study and interpretation of the turbulent
kinetic energy budget for high wave ages. J. Atmos. Sci., 66, 2764–2779.URL link.
Holman, R., 1986: Extreme value statistics for wave run-up on a natural beach. Coastal Eng., 9, 527–544.
Holthuijsen, L., 1983: Stereophotography of ocean waves. Appl. Ocean Res., 5, 204–209.
— 2007: Waves in Oceanic and Coastal Waters. Cambridge University Press, Cambridge.
Holthuijsen, L. H. and T. H. C. Herbers, 1986: Statistics of breaking waves observed as whitecaps in the
open sea. J. Phys. Oceanogr., 16, 290–297.URL link.
Holthuijsen, L. H., M. D. Powell, and J. D. Pietrzak, 2012: Wind and waves in extreme hurricanes. J.
Geophys. Res., 117, C09003. doi:10.1029/2012JC007983.
Huang, N. E., 1979: On surface drift currents in the ocean. J. Fluid Mech., 91, 191–208.
Huang, N. E., Z. Shen, S. R. Long, M. C. Wu, H. H. Shih, Q. Zheng, N.-C. Yen, C. C. Tung, and H. H. Liu,
1998: The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary
time series analysis. Proc. Roy. Soc. Lond. A, 454, 903–995.
Huchet, M., F. Leckler, J.-F. Filipot, A. Roland, F. Ardhuin, M. D. Sikiric, H. Michaud, M. Delpey,
and G. Dodet, 2015: On the high resolution coastal applications with WAVEWATCH III. Proceedings,
14th Int. Workshop of Wave Hindcasting and Forecasting, Key West, Florida.URL link.
Hunt, I. A., 1959: Design of seawalls and breakwaters. Journal of Waterways and Harbours Division,
85, 123–152.
Hwang, P. A. and D. W. Wang, 2004: Field measurements of duration-limited growth of wind-generated
ocean surface waves at young stages of development. J. Phys. Oceanogr., 34, 2316–2326.URL link.
Hwang, P. H., D. W. Wang, E. J. Walsh, W. B. Krabill, and R. N. Swift, 2000: Airborne measurement of
the wavenumber spectra of ocean surface waves. part ii: directional distribution. J. Phys. Oceanogr.,
30, 2768–2787.URL link.
Iribarren, R. and C. Nogales, 1949: Protection des ports. Proceedings XVIIth International Navigation
Congress, Section II, Communication, 4, Lisbon, 31–80.
Ivonin, D. V., P. Broche, J.-L. Devenon, and V. I. Shrira, 2004: Validation of HF radar probing of the
vertical shear of surface currents by acoustic Doppler current profiler measurements. J. Geophys. Res.,
101, C04003. doi:10.1029/2003JC002025.
Jackson, F. C., W. T. Walton, and P. L. Baker, 1985: Aircraft and satellite measurement of ocean wave
directional spectra using scanning-beam microwave radars. J. Geophys. Res., 90, 987–1004.
Janssen, P., 2004: The interaction of ocean waves and wind . Cambridge University Press, Cambridge,
300 pp.
Janssen, P. A. E. M., 1991: Quasi-linear theory of wind wave generation applied to wave forecasting. J.
Phys. Oceanogr., 21, 1631–1642, see comments by D. Chalikov, J. Phys. Oceanogr. 1993, vol. 23 pp.
1597–1600.URL link.
— 2008: Progress in ocean wave forecasting. J. Comp. Phys., 227, 3572–3594.
doi:10.1016/j.jcp.2007.04.029.
— 2009: On some consequences of the canonical transformation in the Hamiltonian theory of water
waves. J. Fluid Mech., 637, 1–44.
— 2011: Ocean wave effects on the daily cycle in SST. J. Geophys. Res., 117, C00J32.
doi:10.1029/2012JC007943.
Janssen, P. A. E. M., K. Hasselmann, S. Hasselmann, and G. J. Komen, 1994: Parameterization of source
terms and the energy balance in a growing wind sea. Dynamics and modelling of ocean waves, G. J.
212 BIBLIOGRAPHY

Komen et al., ed., Cambridge University Press, pages 215–238.

Janssen, T. and J. Battjes, 2006: A note on wave energy dissipation over steep beaches. Coastal Eng.,
54, 711–716.
Jaramillo, S., A. Sheremet, M. A. Allison, A. H. Reed, and K. T. Holland, 2009: Wave-mud interactions
over the muddy atchafalaya subaqueous clinoform, louisiana, united states: Wave-supported sediment
transport. J. Geophys. Res., 114, C04002. doi:10.1029/2008JC004821.
Jensen, B. L., B. M. Sumer, and J. Fredsøe, 1989: Turbulent oscillatory boundary layers at high Reynolds
numbers. J. Fluid Mech., 206, 265–297.
Jessup, A. T. and K. R. Phadnis, 2005: Measurement of the geometric and kinematic properties of
microscale breaking waves from infrared imagery using a PIV algorithm. Meas. Sci. Technol., 16,
1961–1969. doi:10.1088/0957-0233/16/10/011.
Jonsson, I. G., 1967: Proceedings of the 10th International Conference on Coastal Engineering, Tokyo,
Japan, ASCE, 127–148.
Kahma, K. K., 1981: A study of the growth of the wave spectrum with fetch. J. Phys. Oceanogr., 11,
1503–1515.
Kahma, K. K. and C. J. Calkoen, 1992: Reconciling discrepancies in the observed growth of wind-
generated waves. J. Phys. Oceanogr., 22, 1389–1405.URL link.
Kahma, K. K. and M. A. Donelan, 1988: A laboratory study of the minimum wind speed for wind wave
generation. J. Fluid Mech., 192, 339–364.
Kajiura, K., 1968: A model of the bottom boundary layer in water waves. Bull. Earthquake Res. Inst.
Univ. Tokyo, 46, 75–123.
Keller, J. B., 1998: Gravity waves on ice-covered water. J. Geophys. Res., 103, 7663–7669.
Kenyon, K. E., 1969: Stokes drift for random gravity waves. J. Geophys. Res., 74, 6991–6994.

© ©
Kerbaol, V., 1997: Analyse spectrale et statistique vent-vagues des images radar à ouverture
synthà tique : application aux donnà es satellites ERS-1/2 . Ph.D. thesis, Université de Rennes 1.
Kinsman, B., 1965: Wind waves. Prentice-Hall, Englewood Cliffs, N. J., 676 p. Reprinted by Dover
Phoenix editions, Mineola, N. Y.
Kirby, J. T., 1986: Higher-order approximations in the parabolic equation method for water waves. J.
Geophys. Res., 91, 933–952.
Kirby, J. T. and T.-M. Chen, 1989: Surface waves on vertically sheared flows: approximate dispersion
relations. J. Geophys. Res., 94, 1013–1027.
Kirincich, A., 2016: Remote sensing of the surface wind field over the coastal ocean via direct calibration
of HF radar backscatter power. J. Atmos. Ocean Technol., 33, 1377–1392. doi:10.1175//JTECH-D-
15-0242.1.
Kitaigorodskii, S. A., 1962: Applications of the theory of similarity to the analysis of wind-generated
wave motion as a stochastic process. Izv. Geophys. Ser. Acad. Sci., USSR, 1, 105–117.
— 1994: A note on the influence of breaking wind waves on the aerodynamic roughness of the sea surface
as seen from below. Tellus, 46A, 681–685.
Kobayashi, S., Y. Ota, Y. Harada, A. Ebita, M. Moriya, H. Onoda, K. Onogi, H. Kamahori, C. Kobayashi,
H. Endo, K. Miyaoka, and K. Takahashi, 2015: The JRA-55 reanalysis: General specifications and
basic characteristics. J. Met. Soc. Japan, 93, 5–48. doi:10.2151/jmsj.2015-001.
Koga, M., 1982: Bubble entrainment in breaking wind waves. Tellus, 34, 481–489.
Komar, P. D., 1998: Beach processes and sedimentation. Prentice-Hall, second edition, 544 p.
Komatsu, K. and A. Masuda, 1996: A new scheme of nonlinear energy transfer among wind waves: Riam
method. algorithm and performance. Journal of Oceanography, 52, 509–537.URL link.
Komen, G. J., L. Cavaleri, M. Donelan, K. Hasselmann, S. Hasselmann, and P. A. E. M. Janssen, 1994:
Dynamics and modelling of ocean waves. Cambridge University Press, Cambridge, 554 pp.
Komen, G. J., K. Hasselmann, and S. Hasselmann, 1984: On the existence of a fully developed windsea
spectrum. J. Phys. Oceanogr., 14, 1271–1285.URL link.
Korotkevich, A. O., A. Pushkarev, D. Resiod, and V. E. Zakharov, 2008: Numerical verification of the
weak turbulent model for swell evolution. Eur. J. Mech. B/Fluids, 27, 631–387.
Korteweg, D. J. and G. de Vries, 1895: On the change of form of long waves advancing in a rectangular
canal and on a new type of long stationary waves. Phil. Mag., 39, 422–443.
Krogstad, H. E., 1992: A simple derivation of Hasselmann’s nonlinear ocean-synthetic aperture radar
transform. J. Geophys. Res., 97, 2421–2425.
Kudryavtsev, V., D. Hauser, G. Caudal, and B. Chapron, 2003: A semiempirical model of the nor-
malized radar cross-section of the sea surface 1. background model. J. Geophys. Res., 108, 8054.
doi:10.1029/2001JCOO1003.
BIBLIOGRAPHY 213

Kudryavtsev, V., A. Myasoedov, B. Chapron, J. A. Johannessen, and F. Collard, 2012: Imaging mesoscale
upper ocean dynamics using synthetic aperture radar and optical data. J. Geophys. Res., 117, C04029.
doi:10.1029/2011JC007492.
Kudryavtsev, V., M. Yurovskaya, B. Chapron, F. Collard, and C. Donlon, 2017: Sun glitter im-
agery of surface waves. part 1: Directional spectrum retrieval and validation. J. Geophys. Res., 122.
doi:10.1002/2016JC012425.
Kudryavtsev, V. N., 1994: The coupling of wind and internal waves. J. Fluid Mech., 278, 33–62.
Kudryavtsev, V. N., V. K. Makin, and B. Chapron, 1999: Coupled sea surface–atmosphere model. 2.
spectrum of short wind waves. J. Geophys. Res., 104, 7625–7639.
Kuik, A. J., G. P. van Vledder, and L. H. Holthuijsen, 1988: A method for the routine analysis of
pitch-and-roll buoy wave data. J. Phys. Oceanogr., 18, 1020–1034.URL link.
Lacombe, H., 1950: Les fluctuations des caractéristiques de la houle par petites profondeurs. les ”plans
de vagues”. Bulletin d’information du Comité d’Océanographie et d’Etude des Côtes, 180–187.
Lamarre, E. and W. K. Melville, 1991: Air entrainment and dissipation in breaking waves. Nature, 351,
469–472.
Landau, L. D. and E. M. Lifshitz, 1960: Mechanics. Pergamon Press Addison-Wesley, Reading, MA,
524–526 pp.
Langmuir, I., 1938: Surface motion of water induced by wind. Science, 87, 119–123.
Lannes, D., 2013: The water waves problem. American Mathematical Society.
Lannes, D. and P. Bonneton, 2009: Derivation of asymptotic two-dimensional time-dependent equations
for surface water wave propagation. Phys. of Fluids, 21, 16601.
Large, W. G., J. C. McWilliams, and S. C. Doney, 1994: Oceanic vertical mixing: a review and a model
with nonlocal boundary layer parameterization. Rev. of Geophys., 32, 363–403.
Lavrenov, I. V., 1986: Behavior of the surface gravity wave spectrum on a horizontally nonuniform
current. Izv. Atmos. Ocean. Phys., 22, 398–401.
— 2003: Wind-waves in oceans: dynamics and numerical simulations. Springer, Berlin, 376 pp.
Laxague, N. J. M., B. K. Haus, D. Bogucki, and T. Özgökmen, 2015: Spectral characterization of
fine-scale wind waves using shipboard optical polarimetry. J. Geophys. Res., 120, 3140–3156.
Laxague, N. J. M., T. M. Özgökmen, B. K. Haus, G. Novelli, A. Shcherbina, P. Sutherland, C. M. G. B.
Lund, S. Mehta, M. Alday, and J. Molemaker, 2018: Observations of near-surface current shear help de-
scribe oceanic oil and plastic transport. Geophys. Res. Lett., 44, 245–249. doi:10.1002/2017GL075891.
Leckler, F., F. Ardhuin, C. Peureux, A. Benetazzo, F. Bergamasco, and V. Dulov, 2015: Analysis and
interpretation of frequency-wavenumber spectra of young wind waves. J. Phys. Oceanogr., 45, 2484–
2496. doi:10.1175/JPO-D-14-0237.1.
Lentz, S. J., M. F. P. Howd, J. Fredericks, and K. Hathaway, 2008: Observations and a model
of undertow over the inner continental shelf. J. Phys. Oceanogr., 38, 2341–2357.URL link.
doi:10.1175/2008JPO3986.1.
Leonard, B. P., 1991: The ULTIMATE conservative difference scheme applied to unsteady one-
dimensional advection. Computational Methods in Applied Mechanical Engineering, 88, 17–74.
Levi-Civita, T., 1925: Détermination rigoureuse des ondes permanentes d’ampleur finie. Matematische
Annalen, XCII, 264–314.
Li, J.-G., 2010: Global transport on a spherical multiple-cell grid. Mon. Weather Rev., 139, 1536–1555.
doi:http://dx.doi.org/10.1175/2010MWR3196.1.
Li, M. Z. and C. L. Amos, 1999: Sheet flow and large wave ripples under combined waves and cur-
rents: field observations, model predictions and effect on boundary layer dynamics. Continental Shelf
Research, 19, 637–663.
Li, Q. and B. Fox-Kemper, 2017: Assessing the effects of Langmuir turbulence on the entrainment buoy-
ancy flux in the ocean surface boundary layer. J. Phys. Oceanogr., 47, 2863–2886. doi:10.1175/JPO-
D-17-0085.1.
Li, X.-M., S. Lehner, and T. Bruns, 2011: Ocean wave integral parameter measurements us-
ing envisat asar wave mode data. IEEE Trans. on Geosci. and Remote Sensing, 49, 155–174.
doi:10.1109/TGRS.2010.2052364.
Lippmann, T. C., A. H. Brookins, and E. B. Thornton, 1996: Wave energy transformation on natural
profiles. Coastal Eng., 27, 1–20. doi:https://doi.org/10.1016/0378-3839(95)00036-4.
Liu, A. K. and E. Mollo-Christensen, 1988: Wave propagation in a solid ice pack. J. Phys. Oceanogr.,
18, 1702–1712.
Liu, P. C. and A. V. Babanin, 2004: Using wavelet spectrum analysis to resolve breaking events in the
wind wave time series. Annales Geophysicae, 22, 3335–3345.
214 BIBLIOGRAPHY

Liu, P. C., H. S. Chen, D.-J. Doong, C. C. Kao, and Y.-J. G. Hsu, 2008: Monstrous ocean waves during
typhoon Krosa. Annales Geophysicae, 26, 1327–1329.
Liu, Y. and D. K. P. Yue, 1998: On generalized Bragg scattering of surface waves by bottom ripples. J.
Fluid Mech., 356, 297–326.
Long, C. E. and D. T. Resio, 2007: Wind wave spectral observations in Currituck Sound, North Carolina.
J. Geophys. Res., 112, C05001. doi:10.1029/2006JC003835.
Long, R. B. and K. Hasselmann, 1979: A variational technique for extracting directional spectra from
multi-component wave data. J. Phys. Oceanogr., 9, 373–381.
Longuet-Higgins, M., 1976: On the nonlinear transfer of energy in the peak of a gravity-wave spectrum:
A simplified model. Proc. Roy. Soc. Lond. A, 347, 311–328.
Longuet-Higgins, M. S., 1950: A theory of the origin of microseisms. Phil. Trans. Roy. Soc. London A,
243, 1–35.
— 1952: On the statistical distributions of sea waves. J. Mar. Res., 11, 245–265.
— 1953: Mass transport under water waves. Phil. Trans. Roy. Soc. London A, 245, 535–581.
— 1960: Reflection and refraction at a random moving surface. II. number of specular points in a gaussian
surface. J. Opt. Soc. Am., 50, 845–850. doi:10.1364/JOSA.50.000845.
— 1970: Longshore currents generated by obliquely incident sea waves, 1. J. Geophys. Res., 75, 6778–
6789.
— 1984: New integral relations for gravity waves of finite amplitude. J. Fluid Mech., 149, 205–215, see
also Yu and Wu, J. Fluid Mech., 1987.
— 2005: On wave set-up in shoaling water with a rough sea bed. J. Fluid Mech., 527, 217–234.
Longuet-Higgins, M. S., D. E. Cartwright, and N. D. Smith, 1963: Observations of the directional
spectrum of sea waves using the motions of a floating buoy. Ocean Wave Spectra, proceedings of a
conference, Easton, Maryland , National Academy of Sciences, Prentice-Hall, 111–136.
Longuet-Higgins, M. S. and R. W. Stewart, 1962: Radiation stresses and mass transport in surface
gravity waves with application to ‘surf beats’. J. Fluid Mech., 13, 481–504.
— 1964: Radiation stress in water waves, a physical discussion with applications. Deep Sea Research,
11, 529–563.
Lowe, R. J., J. L. Falter, J. R. Koseff, S. G. Monismith, and M. J. Atkinson, 2007: Spectral wave flow
attenuation within submerged canopies: Implications for wave energy dissipation. J. Geophys. Res.,
112, C05018. doi:10.1029/2006JC003605.
Lu, N. Q., A. Prosperetti, and S. W. Yoon, 1990: Underwater noise emissions from bubble clouds. IEEE
J. Oceanic Eng., 15, 275–281.
Lund, B., H. C. Graber, H. Tamura, C. O. Collins, and S. M. Varlamov, 2015: A new technique for the
retrieval of near-surface vertical current shear from marine x-band radar images. J. Geophys. Res.,
120, 8466–8486. doi:10.1002/2015JC010961.
Lygre, A. and H. E. Krogstad, 1986: Maximum entropy estimation of the directional distribution in
ocean wave spectra. J. Phys. Oceanogr., 16, 2,052–2,060.
Ma, B. B., J. A. Nystuen, and R.-C. Lien, 2005: Prediction of underwater sound levels from rain and
wind. J. Acoust. Soc. Amer., 117, 3555–3565.
Madsen, O. S., 1994: Spectral wave-current bottom boundary layer flows. Proceedings of the 24th inter-
national conference on coastal engineering, ASCE, 384–397.
Madsen, O. S., P. P. Mathisen, and M. M. Rosengaus, 1990: Movable bed friction factors for spectral
waves. Proceedings of the 22nd international conference on coastal engineering, ASCE, 420–429.
Magne, R., K. Belibassakis, T. H. C. Herbers, F. Ardhuin, W. C. O’Reilly, and V. Rey, 2007:
Evolution of surface gravity waves over a submarine canyon. J. Geophys. Res., 112, C01002.
doi:10.1029/2005JC003035.
Manasseh, R., A. V. Babanin, C. Forbes, K. Rickards, I. Bobevski, and A. Ooi, 2006: Passive acous-
tic determination of wave-breaking events and their severity across the spectrum. J. Atmos. Ocean
Technol., 23, 599–618.
Marechal, G. and F. Ardhuin, 2021: Surface currents and significant wave height gradients: matching
numerical models and high-resolution altimeter wave heights in the agulhas current region. J. Geophys.
Res., 126, e2020JC016564. doi:10.1029/2020JC016564.
Marié, L., F. Collard, F. Nouguier, L. Pineau-Guillou, D. Hauser, F. Boy, S. Méric, P. Sutherland,
C. Peureux, G. Monnier, B. Chapron, A. Martin, P. Dubois, C. Donlon, T. Casal, and F. Ardhuin,
2020: Measuring ocean surface velocities with the kuros airborne near-nadir doppler radar: a multi-
scale analysis in preparation of the skim mission. Ocean Sci., 16, 1399–1429. doi:10.5194/os-16-1399-
2020.
BIBLIOGRAPHY 215

Marieu, V., P. Bonneton, D. L. Foster, and F. Ardhuin, 2008: Modeling of vortex ripple morphodynamics.
J. Geophys. Res., 113, C09007. doi:10.1029/2007JC004659.
Marin, F., 2004: Eddy viscosity and Eulerian drift over rippled beds in waves. Coastal Eng., 50, 139–159.
Marmorino, G. O., G. B. Smith, and G. J. Lindemann, 2005: Infrared imagery of large-aspect-ratio
Langmuir circulation. Continental Shelf Research, 25, 1–6.
Marsan, D., J. Weiss, E. Larose, and J.-P. Métaxian, 2012: Sea-ice thickness measurement based on the
dispersion of ice swell. J. Acoust. Soc. Amer., 131, 80–91. doi:10.1121/1.3662051.
Martin, S. and P. Kauffman, 1981: A field and laboratory study of wave dampin by grease ice. Journal
of Geology, 27, 283–313.
Massel, S. R., 1993: Extended refraction-diffraction equation for surface waves. Coastal Eng., 19, 97–126.
Masson, D. and P. Chandler, 1993: Wave groups: a closer look at spectral methods. Coastal Eng., 20,
249–275.
Mastenbroek, C., 1996: Wind-wave interaction. Ph.D. thesis, Delft University of Technology, The Nether-
lands.
Mathiesen, M., 1987: Wave refraction by a current whirl. J. Geophys. Res., 92, 3905–3912.
Mathisen, P. P. and O. S. Madsen, 1996: Wave and currents over a fixed rippled bed. 2. bottom and
apparent roughness experienced by currents in the presence of waves. J. Geophys. Res., 101, 16,543–
16,550, see erratum in Mathisen and Madsen JGR 1999.
McFarlane, V., M. Loewen, and F. Hicks, 1982: Measurements of the evolution of frazil ice particle size
distributions. Cold Regions Sci. Tech., 120, 45–55.
McGoldrick, L. F., O. M. Phillips, N. E. Huang, and T. H. Hodgson, 1966: Measurements of third-order
wave interactions. J. Fluid Mech., 25, 437–456.
Mei, C. C., 1989: Applied dynamics of ocean surface waves. World Scientific, Singapore, second edition,
740 p.
Mellor, G., 2002: Oscillatory bottom boundary layers. J. Phys. Oceanogr., 32, 3075–3088.
Mellor, G. and A. Blumberg, 2004: Wave breaking and ocean surface layer thermal response. J. Phys.
Oceanogr., 34, 693–698.
Mellor, G. L. and T. Yamada, 1982: Development of a turbulence closure model for geophysical fluid
problems. Rev. Geophys. Space Phys., 20, 851–875.
Melville, W. K. and P. Matusov, 2002: Distribution of breaking waves at the ocean surface. Nature, 417,
58–63.
Melville, W. K. and R. J. Rapp, 1988: The surface velocity in steep and breaking waves. J. Fluid Mech.,
169, 1–22.
Melville, W. K., R. Shear, and F. Verron, 1998: Laboratory measurements of the generation and evolution
of Langmuir circulations. J. Fluid Mech., 364, 31–58.
Meylan, M. H., L. G. Bennetts, J. E. M. Mosig, W. E. Rogers, M. J. Doble, and M. A. Peter, 2018:
Dispersion relations, power laws, and energy loss for waves in the marginal ice zone. J. Phys. Oceanogr.,
123, 3322–3335. doi:10.1002/2018JC013776.
Miche, A., 1944a: Mouvements ondulatoires de la mer en profondeur croissante ou décroissante. forme
limite de la houle lors de son déferlement. application aux digues maritimes. Troisième partie. Forme
et propriétés des houles limites lors du déferlement. Croissance des vitesses vers la rive. Annales des
Ponts et Chaussées, Tome 114, 369–406.
— 1944b: Mouvements ondulatoires de la mer en profondeur croissante ou décroissante. Première partie.
Mouvements ondulatoires périodiques et cylindriques en profondeur constante. Annales des Ponts et
Chaussées, Tome 114, 42–78.
— 1951: Le pouvoir réfléchissant des ouvrages maritimes exposés à l’action de la houle. Annales des
Ponts et Chaussées, 121, 285–319.
Miles, J., 1996: Surface wave generation: a viscoelastic model. J. Fluid Mech., 322, 131–145.
Miles, J. W., 1959: On the generation of surface waves by shear flows. part 2. J. Fluid Mech., 6, 568–598.
Minster, J. F., D. Jourdan, C. Boissier, and P. Midol-Monnet, 1991: Estimation of the sea-state bias
in radar altimeter Geosat data from examination of frontal systems. J. Atmos. Ocean Technol., 9,
174–187.
Mitsuyasu, H., F. Tasai, T. Suhara, S. Mizuno, M. Onkusu, T. Honda, and T. Rukiiski, 1975: Ob-
servations of the directional spectrum of ocean waves using a cloverleaf buoy. J. Phys. Oceanogr., 5,
751–761.
Monahan, E. C. and D. K. Woolf, 1989: Comments on ”variations of whitecap coverage with wind stress
and water temperature”. J. Phys. Oceanogr., 19, 706–709.URL link.
Monin, A. S., 1962: Empirical data on turbulence in the surface layer of the atmosphere. J. Geophys.
216 BIBLIOGRAPHY

Res., 67, 3103–3109.

Monismith, S. G., 2007: Hydrodynamics of coral reefs. Annu. Rev. Fluid Mech., 39, 37–55.
doi:10.1146/annurev.fluid.38.050304.092125.
Monismith, S. G., J. S. Rogers, D. Koweek, and R. B. Dunbar, 2015: Frictional wave dissipation on a
remarkably rough reef. Geophys. Res. Lett., 112, 4063–4071. doi:10.1002/2015GL063804.
Montagne, R., 1922: Le service de prédiction de la houle au Maroc. Annales Hydrographiques, 157–186.
Mooney, M., 1951: The viscosity of a concentrated suspension of spherical particles. Journal of Colloid
Science, 6, 162–170.
Mosig, J. E. M., F. Montiel, and V. A. Squire, 2015: Comparison of viscoelastic-type models for ocean
wave attenuation in ice-covered seas. J. Geophys. Res., 120, 6072–6090. doi:10.1002/2015JC010881.
Mouche, A. and B. Chapron, 2015: Global C-band Envisat, RADARSAT-2 and Sentinel-1
SAR measurements in copolarization and cross-polarization. J. Geophys. Res., 120, 7195–7207.
doi:10.1002/2015JC011149.
Mouche, A., F. Collard, B. Chapron, K.-F. Dagestad, G. Guitton, J. A. Johannessen, V. Kerbaol, and
M. W. Hansen, 2012: On the use of doppler shift for sea surface wind retrieval from SAR. IEEE Trans.
on Geosci. and Remote Sensing, 50, 2901–2909. doi:10.1109/TGRS.2011.2174998.
Munk, W., 2009: An inconvenient sea truth: Spread, steepness, and skewness of surface slopes. Annu.
Rev. Mar. Sci., 1, 377–415. doi:10.1146/annurev.marine.010908.163940.
Munk, W. H., 1949: Surf beat. Eos Trans. AGU , 30, 849–854.
— 1950: Origin and generation of waves. Proceedings 1st International Conference on Coastal Engineer-
ing, Long Beach, California, ASCE, 1–4.
Munk, W. H., G. R. Miller, F. E. Snodgrass, and N. F. Barber, 1963: Directional recording of swell from
distant storms. Phil. Trans. Roy. Soc. London A, 255, 505–584.
Munk, W. H. and M. A. Traylor, 1947: Refraction of ocean waves: a process linking underwater topog-
raphy to beach erosion. Journal of Geology, LV, 1–26.
Nadaoka, K., S. Beji, and Y. Nakagawa, 1997: A fully dispersive weakly nonlinear model for water waves.
Proc. Roy. Soc. Lond. A, 453, 303–318.
Nielsen, P., 1992: Coastal bottom boudary layers and sediment transport. World Scientific Publishing.
Nikuradse, J., 1933: Strömungsgestze in rauhen rohren. Technical Report 361, VDI, (English translation:
NACA Tech. Memo. 1292, National Advisory Commission for Aeronautics, Washington D.C., 1950).
Noh, Y., 1996: Dynamics of diurnal thermocline formation in the oceanic mixed layer. J. Geophys. Res.,
26, 2189–2195.
Noh, Y. and H. J. Kim, 1999: Simulations of temperature and turbulence structure of the oceanic
boundary layer with the improved near-surface process. J. Geophys. Res., 104, 15621–15634.
Noh, Y., H. S. Min, and S. Raasch, 2004: Large eddy simulation of the ocean mixed layer: the effects of
wave breaking and Langmuir circulation. J. Phys. Oceanogr., 34, 720–733.
Nouguier, F., B. Chapron, F. Collard, A. Mouche, N. Rascle, F. Ardhuin, and X. Wu, 2018: Sea surface
kinematics from near-nadir radar measurements. IEEE Trans. on Geosci. and Remote Sensing, 56,
6169–6179.URL link. doi:10.1109/TGRS.2018.2833200.
Nouguier, F., C.-A. Guérin, and B. Chapron, 2009: “choppy wave” model for nonlinear gravity waves.
J. Geophys. Res., 114, C09012. doi:10.1029/2008JC004984.
Nouguier, F., A. Mouche, N. Rascle, B. Chapron, and D. Vandemark, 2016: Analysis of dual-frequency
ocean backscatter measurements at Ku- and Ka-bands using near-nadir incidence GPM radar data.
IEEE Geoscience And Remote Sensing Letters, 31, 2023–2245. doi:10.1109/LGRS.2016.2583198.
Nwogu, O., 1993: Alternative form of Boussinesq equations for nearshore wave propagation. J. of Wa-
terway, Port Coast. Ocean Eng., 119, 618–637.
Oltman-Shay, J., P. A. Howd, and W. A. Birkemeier, 1989: Shear instabilities of the mean longshore
current, 2, field observations. J. Geophys. Res., 94, 18031–18042.
Onink, V., D. Wichmann, P. Delandmeter, and E. van Sebille, 2019: The role of Ekman currents,
geostrophy and Stokes drift in the accumulation of floating microplastic . J. Geophys. Res., 124,
1474–1490. doi:10.1029/2018JC014547.
O’Reilly, W. C. and R. T. Guza, 1991: Comparison of spectral refraction and refraction-diffraction wave
models. J. of Waterway, Port Coast. Ocean Eng., 117, 199–215.
— 1993: A comparison of two spectral wave models in the Southern California Bight. Coastal Eng., 19,
263–282.
— 1998: Assimilating coastal wave observations in regional swell predictions. part i: inverse methods. J.
Phys. Oceanogr., 28, 679–691.
O’Reilly, W. C., T. H. C. Herbers, R. J. Seymour, and R. T. Guza, 1996: A comparison of directional
BIBLIOGRAPHY 217

buoy and fixed platform measurements of Pacific swell. J. Atmos. Ocean Technol., 13, 231–238.
Osborne, A. R., L. Bergamasco, M. Serio, L. Bianco, L. Cavaleri, M. Drago, L. Iovenitti, and D. Viez-
zoli, 1996: Nonlinear shoaling of shallow water waves: perspective in term of the inverse scattering
transform. Nuovo Cimento, 19C, 151–176.
Osborne, A. R. and T. L. Burch, 1980: Coupling between a surface-wave spectrum and an internal wave:
modulation interaction. Science, 208, 513–460.
Passaro, M., P. Cipollini, S. Vignudelli, G. D. Quartly, and H. M. Snaith, 2014: ALES: A multi-mission
adaptive subwaveform retracker for coastal and open ocean altimetry. Remote sensing of Environment,
145, 173–189.
Passaro, M., L. Fenoglio-Marc, and P. Cipollini, 2015: Validation of significant wave height from improved
satellite altimetry in the German Bight. IEEE Trans. on Geosci. and Remote Sensing, 53, 2146–2156.
Patankar, S., 1980: Numerical heat transfer and fluid flow: Computational methods in mechanics and
thermal science. Hemisphere Publishing Corp., Washington, DC, 210 pp.
Pawka, S. S., 1983: Island shadows in wave directional spectra. J. Geophys. Res., 88, 2579–2591.
Peak, S. D., 2004: Wave refraction over complex nearshore bathymetry. Master’s thesis, Naval Postgrad-
uate School.URL link.
Peral, E., E. Rodriguez, and D. Esteban-Fernandez, 2015: Impact of surface waves on SWOT’s projected
ocean accuracy. Remote Sensing, 7, 14509–14529. doi:10.3390/rs71114509.
Peregrine, D. H., 1967: Long waves on a beach. J. Fluid Mech., 27, 815–827.
Perignon, Y., F. Ardhuin, M. Cathelain, and M. Robert, 2014: Swell dissipation by induced atmospheric
shear stress. J. Geophys. Res., 119, 6622–6630. doi:10.1002/2014JC009896.
Pettersson, H., H. C. Graber, D. Hauser, C. Quentin, K. Kahma, W. M. Drennan, and M. A. Donelan,
2003: Directional wave measurements from three wave sensors during the FETCH experiment. J.
Geophys. Res., 108, 8061. doi:10.1029/2001JC001164.
Peureux, C. and F. Ardhuin, 2016: Ocean bottom pressure records from the cascadia array and short
surface gravity waves. J. Geophys. Res., 121, 2862–2873. doi:10.1002/2015JC011580.
Peureux, C., F. Ardhuin, and P. V. Guimaraes, 2021: On the unsteady steepening of short gravity waves
near the crests of longer waves in the absence of generation or dissipation. J. Geophys. Res., 126.
doi:10.1029/2020JC016735.
Peureux, C., A. Benetazzo, and F. Ardhuin, 2018: Note on the directional properties of meter-scale
gravity waves. Ocean Science, 14, 41–52. doi:10.5194/os-14-41-2018.
Phillips, O. M., 1957: On the generation of waves by turbulent wind. J. Fluid Mech., 2, 415–417.
— 1958: The equilibrium range in the spectrum of wind-generated waves. J. Fluid Mech., 4, 426–433.
— 1960: On the dynamics of unsteady gravity waves of finite amplitude. J. Fluid Mech., 9, 193–217.
— 1977: The dynamics of the upper ocean. Cambridge University Press, London, 336 p.
— 1984: On the response of short ocean wave components at a fixed wavenumber to
ocean current variations. J. Phys. Oceanogr., 14, 1425–1433.URL link. doi:10.1175/1520-
0485(1984)014¡1425:OTROSO¿2.0.CO;2.
— 1985: Spectral and statistical properties of the equilibrium range in wind-generated gravity waves. J.
Fluid Mech., 156, 505–531.
Pierson, W. J., 1972: The loss of two British trawlers – a study in wave refraction. Journal of Navigation,
25, 291–304.
Pierson, W. J., G. Neumann, and R. W. James, 1955: Practical methods for observing and forecasting
ocean waves by means of wave spectra and statistics. U. S. Hydrographic Office, 284 pp.
Pierson, W. J., Jr and L. Moskowitz, 1964: A proposed spectral form for fully developed wind seas based
on the similarity theory of S. A. Kitaigorodskii. J. Geophys. Res., 69, 5,181–5,190.
Pineau-Guillou, L., F. Ardhuin, M.-N. Bouin, J.-L. Redelsperger, B. Chapron, J. Bidlot, and Y. Quil-
fen, 2018: Strong winds in a coupled wave-atmosphere model during a north Atlantic storm event:
evaluation against observations. Quart. Journ. Roy. Meteorol. Soc., 144, 317–332. doi:10.1002/qj.3205.
Plant, W. J., 1982: A relationship between wind stress and wave slope. J. Geophys. Res., 87, 1961–1967.
Poate, T. G., R. T. McCall, and G. Masselink, 2016: A new parameterisation for runup on gravel beaches.
Coastal Eng., 117, 176–190.
Popinet, S., R. M. Gorman, G. J. Rickard, and H. L. Tolman, 2010: A quadtree-adaptive spectral wave
model. Ocean Modelling, 34, 36–49.
Potter, H., C. O. Collins, III, W. M. Drennan, and H. C. Graber, 2015: Observations of wind stress di-
rection during typhoon Chaba (2010). Geophys. Res. Lett., 42, 9898–9905. doi:10.1029/2011JC007833.
Pullen, T., N. W. H. Allsop, T. Bruce, A. Kortenhaus, H. Schüttrumpf, and J. W. van der Meer, 2007:
Wave overtopping of sea defences and related structures: assessment manual . Boyens Medien GmbH,
218 BIBLIOGRAPHY

336 p.URL link.

Putrevu, U. and I. A. Svendsen, 1999: Three-dimensional dispersion of momentum in wave-induced
nearshore currents. Eur. J. Mech. B/Fluids, 18, 410–426.
Quartly, G. D., M. A. Srokosz, and A. C. McMillan, 2001: Analyzing altimeter artifacts: Statis-
tical properties of ocean waveforms. J. Atmos. Ocean Technol., 18, 2074–2091. doi:10.1175/1520-
0426(2001)018¡2074:AAASPO¿2.0.CO;2.
Queffeulou, P. and D. Croizé-Fillon, 2010: Global altimeter SWH data set, version 7, may 2010. Technical
report, Ifremer, uRL:http://tinyurl.com/2cj5sez.URL link.
Quilfen, Y. and B. Chapron, 2019: Ocean surface wave-current signatures from satellite altimeter mea-
surements. Geophys. Res. Lett., 216, 253–261. doi:10.1029/2018GL081029.
Quilfen, Y., B. Chapron, and J. Tournadre, 2010: Satellite microwave surface observations in tropical
cyclones. Mon. Weather Rev., 131, 421–437.
Quilfen, Y., M. Yurovskaya, B. Chapron, and F. Ardhuin, 2018: Storm waves sharpening in the Ag-
ulhas current: satellite observations and modeling. Remote sensing of Environment, 216, 561–571.
doi:10.1016/j.rse.2018.07.020.
Radder, A. C., 1979: On the parabolic equation method for water wave propagation. J. Fluid Mech., 95,
159–176.
Rascle, N. and F. Ardhuin, 2009: Drift and mixing under the ocean surface revisited. stratified conditions
and model-data comparisons. J. Geophys. Res., 114, C02016. doi:10.1029/2007JC004466.
— 2013: A global wave parameter database for geophysical applications. part 2: model
validation with improved source term parameterization. Ocean Modelling, 70, 174–188.
doi:10.1016/j.ocemod.2012.12.001.
Rascle, N., F. Ardhuin, and E. A. Terray, 2006: Drift and mixing under the ocean surface. a coherent
one-dimensional description with application to unstratified conditions. J. Geophys. Res., 111, C03016.
doi:10.1029/2005JC003004.
Rascle, N., B. Chapron, F. Ardhuin, and A. Soloviev, 2013: A note on the direct injection of turbulence
by breaking waves. Ocean Modelling, 70, 145–151. doi:10.1016/j.ocemod.2012.09.001.
Rascle, N., B. Chapron, A. Ponte, F. Ardhuin, and P. Klein, 2014: Surface roughness imaging of
currents shows divergence and strain in the wind direction. J. Phys. Oceanogr., 44, 2153–2163.
doi:10.1175/JPO-D-13-0278.1.
Rascle, N., J. Molemaker, L. Marié, F. Nouguier, B. Chapron, B. Lund, and A. Mouche, 2017: Intense
deformation field at oceanic front inferred from directional sea surface roughness observations. Geophys.
Res. Lett., 48, 5599–5608. doi:10.1002/2017GL073473.
Rayleigh, L., 1896: The Theory of Sound . Macmillan, London, 3rd edition.
Renault, L., M. J. Molemaker, J. Gula, S. Masson, and J. C. McWilliams, 2016a: Control and stabiliza-
tion of the Gulf Stream by oceanic current interaction with the atmosphere. J. Phys. Oceanogr., 46,
3439–3453. doi:10.1175/JPO-D-16-0115.1.
Renault, L., M. J. Molemaker, J. C. McWilliams, A. F. Shchepetkin, F. Lemarié, D. Chelton, S. Illig,
and A. Hall, 2016b: Modulation of wind work by oceanic current interaction with the atmosphere. J.
Phys. Oceanogr., 46, 1685–1704. doi:10.1175/JPO-D-15-0232.1.
Reniers, A. J. H. M. and J. A. Battjes, 1997: A laboratory study of longshore currents over barred and
non-barred beaches. Coastal Eng., 30, 1–22.
Reniers, A. J. H. M., J. A. Roelvink, and E. B. Thornton, 2004: Morphodynamic modeling of an embayed
beach under wave group forcing. J. Geophys. Res., 109, C01030. doi:10.1029/2002JC001586.
Resio, D. T., J. H. Pihl, B. A. Tracy, and C. L. Vincent, 2001: Nonlinear energy fluxes and the finite
depth equilibrium range in wave spectra. J. Geophys. Res., 106, 6985–7000.
Resio, D. T. and J. J. Westerink, 2008: Modeling the physics of storm surges. Physics Today, 33–38.
Reul, N. and B. Chapron, 2003: A model of sea-foam thickness distribution for passive microwave remote
sensing applications. J. Geophys. Res., 108, 3321. doi:10.1029/2003JC001887.
Reul, N., J. Tenerelli, B. Chapron, D. Vandemark, Y. Quilfen, and Y. Kerr, 2006: Smos satellite l-band
radiometer: A new capability for ocean surface remote sensing in hurricanes. J. Geophys. Res., 117,
C02006. doi:10.1029/2011JC007474.
Ricchiuto, M., Á. Csı́k, and H. Deconinck, 2005: Residual distribution for general time-dependent con-
servation laws. J. Comp. Phys., 209, 249–289. doi:10.1016/j.jcp.2005.03.003.
Rice, S., 1944: Mathematical analysis of random noise. Noise and stochastic processes, N. Wax, ed.,
Dover Publications Inc. (published 1954), New York, 133–294.
Roach, L. A., M. M. Smith, and S. M. Dean, 2018: Quantifying growth of pancake sea ice floes using
images from drifting buoys. J. Geophys. Res., 123, 2851–2866. doi:10.1002/2017JC013693.
BIBLIOGRAPHY 219

Roelvink, D., A. Reniers, A. van Dongeren, J. v. T. de Vries, R. McCall, and J. Lescinski, 2009:
Modelling storm impacts on beaches, dunes and barrier islands. Coastal Eng., 56, 1133–1152.
doi:10.1016/j.coastaleng.2009.08.006.
Rogers, W. E., A. V. Babanin, and D. W. Wang, 2010: Observation-consistent input and whitecapping
dissipation in a model for wind-generated surface waves: Description and simple calculations. J. Atmos.
Ocean Technol., 29, 1329–1346.
Rogers, W. E., J. Thomson, H. H. Shen, M. J. Doble, P. Wadhams, and S. Cheng, 2016: Dissipa-
tion of wind waves by pancake and frazil ice in the autumn beaufort sea. J. Geophys. Res., 121.
doi:10.1002/2016JC012251.
Roland, A., 2008: Development of WWM II: Spectral wave modelling on unstructured meshes. Ph.D.
thesis, Technische Universität Darmstadt, Institute of Hydraulic and Water Resources Engineering.
Roland, A. and F. Ardhuin, 2014: On the developments of spectral wave models: numerics and param-
eterizations for the coastal ocean. Ocean Dynamics, 64, 833–846. doi:10.1007/s10236-014-0711-z.
Romero, L., 2019: Distribution of surface wave breaking fronts. Geophys. Res. Lett., 46, 10463–10474.
doi:10.1029/2019GL083408.
Romero, L. and K. W. Melville, 2010: Airborne observations of fetch-limited waves in the gulf of tehuan-
tepec. J. Phys. Oceanogr., 40, 441–465.
Ruessink, B. G., D. J. R. Walstra, and H. N. Southgate, 2003: Calibration and verification of a parametric
wave model on barred beaches. Coastal Eng., 48, 139–149.
Saha, S., S. Moorthi, H.-L. Pan, X. Wu, J. Wang, S. Nadiga, P. Tripp, R. Kistler, J. Woollen,
D. Behringer, H. Liu, D. Stokes, R. Grumbine, G. Gayno, J. Wang, Y.-T. Hou, H. ya Chuang, H.-M.
H. a. J. S. Juang, M. Iredell, R. Treadon, D. Kleist, P. V. Delst, D. Keyser, J. Derber, M. Ek, J. Meng,
H. Wei, R. Yang, S. Lord, H. van den Dool, A. Kumar, W. Wang, C. Long, M. Chelliah, Y. Xue,
B. Huang, J.-K. Schemm, W. Ebisuzaki, R. Lin, P. Xie, M. Chen, S. Zhou, W. Higgins, C.-Z. Zou,
Q. Liu, Y. Chen, Y. Han, L. Cucurull, R. W. Reynolds, G. Rutledge, and M. Goldberg, 2010: The
NCEP Climate Forecast System Reanalysis. Bull. Amer. Meterol. Soc., 91, 1015–1057.
Santala, M. J. and E. A. Terray, 1992: A technique for making unbiased estimates of current shear from
a wave-follower. Deep Sea Res., 39, 607–622.
Saulnier, J.-B., A. Clément, A. F. de O. Falcão, T. Pontes, M. Prevosto, and P. Ricci, 2011: Wave
groupiness and spectral bandwidth as relevant parameters for the performance assessment of wave
energy converters. Ocean Eng., 38, 130–147.
Schlichting, H., 1979: Boundary layer theory. McGraw-Hill, seventh edition.
Schumacher, A., 1939: Stereophotogrammetrische Wellenaufnahmen. Wiss. Ergeb. Dtsch. Atlant. Exped.
Forschungs Vermessung. Meteor 1925-1927. Technical report.
Semedo, A., Ø. Sæatra, A. Rutgersson, K. K. Kahma, and H. Pettersson, 2009: Wave-induced wind in
the marine boundary layer. J. Atmos. Sci., 66, 2256–2270.URL link.
Sénéchal, N., H. Dupuis, and P. Bonneton, 2003: Field experiment on secondary wave generation on a
barred beach and the consequent evolution of energy dissipation on the beach face. Coastal Eng., 46,
233–247.
Shen, H. H., S. F. Ackley, and M. A. Hopkins, 2001: A conceptual model for pancake ice formation in a
wave field. Annales Geophysicae, 33, 361–367. doi:10.3189/172756401781818239.
Sheremet, A., T. Staples, F. Ardhuin, S. Suanez, and B. Fichaut, 2014: Observations of large infragravity-
wave run-up at Banneg island, France. Geophys. Res. Lett., 41, 976–982. doi:10.1002/2013gl058880.
Shields, A., 1936: Anwendung der ähnlichkeits-Mechanik und der Turbulenz-forschung auf die Geschiebe-
bewegung. Preussische Versuchsanstalt für Wasserbau und Schiffbau, 26, 524–526, translated as Cal-
ifornia Inst. Technol. W. M. Keck Lab. for Hydraulic Water Res. Report 167.
Shimura, T., N. Mori, and M. A.Hemer, 2016: Projection of tropical cyclone-generated extreme wave
climate based on cmip5 multi-model ensemble in the western north pacific. Coastal Dyn., 49, 1449–
1462. doi:10.1007/s00382-016-3390-2.
Siddiqui, M. H. K. and M. R. Loewen, 2007: Characteristics of the wind drift layer and microscale
breaking waves. J. Fluid Mech., 573, 417–456. doi:10.1017/S0022112006003892.
Signell, R. P., S. Carniel, L. Cavaleri, J. Chiggiato, J. D. Doyle, J. Pullen, and M. Sclavo, 2005: Assess-
ment of wind quality for oceanographic modelling in semi-enclosed basins. J. Mar. Sys., 53, 217–233.
Sloss, P. W., 1993: Global relief cd-rom. Technical report, Marine Geology and Geophysics Division, U.
S. National Geophysical Data Center (NOAA/NESDIS/NGDC/MGGD).URL link.
Smith, J. A., 1999: Observations of wind, waves, and the mixed layer: the scaling of surface motion. The
wind-driven air-sea interface, M. L. Banner, ed., University of New South Wales, Sydney, Australia,
pages 231–238, iSBN O 7334 0586 X.
220 BIBLIOGRAPHY

— 2006: Wave-current interactions in finite-depth. J. Phys. Oceanogr., 36, 1403–1419.

Snyder, R. L. and C. S. Cox, 1966: A field study of the wind generation of ocean waves. J. Mar. Res.,
24, 141–178.
Snyder, R. L., F. W. Dobson, J. A. Elliot, and R. B. Long, 1981: Array measurement of atmospheric
pressure fluctuations above surface gravity waves. J. Fluid Mech., 102, 1–59.
Soloviev, A. V., R. Lukas, M. A. Donelan, B. K. Haus, and I. Ginis, 2014: The air-sea interface and
surface stress under tropical cyclones. Scientific Reports, 4, 5306. doi:10.1038/srep05306.
Song, J.-B. and M. L. Banner, 2002: On determining the onset and strength of breaking for deep water
waves. Part I: Unforced irrotational wave groups. J. Phys. Oceanogr., 32, 2541–2558.URL link.
Soulsby, R., 1997: Dynamics of marine sands, a manual for practical applications. Thomas Telford
Publications, London, 256 pp.
Squire, V., J. Dugan, P. Wadhams, P. Rottier, and A. Liu, 1995: Of ocean waves and sea ice. Annu.
Rev. Fluid Mech., 27, 115–168.
Squire, V. A., 2020: Ocean wave interactions with sea ice: A reappraisal. Annu. Rev. Fluid Mech., 52,
37–60. doi:10.1146/annurev-fluid-010719-060301.
Squire, V. A., G. L. Vaughan, and L. G. Bennetts, 2009: Ocean surface wave evolvement in the Arctic
basin. Geophys. Res. Lett., 36, L22502. doi:10.1029/2009GL040676.
Stansell, P. and C. MacFarlane, 2002: Experimental investigation of wave breaking criteria based on
wave phase speeds. J. Phys. Oceanogr., 32, 1269–1283.URL link.
Steele, K., C.-C. Teng, and D. W.-C. Wang, 1992: Wave direction measurements using pitch and roll
buoys. Ocean Eng., 19, 349–375.
Stewart, R. H. and J. W. Joy, 1974: HF radio measurements of surface currents. Deep Sea Res., 21,
1039–1049.
Stiassnie, M., 2004: A note on Hasselmann’s energy-transfer model. Annales Hydrographiques, 6e série,
vol. 3, 10–1–10–3.
Stockdon, H. F., R. A. Holman, P. A. Howd, and A. H. Sallenger, Jr., 2006: Empirical parameterization
of setup, swash, and runup. Coastal Eng., 53, 573–588.
Stokes, G. G., 1849: On the theory of oscillatory waves. Trans. Camb. Phil. Soc., 8, 441–455.
Stopa, J. E., F. Ardhuin, A. Bababin, and S. Zieger, 2016a: Comparison and validation
of physical wave parameterizations in spectral wave models. Ocean Modelling, 103, 2–17.
doi:10.1016/j.ocemod.2015.09.003.
Stopa, J. E., F. Ardhuin, and F. Girard-Ardhuin, 2016b: Wave climate in the Arctic 1992-2014: season-
ality and trends. The Cryosphere, 10, 1605–1629. doi:10.5194/tc-10-1605-2016.
Stopa, J. E., F. Ardhuin, R. Husson, H. Jiang, B. Chapron, and F. Collard, 2016c: Swell dissipation from
10 years of envisat asar in wave mode. Geophys. Res. Lett., 43, 3423–3430. doi:10.1002/2015GL067566.
Stopa, J. E. and K. F. Cheung, 2014: Intercomparison of wind and wave data from the ECMWF
reanalysis interim and the NCEP Climate Forecast System Reanalysis. Ocean Modelling, 75, 65–83.
Stopa, J. E., P. Sutherland, and F. Ardhuin, 2018: Strong and highly variable push of ocean waves on
southern ocean sea ice. Proc. Nat. Acad. Sci., 115, 5861–5865. doi:10.1073/pnas.1802011115.
Styles, R. and S. M. Glenn, 2000: Modeling stratified wave and current bottom boundary layers on the
continental shelf. J. Geophys. Res., 105, 24,119–24,139. doi:10.1029/2001JC000864.
Sullivan, P. P. and J. C. McWilliams, 2010: Dynamics of winds and currents coupled to surface waves.
Annu. Rev. Fluid Mech., 42, 19–42.
Sutherland, P. and J. C. Gascard, 2016: Airborne remote sensing of ocean wave directional wavenumber
spectra in the marginal ice zone. Geophys. Res. Lett., 43, 4659–4664. doi:10.1002/grl.53444.
Sutherland, P. and W. K. Melville, 2013: Field measurements and scaling of ocean surface wave-breaking
statistics. Geophys. Res. Lett., 40, 3074–3079. doi:10.1002/grl.50584.
Suzuki, N. and B. Fox-Kemper, 2016: Understanding stokes forces in the wave-averaged equations. J.
Geophys. Res., 121, 3579–3596. doi:10.1002/2015JC011563.
Suzuki, N., B. Fox-Kemper, P. E. Hamlington, and L. P. van Roekel, 2016: Surface waves affect fronto-
genesis. J. Geophys. Res., 121, 3597–3624. doi:10.1002/2015JC011563.
Svendsen, I. A. and U. Putrevu, 1994: Nearshore mixing and dispersion. Proc. Roy. Soc. Lond. A, 445,
561–576.
Sverdrup, H. U. and W. H. Munk, 1947: Wind, sea, and swell: theory of relations for forecasting.
Technical Report 601, U. S. Hydrographic Office.
Symonds, G., D. A. Huntley, and A. J. Bowen, 1982: Two-dimensional surf beat: Long wavegeneration
by a time-varying breakpoint. J. Geophys. Res., 87, 492–498.
Tanaka, M., 2001: Verification of Hasselmann’s energy transfer among surface gravity waves by direct
BIBLIOGRAPHY 221

numerical simulations of primitive equations. J. Fluid Mech., 444, 199–221.

Tayfun, A., 1980: Narrow-band nonlinear sea waves. J. Geophys. Res., 85, 1543–1552.
Tayfun, A. and J.-M. Lo, 1989: Wave envelope and related spectra. J. of Waterway, Port Coast. Ocean
Eng., 115, 515–533. doi:10.1061/(ASCE)0733-950X(1989)115:4(515).
Terray, E. A., W. M. Drennan, and M. A. Donelan, 2000: The vertical structure of shear and dissipation
in the ocean surface layer. Proc. Symp. on Air-Sea Interaction, Sydney, University of New South
Wales, 239–245.
Thomson, J., S. Elgar, and T. Herbers, 2005: Reflection and tunneling of ocean waves observed at a
submarine canyon. Geophys. Res. Lett., 32, L10602. doi:10.1029/2005GL022834.
Thornton, E. B. and R. T. Guza, 1983: Transformation of wave height distribution. J. Geophys. Res.,
88, 5,925–5,938.
— 1986: Surf zone longshore currents and random waves: field data and models. J. Phys. Oceanogr., 16,
1,165–1,178.
Thornton, E. B. and R. F. Kraphol, 1974: Wave particle velocity measured under ocean waves. J.
Geophys. Res., 79, 847–852.
Thorpe, S. A., 1966: On wave interactions in a stratified fluid. J. Fluid Mech., 24, 737–751.
— 1992: The breakup of Langmuir circulation and the instability of an array of vortices. J. Phys.
Oceanogr., 22, 350–360.
Thorpe, S. A., T. R. Osborn, D. M. Farmer, and S. Vagle, 2003a: Bubble clouds and Langmuir circulation:
observations and models. J. Phys. Oceanogr., 33, 122–145.
Thorpe, S. A., T. R. Osborn, J. F. E. Jackson, A. J. Hall, and R. G. Lueck, 2003b: Measurements of
turbulence in the upper-ocean mixing layer using autosub. J. Phys. Oceanogr., 33, 2013–2031.
Toba, Y., 1973: Local balance in the air-sea boundary processes. II partition of wind stress to waves and
current. J. Oceanogr. Soc. Japan, 29, 70–75.URL link.
Tolman, H. L., 1990: The influence of unsteady depths and currents of tides on wind-wave propagation
in shelf seas. J. Phys. Oceanogr., 20, 1166–1174.URL link.
— 1994: Wind waves and moveable-bed bottom friction. J. Phys. Oceanogr., 24, 994–1,009.URL link.
— 1995: On the selection of propagation schemes for a spectral wind wave model. Office Note 411,
NWS/NCEP, 30 pp + figures.
— 2002a: Alleviating the garden sprinkler effect in wind wave models. Ocean Modelling, 4, 269–289.
— 2002b: Limiters in third-generation wind wave models. Global Atmos. Ocean Syst., 8, 67–83.
— 2002c: Testing of WAVEWATCH-III version 2.22 in NCEP’s NWW3 ocean wave model suite. Tech-
nical report, NOAA/NWS/NCEP/MMAB, oMB contribution number 214.
— 2002d: Validation of WAVEWATCH-III version 1.15. Technical Report 213,
NOAA/NWS/NCEP/MMAB.
— 2003: Treatment of unresolved islands and ice in wind wave models. Ocean Modelling, 5, 219–231.
— 2008: A mosaic approach to wind wave modeling. Ocean Modelling, 25, 35–47.
doi:10.1016/j.ocemod.2008.06.005.
Tolman, H. L., M. Accensi, H. Alves, F. Ardhuin, J. Bidlot, N. Booij, A.-C. Bennis, T. Campbell,
D. V. Chalikov, A. Chawla, J.-F. Filipot, M. Foreman, P. Janssen, F. Leckler, Jian-Guo, K. L. M.
Orzech, R. Padilla-Hernández, W. E. Rogers, A. Rawat, A. Roland, M. D. Sikiric, M. Szyszka,
H. L. Tolman, B. Tracy, G. P. van Vledder, A. van der Westhuysen, and S. Zieger, 2014: User
manual and system documentation of WAVEWATCH-IIITM version 2020. Technical Report 316,
NOAA/NWS/NCEP/MMAB.
Tolman, H. L. and J.-H. G. M. Alves, 2005: Numerical modeling of wind waves generated by tropical
cyclones using moving grids. Ocean Modelling, 9, 305–323.
Tolman, H. L. and N. Booij, 1998: Modeling wind waves using wavenumber-direction spectra and a
variable wavenumber grid. Global Atmos. Ocean Syst., 6, 295–309.
Tolman, H. L. and D. Chalikov, 1996: Source terms in a third-generation wind wave model. J. Phys.
Oceanogr., 26, 2497–2518.URL link.
Tournadre, J., N. Bouhier, F. Girard-Ardhuin, and F. Rémy, 2016: Antarctic iceberg distributions 2002–
2010. J. Geophys. Res., 121, 327–349. doi:10.1002/2015JC011178.
Tournadre, J., F. Girard-Ardhuin, and B. Legrésy, 2012: Antarctic iceberg distributions 2002–2010. J.
Geophys. Res., 117, C05004. doi:10.1029/2011JC007441.
Traykovski, P., A. E. Hay, J. D. Irish, and J. F. Lynch, 1999: Geometry, migration, and evolution of
wave orbital ripples at LEO-15. J. Geophys. Res., 104, 1,505–1,524.
Trowbridge, J. and O. S. Madsen, 1984: Turbulent wave boundary layers. 1. model formulation and
first-order solution. J. Geophys. Res., 89, 7989–7997.
222 BIBLIOGRAPHY

Tyler, G. L., C. C. Teague, R. H. Stewart, A. M. Peterson, W. H. Munk, and J. W. Joy, 1974: Wave
directional spectra from synthetic aperture observations of radio scatter. Deep Sea Res., 21, 989–1016.
U. S. Army Corps of Engineers, 2002: Coastal Engineering Manual, Engineering Manual 1110-2-1100 .
U. S. Army Corps of Engineers, Washington, D. C.URL link.
Ursell, F., 1953: The long-wave paradox in the theory of gravity waves. Proceedings of the Cambridge
philosophical society, 49, 685–694.
— 1999: Reminiscences of the early days of the spectrum of ocean waves. Wind-over-wave couplings,
S. G. Sajjadi, N. H. Thomas, and J. C. R. Hunt, eds., Clarendon Press, Oxford, U. K., 127–137.
Vachon, P. W. and J. Wolfe, 2011: C-band cross-polarization wind speed retrieval. IEEE Geoscience
And Remote Sensing Letters, 8, 456–459. doi:10.1109/LGRS.2010.2085417.
van Vledder, G. P., 2006: The WRT method for the computation of non-linear four-wave interactions in
discrete spectral wave models. Coastal Eng., 53, 223–242.
van Vledder, G. P., S. T. C. Hulst, and J. D. McConochie, 2016: Source term balance in a severe storm
in the southern north sea. Ocean Dynamics, 66. doi:10.1007/s10236-016-0998-z.
Vandemark, D., N. Tran, B. D. Beckley, B. Chapron, and P. Gaspar, 2002: Direct estimation
of sea state impacts on radar altimeter sea level measurements. Geophys. Res. Lett., 29, 2148.
doi:10.1029/2002GL015776.
Veron, F., 2015: Sea spray. Annu. Rev. Fluid Mech., 47, 507–538. doi:10.1146/annurev-fluid-010814-
014651.
Veron, F., W. K. Melville, and L. Lenain, 2008: Infrared techniques for measuring ocean surface processes.
J. Atmos. Ocean Technol., 25, 307–326. doi:10.1175/2007JTECHO524.1.
von Gerstner, F. J., 1809: Theorie der wellen. Ann. Phys., 32, 412–440.
Wadhams, P., V. A. Squire, J. A. Ewing, and R. W. Pascal, 1986: The effect of the marginal ice zone
on the directional wave spectrum of the ocean. J. Phys. Oceanogr., 16, 358–376.
WAMDI Group, 1988: The WAM model - a third generation ocean wave prediction model. J. Phys.
Oceanogr., 18, 1775–1810.URL link.
Wang, H., A. B. Villas Bôas, W. R. Young, and J. Vanneste, 2023: Scattering of swell by currents. J.
Fluid Mech., 975. doi:10.1017/jfm.2023.686.
Wang, R. and H. H. Shen, 2010: Gravity waves propagating into an ice-covered ocean: A viscoelastic
model. J. Geophys. Res., 115, C06024. doi:10.1029/2009JC005591.
Wang, X. L., Y. Feng, and V. R. Swail, 2014: Changes in global ocean wave heights as projected using
multimodel cmip5 simulations. Geophys. Res. Lett., 41, 1026–1034. doi:10.1002/2013GL058650.
Wang, X. L. and V. R. Swail, 2004: Trends of atlantic wave extremes as simulated in a 40-yr wave
hindcast using kinematically reanalyzed wind fields. Journal of Climate, 15, 1020–1035.URL link.
Weber, J. E., 1987: Wave attenuation and wave drift in the marginal ice zone. J. Phys. Oceanogr., 17,
2351–2361.
Weber, J. E. and E. Førland, 1990: Effect of the air on the drift velocity of water waves. J. Fluid Mech.,
218, 619–640.
Weeks, W. F., 2010: On sea ice. University of Alaska Press, Fairbanks.
Welch, P. D., 1967: The use of fast Fourier transform for the estimation of power spectra: a method
based on time averaging over short, modified periodograms. IEEE Trans. Audio and Electroacoustics,
15, 70–73.
Weller, R. A., J. P. Dean, J. Marra, J. F. Price, E. A. Francis, and D. C. Boardman, 1985: Three-
dimensional flow in the upper ocean. Science, 277, 1552–1556. doi:10.1126/science.227.4694.1552.
Wessel, P. and W. H. F. Smith, 1996: A Global Self-consistent Hierarchical, High-resolution Shoreline
database. J. Geophys. Res., 101, 8741–8743.
Whitham, G. B., 1962: Mass, momentum and energy flux in water waves. J. Fluid Mech., 12, 135–147.
Wiberg, P. L., 1995: A theoretical investigation of boundary layer flow and bottom shear stress for
smooth, transitional and rough flow under waves. J. Geophys. Res., 100, 22,667–22,679.
Wiberg, P. L. and C. K. Harris, 1994: Ripple geometry in wave-dominated environments. J. Geophys.
Res., 99, 775–789.
Wilson, K. C., 1989: Friction of wave-induced sheet flow. Coastal Eng., 12, 371–379.
Winterwerp, J., R. de Graaff, J. Groeneweg, and A. P. Luijendijk, 2003: Modelling of wave damping at
Guyana mud coast. Coastal Eng., 54, 249–261, 10.1016/j.coastaleng.2006.08.012.
Wolf, J. and D. Prandle, 1999: Some observations of wave–current interaction. Coastal Eng., 37, 471–485.
Woodward, P., 1952: The spectrum of random frequency modulation. Technical Report 168, Telecom-
munications Research Establishement, Great Malvern, Worcs., England.
Wyatt, L. R., 2000: Limits to the inversion of HF radar backscatter for ocean wave measurement. J.
BIBLIOGRAPHY 223

Atmos. Ocean Technol., 17, 1651–1665.

Xu, Z. and A. J. Bowen, 1994: Wave- and wind-driven flow in water of finite depth. J. Phys. Oceanogr.,
24, 1850–1866.URL link.
Yanenko, N. N., 1971: The method of fractional steps. Springer-Verlag.
Young, I. R., 1986: Probability distribution of spectral integrals. J. of Waterway, Port Coast. Ocean
Eng., 112, 338–341.
— 1998: An experimental investigation of the role of atmospheric stability in wind wave growth. Coastal
Eng., 34, 23–33.
— 1999: Wind generated ocean waves. Elsevier Science, Oxford, 288 pp.
Young, I. R., W. Rosenthal, and F. Ziemer, 1985: A three-dimensional analysis of marine radar images for
the determination of ocean wave directionality and surface currents. J. Geophys. Res., 90, 1049–1059.
Young, I. R. and G. P. van Vledder, 1993: A review of the central role of nonlinear interactions in
wind-wave evolution. Phil. Trans. Roy. Soc. London A, 342, 505–524.
Young, I. R., L. A. Verhagen, and M. L. Banner, 1995: A note on the bimodal directional spreading of
fetch-limited wind waves. J. Geophys. Res., 100, 773–778.
Yueh, S. H., W. Tang, A. G. Fore, G. Neumann, A. Hayashi, A. Freedman, J. Chaubell, and G. S. E.
Lagerloef, 2013: L-band passive and active microwave geophysical model functions of ocean surface
winds and applications to aquarius retrieval. IEEE Trans. on Geosci. and Remote Sensing, 51, 4619–
4632.
Yurovskaya, M. V., V. A. Dulov, B. Chapron, and V. N. Kudryavtsev, 2013: Directional short
wind wave spectra derived from the sea surface photography. J. Geophys. Res., 113, C12024.
doi:10.1002/jgrc.20296.
Zakharov, V. E., 1968: Stability of periodic waves of finite amplitude on the surface of a deep fluid. J.
Appl. Mech. Tech. Phys., 2, 190–194.
— 1999: Statistical theory of gravity and capillary waves on the surface of a finite-depth fluid. Eur. J.
Mech. B/Fluids, 18, 327–334.
Zakharov, V. E. and M. M. Zaslavskii, 1982: The kinetic equation and Kolmogorov spectra in the weak
turbulence theory of wind waves. Izv. Atmos. Ocean. Phys., 18, 747–753.
Zappa, C. J., M. L. Banner, H. Schultz, A. Corrada-Emmanuel, L. B. Wolff, and J. Yalcin, 2008: Retrieval
of short ocean wave slope using polarimetric imaging. Meas. Sci. Technol., 19, 055503.
Zhukovets, A. M., 1963: The influence of bottom roughness on wave motion in a shallow body of water.
Izv. Geophys. Ser. Acad. Sci., USSR, 10, 1561–1570.
Zieger, S., A. V. Babanin, W. E. Rogers, and I. R. Young, 2015: Observation-based
source terms in the third-generation wave model WAVEWATCH. Ocean Modelling, 96, 2–25.
doi:10.1016/j.ocemod.2015.07.014.
Zijlema, M., G. Stelling, and P. Smit, 2011: Swash: An operational public domain code for
simulating wave fields and rapidly varied flows in coastal waters. Coastal Eng., 58, 992–1012.
doi:10.1016/j.coastaleng.2011.05.015.
Zilitinkevitch, S. S. and D. V. Chalikov, 1968: Determinint the universal wind-velocity and temperature
profiles in the atmospheric boundary layer. Izv. Atmos. Ocean. Phys., 4, 394–302.
Zou, Q., 2004: A simple model for random wave bottom friction and dissipation. J. Phys. Oceanogr., 34,
1460–1467.