E/MEarth in MotionA VISUAL SCIENCE PROJECT31 CHAPTERS / 5 WAVE LAB EXPERIENCES
Skip to chapter navigation
Browse

31 chapters

View the learning journey →
Explore ocean & coasts
Choose a chapter
Wave Lab ↗Overview ↑
OBSERVE / QUESTION / EXPERIMENT

The planet moves.
Make sense
of it.

An open learning space for the ocean, atmosphere and the patterns hidden in data. Start with a wave. Follow your curiosity.

Start with ocean waves ↗

Ocean · Weather · Climate · Data

FIELD 001 / OSCILLATION
ONE PATTERN. MANY PROCESSES.
Waves · Wind · Change · Chance↓ DISCOVER
Coastal modelling / Waves first

Coastal modelling.
Start with waves.

Begin with the numerical wave modelling pathway: model selection, wave forcing, boundaries, convergence, independent evaluation and uncertainty. This coastal modelling hub is designed to expand with future hydrodynamic, sediment and coupled-modelling modules.

01 Ocean & coasts Explore now ↗
02 Weather Planned
03 Climate Planned
04 Data & statistics Planned
COLLECTION 01 / OCEAN & COASTS

From a single wave
to a changing coast.

01 / THE FOUNDATIONS

Meet a wave.

Learn the language of crests, troughs and periods. Separate the travelling pattern from the moving water.

Begin here →
02 / THE JOURNEY

One wave.
Three depth regimes.

Explore why the same wave behaves differently as the water beneath it becomes shallower.

Open the experiment →
03 / THE NEARSHORE

Why waves
change height.

Explore shoaling: how changing energy-transport speed affects wave height before breaking.

Explore shoaling →
04 / DIRECTION

Why wave crests turn.

Explore refraction over parallel depth contours, and connect it to Snell’s law.

Explore refraction →
06 / BREAKING WAVES

Where surf begins.

Explore approximate breaking limits and recognise spilling, plunging and surging behaviour.

Explore wave breaking →
08 / WAVE SUPERPOSITION

From one wave
to many.

Add frequencies and phases one at a time. Watch a regular oscillation become an irregular-looking sea-surface record.

Build an irregular wave →
09 / IRREGULAR SEAS

Wave-by-wave statistics

Count waves, then compare their heights, periods and probabilities.

Explore chapter 09 →
10 / IRREGULAR SEAS

FFT and spectral moments

Connect a sampled record to its frequency variance density.

Explore chapter 10 →
11 / IRREGULAR SEAS

Spectral shapes and models

Compare normalized PM-shaped, JONSWAP and TMA examples.

Explore chapter 11 →
12 / IRREGULAR SEAS

Directional spectra

Reveal crossing systems, spreading and directional marginals.

Explore chapter 12 →
13 / IRREGULAR SEAS

Random seas and responses

Change phases, inspect grouping and follow pressure and velocity below the surface.

Explore chapter 13 →
14 / IRREGULAR SEAS

Spectra toward shore

Explore ideal spectral shoaling and energy-flux conservation.

Explore chapter 14 →
15 / WAVE THEORY

When a crest grows sharper.

Compare linear and second-order profiles and recognise the limits of the expansion.

Explore chapter 15 →
16 / WAVE THEORY

Choose the theory for the question.

Explore relative depth, height and the assumptions behind different wave descriptions.

Explore chapter 16 →
17 / SPECTRAL MODELLING

Each component takes its own path.

Follow directional systems over parallel depth contours.

Explore chapter 17 →
18 / SPECTRAL MODELLING

Why a spectrum changes.

Separate local gains, losses and conservative redistribution.

Explore chapter 18 →
19 / SPECTRAL MODELLING

A spectrum on the move.

Connect transport, source terms and the action balance.

Explore chapter 19 →
20 / MODELLING PRACTICE

Turn equations into a calculation.

Explore geographic and spectral resolution, boundaries and convergence.

Explore chapter 20 →
21 / MODELLING PRACTICE

Agreement needs more than one number.

Separate verification, calibration and independent performance evaluation.

Explore chapter 21 →
22 / MODELLING PRACTICE

Follow a sea state toward the coast.

Run a transparent spectral transect and export its conservation account.

Explore chapter 22 →
23 / STRUCTURES

Behind the tip.

Explore a wave field spreading into the shadow of a breakwater.

Explore chapter 23 →
24 / STRUCTURES

Follow the wave power.

Separate reflection, transmission and dissipation with a linked wave and power experiment.

Explore chapter 24 →
25 / SHORELINE RESPONSE

Where does the water reach?

Separate run-up, crest freeboard and mean overtopping discharge.

Explore chapter 25 →
26 / LONG WAVES

A slower coastal rhythm.

Distinguish a short-wave group envelope from low-frequency surface motion.

Explore chapter 26 →
27 / BASIN OSCILLATIONS

A basin has its own rhythm.

Explore standing modes, natural periods and the role of damping in resonance.

Explore chapter 27 →
28 / OBSERVING WAVES

From sensor to sea state.

Explore pressure response, wave instruments and the limits of measurement.

Explore chapter 28 →
29 / DATA QUALITY

Keep the evidence.

Inspect faults, spectra and flags through a practical QC method guide.

Explore chapter 29 →
Chapter 01 / Understanding ocean waves

Read the sea.
One idea at a time.

A wave arriving at the beach connects distant weather to local coastal change. Before exploring the equations, learn what moves, what we measure and why the same wave behaves differently as the water becomes shallower.

01 / A travelling disturbance

The pattern travels.
The water oscillates.

Watch a floating object as a crest passes: it rises and falls while the crest continues onwards. Water particles also move horizontally. In the simplest small-amplitude theory, their paths are closed circles or ellipses; real waves can produce net drift, and currents can carry water along.

Gravity pulls a displaced sea surface back towards equilibrium. Inertia carries the motion onwards, sustaining the oscillation. For very small ripples, surface tension also matters.

A wave is not a solid ridge of water travelling intact across the ocean. It transfers energy through moving water.

02 / From wind to swell

A storm can send a distant signal.

Wind transfers energy to the sea surface. Wave growth depends on wind strength, how long it blows and the open-water distance over which it acts—the fetch. Depth can also constrain growth.

Wind sea is still being generated by the local wind. Swell has travelled away from its generating area and often looks more organised. Both can occur together, arriving from different directions.

In deep water, longer-period components travel faster. This sorting, called dispersion, helps explain why distant swell can arrive before the shorter waves from the same storm.

Explore: NOAA JetStream — waves and their generation ↗

03 / Learn to describe a wave

Space and time tell different stories.

Anatomy of an idealised regular waveTwo crests separated by wavelength L, a trough, still water level, and height H measured vertically between crest and trough elevations.Wavelength LCrestTroughHeight HStill water levelOne snapshot in space · schematic
Height · HVertical distance between trough and crest. For a sinusoid, amplitude a = H/2.
Wavelength · LHorizontal spacing between successive crests, measured along travel direction.
Period · TTime between successive crests passing one fixed location.
Depth · hDistance from still water level down to the seabed.
Go deeper: speed, frequency and steepness

Frequency counts cycles per second: f = 1/T. Phase speed follows an individual crest: c = L/T. Steepness H/L compares height with length. These describe different properties: a tall wave is not necessarily a steep wave.

Group speed Cg describes the propagation of a narrow wave packet and its energy in linear theory without currents. It is half the phase speed in deep water and approaches the phase speed in shallow water.

Watch a crest move through a group.

Blue: the sum of two nearby frequencies. Coral dashed: its amplitude envelope. In this deep-water example, crests move at approximately twice the envelope speed. The drawing shows a repeating beat pattern, not a single isolated packet. Vertical size is normalised; horizontal distance is in carrier wavelengths.

04 / Depth is relative

“Deep” compared with what?

Compare water depth with wavelength using h/L. These conventional ranges describe a gradual transition, not sudden physical boundaries.

h/L ≥ 0.5

Deep water

Motion decays strongly with depth. The seabed has little influence on wave speed. Near-surface particle paths are approximately circular.

0.05 < h/L < 0.5

Intermediate water

Both period and depth affect speed. Particle paths flatten towards the bottom as the seabed restricts vertical motion.

h/L ≤ 0.05

Shallow water

Motion extends throughout the depth. Paths are strongly flattened, and wave speed approaches √(gh).

The same location can be deep for short waves and shallow for much longer waves. A tsunami can behave as a shallow-water wave even while crossing the deep ocean.

Test depth and period in the wave lab →
05 / The coast changes the wave

Several processes act together.

Shoaling changes wave height as the speed of energy transport changes with depth. Height can initially decrease before growing nearer shore.

Refraction turns the wave direction when one part of a crest travels more slowly than another. Diffraction spreads wave energy around obstacles or through openings. Reflection sends part of the wave energy back from a boundary.

Breaking releases organised wave energy into turbulence and currents. Bottom friction also dissipates energy. Waves help move sediment and reshape beaches, with the outcome depending on the waves, currents and seabed.

Explore the animated shoaling experiment →
06 / Beyond everyday surf

Not all waves come directly from wind.

Infragravity waves: slower coastal oscillations

Interactions among shorter waves and variations in breaking can generate much longer oscillations. These can become important in the surf zone, influencing shoreline water motion and run-up.

Seiches: water sloshing within a basin

A lake, bay or harbour has natural oscillation periods set by its geometry and depth. Disturbances can excite these modes; forcing near a natural period can amplify the response through resonance.

Tsunamis: displacement of a large water volume

Seabed movement or a landslide can generate very long waves. Their long wavelength makes depth important even offshore. They are physically different from ordinary wind-generated surf.

Broader overview: Coastal Wiki — Waves, Karsten Mangor and contributors ↗. Further context for this learning journey; this introduction and its diagram are independently written and drawn.

07 / From one wave to a sea state

The ocean does not repeat a perfect sine wave.

Real seas combine waves with different heights, periods and directions. Their superposition produces an irregular surface and groups of larger and smaller waves. Our regular-wave experiments isolate one process so its effect is easier to see.

What does “2-metre seas” mean?

Significant wave height, traditionally H1/3, is the average height of the highest third of individual waves in a record. It is not the largest wave: individual waves can be considerably higher.

Buoy products commonly estimate a closely related quantity, Hm0, from the wave spectrum. A spectrum shows how surface-elevation variance is distributed across frequencies; directional information adds where the components come from.

NOAA NDBC — how buoy wave statistics are calculated ↗

Observe, then interpret.

A buoy records motion over time. Other instruments use pressure, sound or radar to infer wave properties. A useful description includes height, period, direction, location and the time interval measured.

A sea state describes conditions during a limited interval. A wave climate describes their longer-term patterns, including seasons and extremes. This distinction will connect our ocean chapters with weather, climate and statistics.

Check your understanding: do taller waves always travel faster?

No. In small-amplitude linear theory, wave speed depends on period and depth, not height. Changing height in the shoaling lab changes the energy and breaking estimate; it does not change the calculated wavelength. Strongly nonlinear waves require a more detailed theory.

Start with the ideas above, then use the experiments to connect the picture to the mathematics. The models deliberately omit many real-world effects; each lab explains its assumptions.

Continue to Chapter 02: depth & dispersion →
EXPERIMENT 001 / DEPTH & DISPERSION

How deep is “deep water”?

Change the water depth or wave period. Compare the calculated wavelength and the changing shapes of water-particle orbits. The profile rescales to display two wavelengths.

Top: two wavelengths in a rescaled view, elevation exaggerated. Bottom: equal-axis orbit diagrams at three horizontal stations; one common enlargement factor. These are separate views, not a single water cross-section.
WAVELENGTH
PHASE SPEED
RELATIVE DEPTH

Try this.

Choose Intermediate water. Keep the depth fixed, then increase the period. Longer waves feel the bottom more strongly—even though the seabed has not moved.

Three regimes, gradual transitions.

Deep: h/L ≥ 0.5. Particle motion decays rapidly below the surface. Depth has little influence on phase speed.

Intermediate: 0.05 < h/L < 0.5. The seabed increasingly influences motion and speed. Particle paths become flatter with depth.

Shallow: h/L ≤ 0.05. Motion extends through the water column. Phase speed approaches √(gh).

Why does the pattern change?

Water cannot flow through a solid seabed. Vertical motion diminishes near the bottom, flattening particle paths. When depth is small relative to wavelength, that constraint affects the entire water column.

Show me the mathematics
ω2=gktanh(kh)

ω = 2π/T is angular frequency; k = 2π/L is wavenumber; g = 9.81 m/s². We solve this equation for k, then calculate L and phase speed c = L/T.

Deep-water limit: c ≈ gT/(2π). Shallow-water limit: c ≈ √(gh). The full equation gives a continuous transition.

Assumptions & visual interpretation

Small-amplitude, regular, nonbreaking waves in uniform depth with no current. Each slider setting is a separate local wave state, not a model of a wave crossing the entire beach. No dimensional wave height is assigned to the normalised orbit display; shoaling is not calculated here.

Orbit shapes and phase use linear theory. Within each setting, all orbits share one enlargement factor, preserving relative sizes and ellipse aspect ratios. Orbit size is normalised again when settings change, so absolute excursions cannot be compared across settings. The surface graph has a different scale. Motion follows the selected period. Closed orbits omit higher-order drift.

Explore the open references

NOAA NDBC: How are ocean waves described? ↗
Start with wave terminology.

Theory reference: Holthuijsen (2007), Chapter 5, §§5.4–5.5, for particle motion, dispersion and energy transport.

CHAPTER 02 / EQUATION REFERENCE

One theory, three useful forms.

The full linear dispersion relation, ω² = gk tanh(kd), connects depth, frequency and wavelength. The shallow and deep expressions are its limiting approximations. The middle column remains valid across depths within linear theory; it is not restricted mathematically to the transitional range.

Use a = H/2, k = 2π/L, ω = 2π/T and ψ = kx − ωt. Depth d is positive and corresponds to h in the experiment above; z is positive upward from still water, so the bed is z = −d. Waves travel in +x. The symbol ψ denotes wave phase, leaving θ available for a compass bearing in later chapters.

Linear progressive waves: shallow and deep limits of the finite-depth solution
QuantityShallow limitGeneral finite depth
including transitional water
Deep limit
Relative depthdL<125125≤dL≤12dL>12
Elevation η (m)acos⁡ψacos⁡ψacos⁡ψ
Phase speed C (m/s)gdωk=gωtanh⁡(kd)gω
Wavelength L (m)TgdgT22πtanh⁡(kd)gT22π
Group speed Cg (m/s)gd=C12[1+2kdsinh⁡(2kd)]CC2=g2ω
Horizontal velocity u (m/s)agdcos⁡ψaωcosh⁡[k(z+d)]sinh⁡(kd)cos⁡ψaωekzcos⁡ψ
Vertical velocity w (m/s)aω(1+zd)sin⁡ψaωsinh⁡[k(z+d)]sinh⁡(kd)sin⁡ψaωekzsin⁡ψ
Horizontal acceleration ax (m/s²)aωgdsin⁡ψaω2cosh⁡[k(z+d)]sinh⁡(kd)sin⁡ψaω2ekzsin⁡ψ
Vertical acceleration az (m/s²)−aω2(1+zd)cos⁡ψ−aω2sinh⁡[k(z+d)]sinh⁡(kd)cos⁡ψ−aω2ekzcos⁡ψ
Horizontal displacement ξ (m)−aωgdsin⁡ψ−acosh⁡[k(z+d)]sinh⁡(kd)sin⁡ψ−aekzsin⁡ψ
Vertical displacement ζ (m)a(1+zd)cos⁡ψasinh⁡[k(z+d)]sinh⁡(kd)cos⁡ψaekzcos⁡ψ
Total gauge pressure p (Pa)ρg(η−z)ρgacosh⁡[k(z+d)]cosh⁡(kd)cos⁡ψ−ρgzρgaekzcos⁡ψ−ρgz

The table uses the CEM's conservative shallow-water boundary d/L = 1/25. The introductory lab uses the also-common 1/20 convention. These are accuracy guides, not abrupt physical transitions. Deep water requires d/L greater than about 1/2; the opposite inequality would be incorrect.

How to read velocities, accelerations and displacements

u and w are horizontal and upward velocities at a fixed mean position. To first order, differentiating the displacement in time gives velocity, and differentiating velocity gives acceleration: ∂ψ/∂t = −ω fixes the signs. The displacement describes the oscillatory orbit about a mean position; this first-order theory does not include net drift.

In shallow water, horizontal velocity is nearly uniform with depth, while vertical velocity vanishes at the bed. In deep water the orbital amplitudes decrease approximately as ekz. An infinitely deep approximation does not enforce a finite-bed boundary exactly.

Pressure and validity

Total gauge pressure is measured relative to atmospheric pressure. Separate it into hydrostatic pressure −ρgz and wave-induced pressure p′. The pressure transfer factor Kp = cosh[k(z+d)]/cosh(kd) attenuates the oscillating pressure below the surface. At z = 0, linear theory gives p′ = ρgη; it applies the boundary condition at the still-water level rather than evaluating the exact moving surface.

The approximations assume small amplitude, inviscid and irrotational flow, a level impermeable bed and no background current. Small d/L alone does not guarantee linear behaviour: large H/d and strong shallow-water nonlinearity require a more appropriate wave theory.

CHAPTER 03 / SHOALING

Shallower water.
Always taller waves?

Not immediately. A wave can become slightly smaller before it grows towards shore. To understand why, follow the speed at which its energy travels.

MAIN INTERACTIVE EXPERIMENT · THREE LINKED VIEWS

Shoaling: one wave, three scientific perspectives

Adjust a regular incident wave and follow the same calculation through a 1D virtual gauge, a 2D cross-shore section and a rotatable 3D surface view. Period, local wavelength and height are linked by linear dispersion and idealised energy-flux conservation.

The interactive shoaling experiment loads when this section is in view. Alternatively explore it in Wave Lab.

Model boundary: Regular, normally incident linear waves, gradual depth changes and constant energy flux until a depth-limited breaking screen. The 3D visualization is an animated perspective of this solution, not a three-dimensional hydrodynamic solver or a breaking-wave calculation.

The full Wave Lab version remains available for standalone exploration.

Supplementary experiment — original shoaling profile and energy-flux diagram

This earlier analytical experiment retains the detailed depth-profile plot, height envelope and point-by-point energy interpretation. Use it to check the linked views above.

OBSERVE / INSPECT / EXPLAIN

The journey over a sloping seabed

Linear sinusoidal waves are illustrated; real waves become asymmetric near breaking. Click to inspect a depth. Coral: selected point · Dashed boundary: estimated breaking · Beyond the boundary: no wave prediction.

Blue: calculated height before breaking · Grey dashed: input height · Coral: selected depth. The horizontal depth axis is logarithmic; it does not represent distance.

LOCAL HEIGHT
HEIGHT / REFERENCE
GROUP SPEED

Follow the energy.

Shoaling is the change in wave height caused by changing water depth. In this idealised experiment, no energy is added or lost before breaking. The same energy flux must pass each section.

Wave energy travels at group speed, which differs from the speed of an individual crest. When group speed increases, wave height decreases. When it decreases towards shallow water, wave height increases.

Try this.

  1. Keep the default period of 8 s and input height of 1 m.
  2. Move the selected depth from 50 m towards 10 m. Look for the small dip in height.
  3. Continue towards 2 m. Compare the growing height with the decreasing group speed.
  4. Increase the input height. Notice where the calculation reaches its breaking boundary.

In this stationary-depth, no-current model, the wave period stays constant along the propagation path. Changing the period slider starts a different experiment.

Connect the picture to the equation
01 / WAVE ENERGYE=ρgH28

Energy per unit surface area grows with height squared.

02 / CONSERVE ENERGY FLUXECg=constant

In this model, energy is transported without loss before breaking.

03 / FIND THE NEW HEIGHTKs=Cg,refCg

Compare local group speed with the reference speed.

H=HrefKs

H is crest-to-trough height (m); Cg is group speed (m/s); Ks is dimensionless. The reference is at 50 m depth, which is not necessarily deep water for every period. Here Kₛ denotes a height ratio relative to that reference, rather than necessarily the conventional deep-water shoaling coefficient.

Where does the calculation stop?

The curve ends at the first point where the predicted regular-wave height reaches the smaller of 0.78h and 0.142L tanh(kh). These approximate depth and steepness limits flag the end of the unbroken-wave illustration. Actual breaking depends on wave and seabed conditions.

Beyond that point, local height and the shoaling factor are hidden. We do not extend conservation of unbroken-wave energy flux into the surf zone.

Assumptions and group speed

Small-amplitude regular waves approach a slowly varying bed normally. No currents, refraction, reflection, wind input or dissipation before breaking. This is a local-depth transformation, not a resolved beach or breaking model.

Cg=L2T[1+2khsinh(2kh)]

Wavelength L comes from the same linear dispersion relation as Chapter 02. Near breaking, linear theory becomes less reliable; the boundary is an educational screening rule.

Theory references: Holthuijsen (2007), §7.3.1, pp. 199–201; Reeve, Chadwick & Fleming (2018), §2.3.2, pp. 33–34. The breaking approximation is discussed separately from the linear shoaling solution.

← Back to the learning journey
CHAPTER 04 / REFRACTION

One crest.
Different speeds.

Why do waves often approach a beach more directly than they did offshore? The part of a crest in shallower water travels more slowly, turning the wave towards the shore-normal.

01 Observe the turn02 Change the approach03 Connect it to Snell’s law
MAIN INTERACTIVE EXPERIMENT · THREE LINKED VIEWS

Refraction: one analytical phase field, three linked perspectives

Use the corrected linked simulator to compare the same regular-wave refraction calculation through a 1D virtual gauge, a 2D wavefront navigation map and a 3D travelling-wave surface. The new Match 2D (top) view lets you check geometry directly; Oblique 3D restores the illustrative surface relief. A height-exaggeration control is included for easier interpretation.

The interactive refraction experiment is loading. The same simulator is also included as a standalone file in the ZIP package.

Model boundary: Regular linear waves over straight, parallel depth contours with no currents and no dissipation. The 3D panel is a linked rendering of the same analytical phase field shown in the 2D map; it is not a three-dimensional fluid solver. Use Match 2D (top) to confirm crest alignment, then switch to Oblique 3D for an illustrative surface view.

The earlier Chapter 4 plan-view experiment is retained below. You can also open the refraction studio in Wave Lab for a dedicated full-size exploration.

Supplementary experiment — original Chapter 4 refraction lab

This earlier analytical experiment retains the original plan-view construction, local metrics and worked Snell’s-law calculation. Use it as a cross-check against the linked 1D–2D–3D studio above.

Plan view · Equal horizontal distance scales · Blue lines: crests · Coral lines: rays (normal to crests) · Dashed lines: depth contours. This domain ends at 5 m depth, before the shoreline. Line colour and thickness do not represent wave height.

LOCAL ANGLE
LOCAL WAVELENGTH
LOCAL PHASE SPEED
UNDERSTAND / DEPTH AND DIRECTION

The seabed changes the speed.

The wave’s period is fixed in this stationary setting. As depth decreases, the dispersion relation gives a shorter wavelength and lower phase speed. An oblique crest spans different depths, so its shallow-water portion falls behind its deeper-water portion. The crest rotates towards alignment with the depth contours.

A wave ray follows the direction of propagation. Here it is perpendicular to the crest; it is not a water-particle trajectory. As the crest aligns with the contours, the ray turns towards their normal.

Try three comparisons.

  1. Set the approach angle to 0°. Wavelength still decreases, but the waves do not turn.
  2. Return to 40°. Compare the ray direction at 50 m and 10 m.
  3. Keep the incident angle fixed and compare 4 s and 16 s. Longer-period waves respond to the seabed farther offshore.

Shoaling changes height through changing group speed. Refraction changes direction and can also redistribute energy through changing ray spacing. This lesson calculates crest geometry and direction. Explore combined height changes in Chapter 05.

Connect the picture to Snell’s law

For straight, parallel depth contours with no currents:

sinθc=sinθrefcref=constant

θ is measured from the contour normal; c = L/T is phase speed. As c decreases, sin θ decreases, so the propagation direction turns towards the normal.

Go deeper: why parallel contours matter

The alongshore wavenumber stays constant: k sin θ = kref sin θref. Each crest is a line of constant phase. The crest positions follow a phase calculation that accounts for the changing cross-shore wavenumber.

For a stationary, lossless ray tube, energy flux is E Cg b = constant, where b is spacing measured perpendicular to the rays. With parallel contours this gives H/Href = √[Cg,ref cos θref / (Cg cos θ)]. Try the combined height calculation in Chapter 05.

Assumptions and limits

Small-amplitude, monochromatic waves over a gently varying, alongshore-uniform bed. No currents, reflection, diffraction, wind input, breaking or dissipation. The plotted crests show phase, not a resolved three-dimensional water surface.

The 50 m reference is not necessarily deep water. At every depth the full linear dispersion relation is used. Curved contours, shoals and headlands need a more general calculation; this straight-contour experiment cannot predict their focusing patterns.

Check your understanding

Will waves turn at normal incidence? No: θ = 0 remains a solution at every depth. Their wavelength and speed still change.

Is a ray the path of a floating object? No. It follows wave propagation; water particles mainly oscillate, with additional drift and currents possible in real seas.

Scientific basis and further reading

Holthuijsen (2007), Waves in Oceanic and Coastal Waters, §7.3.2, pp. 202–210; Sorensen (2006), Basic Coastal Engineering, Chapter 4, §4.2; Reeve, Chadwick & Fleming (2018), Coastal Engineering, §§2.3.1–2.3.3. These are supporting references; a purchase is not required to use this lesson.

Free overview: Coastal Wiki — Waves ↗

Verification: zero-angle limit, conservation of alongshore wavenumber, and Snell’s-law benchmarks. Original explanatory text and graphics. Model reviewed November 2026.

CHAPTER 05 / SHOALING + REFRACTION

A turn changes
the height story.

A wave slows as it enters shallower water, but slowing alone does not tell us its new height. The direction of energy travel matters too. Compare two approaches to the same straight coast.

01 Compare two predictions02 Follow the energy03 Check the limits
ONE WAVE FIELD / TWO VIEWS

Look from the side. Then look from above.

Both views show the selected oblique-incidence case (the coral curve in the comparison chart). Fixed beach slope: 1:80. Adjust the shared conditions ↓. The same dark dashed line marks your selected depth.

A · Profile view — a cross-shore slice

B · Bird’s-eye view — the same field from above

Blue: normal incidence, shoaling alone · Coral: selected oblique incidence, shoaling and refraction · Each curve ends at its own estimated breaking onset. Depth decreases to the right on a logarithmic axis; this is not distance offshore.

SHOALING ALONE
COMBINED HEIGHT
LOCAL ANGLE

What exactly are we comparing?

Both cases start with the same wave height and period at 50 m depth. The blue case approaches the depth contours at right angles. The coral case starts at your selected angle. They therefore begin with equal energy density, but different energy flux towards shore.

For these parallel contours, turning towards the normal spreads neighbouring ray paths farther apart when their spacing is measured perpendicular to the rays. The resulting refraction factor is at most one. It reduces the height relative to the normal-incidence calculation at the same depth.

This does not mean refraction always lowers waves. A shoal or headland can concentrate ray paths and increase local height. The parallel-contour model here does not represent that geometry.

Try a controlled comparison.

  1. Choose 0°. The curves should coincide: there is no directional correction.
  2. Choose 40°, then 65°. Inspect 5 m depth and compare both heights.
  3. Increase the input height. Observe where each curve stops; the same breaking rule is applied separately to each case.
Check your understanding

Can the combined wave still grow above its input height? Yes. A sufficiently large shoaling factor can outweigh the reduction associated with refraction.

Does the coral curve lose energy through friction? No. This comparison contains no dissipation before breaking. Geometry and group speed account for its height change.

Connect both processes to the equation

For this stationary, alongshore-uniform coast without currents or energy losses, the shoreward energy flux per unit alongshore width is conserved:

ECgcosθ=constantHHref=Cg,refcosθrefCgcosθ=KsKr

Ks = √(Cg,ref/Cg) represents the group-speed change. Kr = √(cos θref/cos θ) represents refraction here. Angles are measured from the shore-normal. Both factors refer to 50 m, which need not be deep water.

Model limits and breaking markers

Small-amplitude regular waves, gently varying parallel contours, no reflection, diffraction, current, wind input or bottom friction. This is a comparison of two idealised incident directions, not a forecast of an irregular sea.

Each curve stops when its height first reaches the smaller of 0.78h and 0.142L tanh(kh). These are approximate limits, not a calibrated surf-zone model. Nonlinear shape changes can matter before this boundary. No post-breaking height is calculated. If no crossing occurs above 0.5 m, the chart simply ends at the minimum depth shown.

Scientific basis and attribution

The energy-flux formulation is checked against Reeve, Chadwick & Fleming (2018), Coastal Engineering, §2.3.3, printed pp. 34–35, equation 2.20; Sorensen (2006), Basic Coastal Engineering, §4.2; and Holthuijsen (2007), Waves in Oceanic and Coastal Waters, §§7.3.1–7.3.2.

Text, code and chart are original. The equations express established wave theory; the source sections above explain their derivation. Free background overview: Coastal Wiki — Waves ↗.

CHAPTER 06 / WAVE BREAKING

When a wave
becomes surf.

The earlier experiments stop at a boundary. Here we examine what that boundary means, why its position is uncertain, and what changes when an organised wave begins to break.

01 Test a breaking-onset assumption02 Explore sensitivity03 Recognise breaker behaviour
BREAKING DEPTH
BREAKING HEIGHT
DISTANCE FROM SHORE

What changes at breaking?

A regular crest can no longer maintain its organised shape. Air is entrained, turbulence develops and wave energy is dissipated. Breaking also changes wave momentum flux, helping drive mean water-level changes and nearshore currents.

Before this point, real waves can already have narrow high crests, broad troughs and a steep front face. The sine waves in our earlier labs do not reproduce that nonlinear deformation.

Try this.

  1. Keep T = 8 s and γ = 0.78. Increase the input height and watch the estimated break move into deeper water.
  2. Restore H = 1 m. Compare γ = 0.60 and 0.90. A smaller assumed limit usually gives an earlier crossing in deeper water.
  3. Change the slope. In this simplified calculation, the breaking depth stays the same but its distance from the shoreline changes.

Real beach slope affects breaking behaviour too. Here it only converts depth into distance on a plane bed. That deliberate simplification makes the sensitivity experiment easy to interpret.

The two screening limits
Hlimit=min(γh,0.142Ltanh(kh))

The first term limits height relative to depth. The second is an approximate steepness limit associated with Miche. The lab searches from 50 m towards shore for the first intersection with the lossless, normal-incidence shoaling prediction.

This combination is a teaching rule. γ is not a universal constant, and combining two approximate limits does not create an exact breaking model.

Why we stop the height calculation here

Continuing the same lossless formula would send height towards unrealistically large values as depth approaches zero. In the surf zone, dissipation, nonlinear waves, rollers and mean flows become important. A separate model is needed to predict their evolution.

For irregular waves, different waves break at different locations; the surf-zone boundary is not a single permanent line. A criterion for individual regular-wave height should not be applied uncritically to significant wave height.

Scientific basis and model scope

Sorensen (2006), Basic Coastal Engineering, §2.8, printed pp. 38–41, discusses asymmetry, Miche’s approximate limit and breaker types. Reeve, Chadwick & Fleming (2018), Coastal Engineering, §2.6.2, printed pp. 61–65, discusses breaking, variable breaker indices and surf similarity. Holthuijsen (2007), §8.4.5, distinguishes depth-induced breaking parameterisations for irregular seas.

The predicted onset is approximate, based on linear shoaling with no currents, refraction, reflection or pre-breaking losses. The illustrated breaker shapes below are original qualitative drawings, not outputs of that model.

RECOGNISE / QUALITATIVE SHAPES

Three familiar ways a crest breaks.

Wave steepness and beach slope influence the behaviour. These are useful categories with gradual transitions, not three sharply separated states.

Spilling

Foam spreads down the face →

Foam develops near the crest and spreads down the front as the wave advances. Often associated with gentler beaches and relatively steep incident waves.

Plunging

Crest projects forwards →

The crest curls over and falls onto the water ahead, often enclosing air. Energy release is more concentrated than in a spilling breaker.

Surging

Uprush on a steep face →

Water surges up a steep boundary with limited development of an overturning crest. Reflection can be substantial. Collapsing breakers occupy part of the transition towards this behaviour.

Schematic snapshots only. The coral marks highlight the feature to notice; they are not a numerical foam or velocity field. No breaker type is assigned by the onset calculator.

CHAPTER 07 / SET-DOWN & SETUP

Waves pass.
The average level changes.

A crest is a temporary rise. Setup is a rise in the time-averaged water level. Follow the same beach from unbroken waves into an idealised surf zone, and discover why these are different.

01 Separate wave and mean02 Follow momentum flux03 Examine the assumptions

A · A shallow beach section

B · Magnify the mean-level change

C · Two budgets, different stories

Blue: energy flux F/FᵣCoral dashed: momentum flux Sₓₓ/Sₓₓ,ᵣ

Ratios are dimensionless; these are different physical quantities. Energy flux stays constant before breaking. Momentum flux grows before breaking, then decreases with depth-limited height in the surf zone.

New experiment: normal incidence; fixed period 20 s; offshore depth 4 m; mean level fixed to zero there. This is a shallow-water approximation, not a continuation of the 50 m full-dispersion labs.

The domain ends at h = 0.20 m. It does not reach the shoreline or predict run-up.

LOCAL MEAN LEVEL
TOTAL DEPTH D = h + η̄
MODEL BREAKING DEPTH hᵦ

Why can the mean surface slope?

Waves carry momentum as well as energy. Radiation stress is the excess, wave-averaged momentum flux; the name does not refer to electromagnetic radiation.

In this simplified cross-shore balance, a change in wave momentum flux is balanced by a pressure force associated with a sloping mean surface. Before breaking, increasing momentum flux accompanies set-down. Across the model surf zone, decreasing momentum flux accompanies a shoreward rise in mean level.

“Breaking pushes water up” is an incomplete explanation: the important quantity in this balance is the spatial gradient of momentum flux.

Predict, then experiment.

  1. Turn on mean level only. Does the raised mean disappear when individual crests are hidden?
  2. Increase the input height. Compare the breaking position, the set-down minimum and the nearshore mean level.
  3. Change slope while inspecting the same h. The level is unchanged here, but the distance over which it changes is different.

This is setup, not swash. We are calculating a steady average, not individual uprush events, shoreline motion, tide or storm surge.

Connect the picture to the balance

dSxxdx+ρgDdη¯dx=0

x increases shoreward; h is bed depth below the still-water datum; η̄ is mean surface displacement; D = h + η̄ is total mean water depth. Sₓₓ has units N/m and its gradient N/m².

E=ρgH28,Sxx≈32E,F≈EgD

These are normal-incidence, shallow-water approximations. Before breaking, F is constant. After breaking, we prescribe H = γD; F then decreases. The post-breaking height rule is a closure assumption, not an output of a resolved breaker simulation.

Go deeper: the piecewise solution

Let Dᵣ = 4 m and η̄ᵣ = 0. Before breaking:

H = Hᵣ(Dᵣ/D)1/4
η̄ = −(Hᵣ²√Dᵣ/16)(D−3/2 − Dᵣ−3/2).

Find Dᵦ from H(Dᵦ) = γDᵦ. In the surf zone define A = 3γ²/8, then η̄ = η̄ᵦ + A(Dᵦ − D). Combining D = h + η̄ gives dη̄/dx = [A/(1 + A)]/N on a plane slope 1:N. H, η̄ and Sₓₓ are continuous at the transition; their gradients need not be.

This analytic solution retains D in the hydrostatic balance. It does not resolve rollers, undertow momentum, bed shear, reflection, nonlinear wave shape or irregular breaking.

Boundary condition, validity and evidence

The offshore mean level is prescribed, so every displacement is relative to that finite-depth boundary. It is not an absolute water-level forecast. Fixing T = 20 s keeps h/L small throughout the domain; shallow-water kinematics still do not make linear wave shape accurate near breaking.

Verification checks cover continuity at breaking, the momentum-balance residual, pre-breaking flux conservation, decreasing surf-zone flux, the analytic surf-zone slope and the zero-wave limit. These are mathematical checks, not validation against measurements.

Scientific basis: Reeve, Chadwick & Fleming (2018), §2.6.3, printed pp. 65–67. The governing balance can also be read freely in SWAN’s technical documentation: wave-induced setup ↗. SWAN is a more general spectral model; this lab does not reproduce its full implementation. Text, diagrams and implementation here are original.

Check your understanding

Is setup the height of a passing crest? No: it is a mean-level change measured relative to a stated reference.

Does constant energy flux imply constant momentum flux? No. Panel C shows why set-down can occur even before energy is dissipated.

Can this model tell us how far a wave runs up the beach? No. That requires shoreline and swash dynamics, outside this chapter.

CHAPTER 08 / FROM REGULAR TO IRREGULAR WAVES

One wave is simple.
Real seas are a sum.

Our earlier experiments followed one idealised, repeating wave. Now imagine a fixed wave gauge offshore, recording the surface as several oscillations arrive together. Add them one at a time and watch how a more complicated time series emerges.

01 Add one frequency02 See interference and groups03 Add more components and phases
ONE EXPERIMENT / TWO SYNCHRONISED VIEWS

A · Individual ingredients

Each coloured line is one component at the same fixed sensor. All lines use the same amplitude scale, ±0.6 m; they are not stacked vertically.

B · The surface-elevation record

Sum η(t)Zero referenceDashed group envelope (two-wave stage only)Time marker

Fixed vertical scale, ±1.75 m, across all stages: the increased range is real superposition, not axis rescaling. The dashed envelope appears with exactly two equal-amplitude frequencies. The marker can be moved with the time slider or by clicking either graph.

One regular wave: the same amplitude and period repeat throughout the record.

ACTIVE COMPONENTS1 / 5
SURFACE ELEVATION AT MARKER+0.45 m
A + B BEAT INTERVAL—

All plots are deterministic, 150-second synthetic examples; colours identify components, not measured wave systems. A surface-elevation record at one point contains time/frequency information but, by itself, does not determine wave direction.

From a repeated rhythm to a changing pattern

One component has one frequency, one amplitude and a fixed phase. Its crest-to-trough height is H = 2a, and its shape repeats every T = 1/f seconds.

Two nearby frequencies can add when their crests align and partially cancel when their phases oppose. The sum forms alternating stronger and weaker sections: a temporal wave group or beat pattern at the sensor. The individual frequencies do not change when you combine them.

More frequencies and phases create a less repetitive record. A short sequence can appear irregular even though the five ingredients here are deterministic. An actual sea state involves many interacting components and is commonly described statistically.

Try to predict what changes.

  1. Add B and inspect the orange dashed envelope. Does it describe every individual crest, or the slower changing range of crest heights?
  2. Set B’s phase to 180°. At t = 0, do A and B now reinforce or cancel? What stays the same about their periods?
  3. Add C, D and E. Can you still assign one unambiguous period or crest-to-trough height to the whole 150 s record?

This chapter is about linear superposition at a point. A wave group in time is not itself a prediction of group velocity across space. We do not estimate significant wave height or build a spectral density here.

Look under the surface of the mathematics

Go deeper: write each wave and add the components
η(t)=∑n=1Nancos⁡(2πfnt+φn)

η(t) is elevation in metres at the fixed sensor; an is amplitude (m), fn frequency (Hz), φn phase (radians in the calculation), t time (s), and N the number of components. The component phases in the controls are displayed in degrees.

This is a linear, first-order construction: elevations add algebraically. Positive and negative values can cancel. Fixed-frequency, fixed-phase components do not represent nonlinear wave interaction or stochastic forcing.

Go deeper: derive the two-frequency group envelope

Waves A and B have equal amplitudes a = 0.45 m, fA = 0.100 Hz and fB = 0.120 Hz. Define Δf = fB − fA and Δφ = φB − φA.

ηA+ηB=2acos⁡(πΔft+Δφ2)cos⁡(2πf¯t+φ¯)

The rapidly oscillating carrier has mean frequency f̄ = (fA + fB)/2 = 0.110 Hz. The slowly varying amplitude envelope is ±2a|cos(πΔf t + Δφ/2)|. Successive strong-group maxima are 1/|Δf| = 50 s apart. A phase shift moves the envelope in time but does not change that interval.

At the default phase of 0°, the two components sum to +0.90 m at t = 0. At 180°, their elevations cancel at t = 0. The dashed curves are amplitude envelopes, not separate physical water surfaces.

Applicability, evidence and the route to spectra

The graph displays a 150 s record at one location, sampled densely for drawing. No random phases, measured buoy data, frequency-bin estimation, directional information, finite-depth propagation or breaking are included. The reference level is η = 0; this should not be confused with the nonzero mean setup calculated in Chapter 7.

The construction follows the harmonic-superposition description in Holthuijsen (2007), Waves in Oceanic and Coastal Waters, §3.5.1–3.5.3 (printed pp. 31–36), and Stewart (2008), Introduction to Physical Oceanography, §16.3 (printed pp. 278–284). The component values, graphics and interaction are original teaching examples, not reproduced figures.

Next: Chapter 9 first defines crossing-based waves and their statistics. Chapters 10 and 11 develop frequency analysis and spectral shapes; Chapter 12 adds direction, Chapter 13 explores realizations and subsurface response, and Chapter 14 follows ideal spectral shoaling. A single point-elevation trace alone cannot uniquely identify directional spreading.

CHAPTER 09 / WAVE-BY-WAVE STATISTICS

A sea state contains
more than one wave height.

Real waves vary in height, period and direction. A characteristic height summarizes a sea state; it does not describe every wave within it. Begin by dividing an elevation record into complete waves, then ask what their statistics reveal—and what they conceal.

01 Define a wave02 Rank its height03 Compare height and period04 Interpret probability

Wind sea, swell and mixed conditions

Wind sea receives energy from the local wind. Swell has propagated away from its generating weather system and often occupies a narrower range of frequencies and directions. Neither is perfectly monochromatic. A mixed sea can contain several systems at once, so a single characteristic period may hide important structure.

Two complementary descriptions

Wave-by-wave analysis identifies events between selected crossings and retains their individual heights and periods. Spectral analysis distributes variance over frequency and, when additional measurements are available, direction. An individual crossing-defined wave can be the local sum of several travelling components; it is not necessarily one identifiable wave system.

A · Define complete waves in the record

A 150 s synthetic A–E record, sampled every 0.05 s and demeaned using those samples. Shading identifies one complete wave; partial waves at the ends are excluded. Orange markers locate its crossing boundaries.

B · Rank the heights

Coral bars form the highest third, rounded up to a whole number of waves in this small sample. A rank is not the time order of a wave.

C · Keep height and period paired

A height is a vertical difference.

Between two successive crossings in the same direction, find the highest elevation and lowest elevation. Their difference is Hᵢ; the elapsed crossing-to-crossing time is Tᵢ. Several local crests may lie inside one crossing-defined wave. Counting every local maximum instead would answer a different question.

Which average answers the question?

The arithmetic mean H̄ describes a typical height averaged over all waves. The root-mean-square height Hrms gives more weight to large waves. H1/3 averages the highest third and H1/10 the highest tenth. Hmax is the largest observed height, so its value depends strongly on record length. None is automatically the height required for an engineering design.

Definitions and a short worked example

H̄ = ΣHᵢ/N;   Hrms = √(ΣHᵢ²/N);   T̄z = ΣTᵢ/N

For sorted heights 3.0, 2.4, 2.0, 1.6, 1.2 and 0.8 m, H̄ = 1.833 m, Hrms = 1.975 m and H1/3 = (3.0+2.4)/2 = 2.7 m. The maximum is 3.0 m. Averaging all six heights cannot produce H1/3.

For complete crossing intervals, T̄z equals (last crossing time − first crossing time)/N. Dividing the entire recording duration by the number of complete waves incorrectly includes the partial end intervals.

Elevation, height and probability are different quantities.

A linear sea made from many sufficiently independent components is often approximated by Gaussian surface elevation. Elevation may be positive or negative. Individual wave height is non-negative; for a narrow-band linear Gaussian sea, its distribution is approximately Rayleigh. A Gaussian elevation distribution does not by itself make every broad-band or breaking-wave height distribution Rayleigh.

Stationarity, ergodicity and covariance

Weak stationarity requires a constant mean and variance and covariance depending on time lag rather than absolute time. A slowly changing storm may be approximately stationary over a suitable short record, but the assumption must be checked. Ergodicity is the additional condition that long time averages can represent ensemble statistics; one short record does not prove it. Spatial homogeneity is a separate assumption and is not required merely to define a spectrum at one fixed station.

Cη(τ) = E[(η(t)−μ)(η(t+τ)−μ)]

For a stationary real process: Cη(τ) = ∫₀∞S(f) cos(2πfτ) df; Cη(0) = m₀

Covariance uses departures from the mean; uncentred correlation equals covariance only when the mean is zero. These statistical assumptions make a record informative about a sea state, but they do not turn its largest wave into a universal maximum.

Height–period dependence and shallow-water limitations

The scatterplot retains paired Hᵢ and Tᵢ values. In real records their joint density p(H,T) generally cannot be replaced by p(H)p(T) without testing independence. Marginal height or period distributions alone lose this pairing. Nonlinearity, wave breaking and finite depth can alter both distributions.

Near breaking, H1/3 and Hm0 can differ systematically as wave shape changes. Their ratio is not a universal depth correction. Report the measured or modelled definition explicitly rather than relabelling every height as Hs.

Probability within one idealized sea state

The spectral height Hm0 = 4√m₀ is defined from elevation variance m₀ and developed in Chapter 10. For the narrow-band Rayleigh approximation, set Hrms = Hm0/√2. The probability of a single wave exceeding h is Q(h) = exp[−2(h/Hm0)²]. The calculation below illustrates this model; it is not a fit to the five-component record above.

Single-wave exceedance versus the largest wave

For independent heights, P(Hmax > h) = 1 − [1−Q(h)]N. Setting NQ(h) = 1 gives h/Hm0 = √(ln N/2), a useful “one expected exceedance” level, not a guaranteed maximum or the exact expected maximum. Real grouping violates independence and changes maximum statistics.

At h = Hm0, Q = e−2 ≈ 13.5%. H1/3 is the mean of the largest third, not the threshold that exactly one third exceeds. Under Rayleigh assumptions that threshold is Hrms√ln3, and H1/10 is about 1.27 H1/3; these are model relationships rather than identities for every record.

Short-term exceedance concerns individual waves within a specified sea state. A 100-year return-level sea state concerns the long-term distribution of sea-state conditions. The two probabilities must not be substituted for each other.

CHAPTER 10 / FREQUENCY SPECTRA & WAVE STATISTICS

From the time record
to the frequency spectrum.

A wave gauge records surface elevation through time. Fourier analysis asks which rhythms are present in that record. Follow one signal from sampled elevations to complex Fourier coefficients, then convert their squared magnitudes into a spectrum whose area measures variance.

01 Observe η(t)02 Resolve frequencies and phase03 Estimate S(f)04 Interpret weighted areas
ONE EXPERIMENT / TIME AND FREQUENCY

A · Surface elevation at a fixed sensor

The dark line joins samples taken every 0.05 s; the coloured line isolates the selected known component. The vertical scale stays at ±1.75 m. At the default A-only phase, the crest, trough and 0.90 m crest-to-trough bracket appear on this graph. Optional orange zero-upcrossing lines separate complete waves for H1/3. This is a synthetic record.

B · Resolve the signal into sinusoidal components

C · FFT periodogram and the prescribed reference

Tp · peak10.00 s Tm01 · m₀/m₁10.00 s Tm02 · √(m₀/m₂)10.00 s

The vertical markers sit at the inverse of each labelled period. Only Tp marks a highest component bar; the other two indicate moment-derived equivalent frequencies, not additional wave components. Each bar's area = a²/2 (m²), with fixed illustrative Δf = 0.010 Hz and S(f) in m²/Hz.

Turn on the FFT curve to see a finite-record estimate. The bars and reported Hm0/periods remain based on the exact prescribed components.

One component: one frequency bin and a perfectly repeating record.

SPECTRAL VARIANCE m₀0.1013 m²
SPECTRAL HEIGHT Hm01.27 m
RECORD HEIGHT H1/30.90 m
PEAK PERIOD Tp10.00 s
MEAN PERIOD Tm0110.00 s
MEAN ZERO-CROSSING ESTIMATE Tm0210.00 s

This independent component experiment uses the prescribed zero level for crossings; Chapter 9 instead removes the finite sample mean before segmenting its record. The height Hm0 and periods are calculated from the prescribed variance distribution. H1/3 comes from complete individual waves in this finite sampled trace. Neither is a field observation. Tm02 is a spectral estimate, not the arithmetic mean of detected crossing periods.

How was 0.90 m obtained?

For wave A, the crest is +0.45 m and the trough is −0.45 m: H = +0.45 − (−0.45) = 0.90 m.

What is the peak period?

The strongest bin is at 0.100 Hz; Tp = 1 / 0.100 = 10.00 s.

What are the mean periods?

First define the spectral moments: m₀ = ∫S(f)df (total variance, the area under the spectrum), m₁ = ∫fS(f)df and m₂ = ∫f²S(f)df. Then Tm01 = m₀/m₁ and Tm02 = √(m₀/m₂). They use the whole spectrum, not just the highest bar.

ANALYSIS / FROM THE RECORD TO THE COEFFICIENTS

An FFT is a calculation, not yet a spectrum.

The discrete Fourier transform compares the sampled signal with sine and cosine patterns at a set of frequencies. A complex coefficient records these two projections together: its magnitude describes the strength of the matching pattern and its argument describes phase. The fast Fourier transform is an efficient algorithm for computing exactly this transform. Panel B shows the known ingredients so that the result can be checked; the coefficient readout below it is calculated from the sampled signal, not copied from those ingredients.

To estimate variance density, first remove the sample mean, optionally taper the record, and then square and normalize the Fourier magnitudes. Combine matching positive and negative frequencies into a one-sided spectrum. Phase is absent from this density: two records can have the same underlying spectral variance but different arrangements of crests. For a short record, changing phase can also change leakage and its measured variance.

Read the spectrum as an area.

Variance is the average squared departure from the mean elevation. It measures the size of the fluctuations in m², not a crest-to-trough distance. In a spectrum, density times frequency width gives variance. A harmonic of amplitude 0.45 m contributes 0.45²/2 = 0.10125 m². Drawing that contribution in an illustrative 0.010 Hz-wide bin gives 10.125 m²/Hz. A narrower drawing bin would be taller without adding variance.

Worked moments · exact known components

The zeroth moment m₀ counts every contribution equally. The first moment m₁ weights it by frequency; m₁/m₀ is the variance-weighted mean frequency. The second moment m₂ weights it by frequency squared, making it more sensitive to rapid oscillations and high-frequency noise. Thus Tm01 is the inverse mean frequency, whereas Tm02 is the inverse root-mean-square frequency. The latter relates to the mean interval between like-sense zero crossings for a stationary, differentiable Gaussian process; it is not an identity for a short deterministic record.

Derivation: coefficients, one-sided normalization and Parseval

Xk = Σn=0N−1 wn(ηn − η̄) exp(−i2πkn/M)

fk = k/(MΔt),   U = (1/N)Σwn²

Ŝk = ck Δt |Xk|²/(NU)

Σk=0M/2Ŝk/(MΔt) = Σwn²(ηn−η̄)²/(NU)

Here ck = 2 for interior positive frequencies, and 1 at DC and Nyquist. N counts observed samples; M = 16,384 includes appended zeros. Rectangular weights are w = 1; the symmetric Hann taper is wn = [1 − cos(2πn/(N−1))]/2. Dividing by U corrects the taper's mean-square gain. For a rectangular window, the integral equals the demeaned sample variance with divisor N. For Hann it equals the window-weighted mean square divided by U, which need not equal the unweighted variance of this particular record. Mean removal before tapering does not guarantee zero DC after tapering.

For an unpadded rectangular DFT and an isolated, bin-centred sinusoid, a = 2|Xk|/N and φ = arg(Xk). Off-bin components, overlapping leakage and tapering require more care. A coefficient at a nearby padded bin is therefore not labelled as an exactly recovered physical component. All FFT-integral checks include the entire 0–10 Hz band; the figure zooms to 0.06–0.22 Hz.

Derivation: spectral moments and the illustrative bins

mn = ∫₀∞ fnS(f) df ≈ Σ fknSkΔf

Known discrete lines: mn = Σ finai²/2

Hm0 = 4√m₀;   Tm01 = m₀/m₁;   Tm02 = √(m₀/m₂)

The moment readouts use the exact line frequencies. The coloured 0.010 Hz rectangles are an area-preserving illustration, not an inferred bandwidth or an FFT grid. Integrating f² over a genuinely flat rectangle would add vΔf²/12 to its second moment; that artificial broadening is deliberately excluded. Period guides are placed at 1/T on the frequency axis and are not extra spectral peaks.

What can a finite record tell us?

Sampling and aliasing

At Δt = 0.05 s, the sampling frequency is 20 Hz and the Nyquist frequency is 10 Hz. Frequencies above Nyquist can masquerade as lower ones: a 19.9 Hz cosine sampled at 20 Hz has the same samples as a 0.1 Hz cosine of the same phase when that phase is zero. An analogue anti-alias filter must suppress unwanted high frequencies before sampling; an FFT cannot undo this ambiguity.

Duration and leakage

A 150 s observation has nominal Fourier spacing 1/150 ≈ 0.006667 Hz. A component that does not complete an integer number of cycles spreads across Fourier bins. Even a bin-centred sinusoid has sidelobes between those bins, visible on a zero-padded plot. Longer records narrow the spectral response, provided the sea state remains sufficiently stationary.

Windows and zero-padding

A Hann taper reduces the influence of the record's ends. It suppresses distant sidelobes but broadens the main peak, trading separation of nearby frequencies for reduced leakage. Appending zeros gives a finer plotting interval, here 20/16,384 ≈ 0.001221 Hz. It interpolates the same finite-record transform and creates no new observations or resolving power.

Uncertainty and changing seas

A raw periodogram of a random sea is a noisy estimate, even when the transform is exact. Averaging spectra from segments reduces sampling variability but sacrifices frequency resolution. Longer records can mix changing conditions, so duration, averaging and stationarity must be considered together. This five-component experiment demonstrates analysis errors and finite-record effects; it is not a stochastic confidence-interval model.

Experiment: does phase change a component's variance?

Add B, select it and change its phase to 180°. The record changes and the known variance bars stay fixed. Finite-record cross terms can change the estimated periodogram. For A and B alone over the default 150 s, both complete integer cycles, so their rectangular-window integrated variance remains 0.20250 m².

Experiment: where does leakage go when the record gets longer?

Add all five components and compare 150 s with 300 s. The nominal spacing halves, but not every component becomes bin-centred. Change to Hann and observe lower distant sidelobes with broader peaks. The reference moments do not change because the component amplitudes and frequencies have not changed.

Two lenses on the same motion.

Time domain: a zero-upcrossing analysis separates the record into individual waves. Each complete wave has a crest-to-trough height and a period. H1/3 is the average height of the highest third of those complete waves (top third rounded up for this short exercise).

Frequency domain: a variance-density spectrum describes how surface-elevation variance is distributed over frequency. The integral under S(f) is m₀, with units m². The widely used spectral significant wave height is Hm0 = 4√m₀.

Different period summaries: Tp is the inverse of the frequency at the largest spectral bin; Tm01 and Tm02 weight the whole spectrum, so they respond to changing frequency composition. Equal strongest bins produce a peak-frequency tie; this demonstration reports the lowest tied frequency.

Hm0 and H1/3 are not interchangeable identities. The former is derived from variance and commonly relates to significant heights under random-sea assumptions; the latter is a direct sample statistic. In a narrow-band linear Gaussian sea, the Rayleigh approximation brings the two close; breaking, nonlinearity and finite samples can change that relationship. Their difference is particularly clear for a single deterministic sinusoid, whose H1/3 is exactly 2a while Hm0 = 4a/√2.

Questions to test the model

  1. Start with A. Why is Hm0 larger than its single-wave height 2a?
  2. Add B. Which quantities increase, and why do we now have two equally high frequency bars?
  3. Add all five. Why do Tm01 and Tm02 respond even though the strongest frequency bin has not changed?

Go deeper: from bins to moments

How is a harmonic's variance converted into spectral density?
vi=ai22,Si=viΔf

A sinusoid with amplitude ai has variance vi = ai²/2 in m² (averaged over complete cycles). A hypothetical flat bin of width Δf = 0.010 Hz has density Si = vi/Δf in m²/Hz. Changing a bin width changes the density height but not its area or variance.

The model sums exact known harmonic contributions. Computing a spectrum from an observed record instead requires frequency analysis, finite-record treatment and averaging. The chosen 0.010 Hz bins should not be mistaken for the resolution or output of an FFT of this trace.

Why do we integrate spectral moments?
mn=∫0∞fnS(f)df≈∑finviHm0=4m0,Tm01=m0m1,Tm02=m0m2

m0 is total elevation variance. m1 and m2 shift the weighting toward higher frequencies. Here the exact discrete-harmonic expression Σfinvi supplies each moment. Tp = 1/fp is reported separately. These are frequency-based, not angular-frequency-based, moments.

Why can a short record disagree with spectral Hm0?

Because the fixed-phase five-wave record is deterministic and finite. Its beginning and ending waves may be incomplete, its measured heights depend on the phase pattern, and the prescribed harmonic variances represent component-wise long-time contributions rather than a finite-record variance estimator. The usual near-equivalence between Hm0 and significant individual-wave heights is a statistical approximation under appropriate random-sea conditions, not a mathematical identity.

CHAPTER 11 / SPECTRAL SHAPES & MODELS

Equal wave height.
Different spectral shapes.

A spectrum contains more information than its total variance. Two sea states can share Hm0 while distributing that variance across very different frequencies. Parametric spectra describe useful families of shapes; they do not replace measurements or a wave model.

The area is held at m₀ = 0.36 m², so Hm0 = 2.40 m. The reference frequency fp = 0.125 Hz belongs to the underlying deep-water shape. Finite-depth weighting can move the sampled maximum. The computation uses 0.020–0.800 Hz; the chart shows 0.02–0.50 Hz.

What the families represent

The original Pierson–Moskowitz spectrum describes a fully developed deep-water wind sea under restrictive wind conditions. A two-parameter PM-shaped or Bretschneider form specifies height and peak period independently. JONSWAP adds a peak-enhancement factor to represent more sharply peaked developing seas. γ = 1 removes that enhancement.

TMA multiplies a JONSWAP form by a finite-depth equilibrium factor. It describes a family of finite-depth spectra; applying it is not equivalent to transporting an offshore spectrum shoreward with conserved energy flux. Chapter 14 examines that separate transformation problem.

A mixed swell and wind sea often needs multiple partitions. Ochi–Hubble is one two-component parametric family, but it is not the only way to represent two peaks. More than one directional system can also share a single frequency peak, so the number of peaks does not uniquely count wave systems.

Shape equations and the normalization used here

B(f) = f−5 exp[−(5/4)(fp/f)4]

J(f) = B(f) γr(f),   r(f) = exp[−(f/fp−1)²/(2σ²)]

σ = 0.07 for f ≤ fp; σ = 0.09 for f > fp

Sᵢ = (Hm0/4)² qᵢ / Σ(qⱼ Δf)

Choose q = B, J or JΦ. This explicit finite-grid normalization preserves the selected variance. It does not estimate the original Phillips coefficient α from wind or fetch. All densities are per hertz.

Bandwidth is a definition, not just a visual impression.

The coefficient ν = √(m₀m₂/m₁² − 1) is the coefficient of variation of frequency when variance is used as the weight. It is zero for a single spectral line and increases as frequency content spreads. It is non-negative but is not universally bounded by one. Distinct bandwidth measures need not rank spectra identically.

Higher moments, peakedness and frequency conventions

ε = √[1 − m₂²/(m₀m₄)]

Qp = (2/m₀²) ∫₀∞ f S(f)² df

ω = 2πf;   Sω(ω) = Sf(ω/2π)/(2π);   mn,ω = (2π)nmn,f

ε lies between zero and one when the required finite moments exist. An ideal f−5 tail has a divergent fourth moment if extended indefinitely, so m₄ and ε from a finite grid depend strongly on the upper cutoff. Qp measures concentration differently; it involves the square of the density, not merely m₁. The readouts above deliberately report ν and the integration limits instead of presenting a cutoff-sensitive ε as a universal sea-state property.

With moments in Hz, Tm01 = m₀/m₁ and Tm02 = √(m₀/m₂). With angular-frequency moments, both expressions require a factor 2π. A formula cannot be transferred between conventions without changing its density and moments.

The finite-depth equilibrium factor

Φ(f,d) = tanh²(kd) / [1 + 2kd/sinh(2kd)]

(2πf)² = gk tanh(kd)

This factor follows from the finite/deep ratio of k−3dk/dω. It tends to one in deep water and to ω²d/(2g) as kd tends to zero. The numerical solution uses the exact linear dispersion relation. The chosen normalization subsequently rescales JΦ; it does not conserve offshore energy flux or simulate breaking.

CHAPTER 12 / DIRECTIONAL SPECTRA

Not just how fast waves oscillate.
Which way do they travel?

Two sea states can share exactly the same frequency spectrum and spectral wave height, yet look completely different across the water. Spread the variance across directions, then split it into crossing swell and wind-sea systems. Compare three representations of one two-dimensional spectrum.

01 Hold S(f) fixed02 Change directional spreading03 Separate crossing systems
ONE EXPERIMENT / THREE VIEWS

A · Frequency–direction variance density S(f, θ)

S(f, θ) · m²/(Hz·rad)032 · fixed scale

Horizontal: frequency (Hz). Vertical: propagation direction θ in degrees, plotted on a 0° to 360° compass-style axis, clockwise from north. 360° is the same direction as 0°, so a lobe close to north can wrap across the top and bottom edges of the plot. Colour represents S(f, θ) (m²/Hz/rad); the colour scale stays fixed at 0–32 m²/(Hz·rad) across scenarios. Markers identify the mean directions of the two modelled systems.

B · Directional rose (integrated over frequency)

Each sector spans exactly 5°. Its radius is the angular marginal G(θ) = ∫S(f, θ)df in m²/rad, on a fixed 0–1.5 scale. Radius, not wedge area, encodes density. Thin coloured arrows identify the prescribed system means; they do not measure density.

C · The unchanged frequency marginal S(f)

Blue = swell partition; coral = wind-sea partition; dark curve = sum. Integrating S(f, θ) over all directions recovers this identical frequency spectrum in every preset.

TOTAL VARIANCE m₀0.480 m²
SPECTRAL HEIGHT Hm02.77 m
SYSTEM PEAK PERIODS10.0 / 6.25 s
MEAN-DIRECTION SEPARATION0°

Illustrative two-partition spectrum: Gaussian frequency shapes (discretely normalized), wrapped Gaussian direction shapes (discretely normalized), 0.005 Hz × 5° grid. Angle is the direction of travel on a compass-style bearing axis (0° north, 90° east, 180° south, 270° west), not the nautical “coming from” convention. Radians are used in integration and in the S(f, θ) units.

Read one cell, then sum it in two ways.

A cell contributes Δm₀ ≈ S(f, θ) × 0.005 Hz × (5π/180 rad). Sum across direction to obtain the frequency marginal S(f) in m²/Hz; sum across frequency to obtain the angular marginal G(θ) in m²/rad. Integrating either marginal gives 0.48 m². Spreading the same variance over more angles lowers the density peaks without changing Hm0.

Three views, three different questions

A frequency–direction heatmap reveals which periods travel in which directions. Two separate patches can indicate crossing swell and wind sea, even when the total variance is unchanged.

A directional rose collapses all frequencies into one angular distribution. It helps show whether waves propagate toward one dominant direction or several—but it cannot tell which direction belongs to the longer-period system.

A frequency marginal collapses directions instead. It reports variance at each frequency but cannot tell whether the corresponding waves travel together or cross at different angles.

A single-point surface-elevation time series cannot uniquely resolve directions. Directional instruments combine additional spatial, slope, velocity or other information. Directional wave modelling likewise tracks variance or action across frequency and direction, subject to physical assumptions and source terms.

Try a thought experiment

  1. Choose Aligned, then Broad. What stays identical on the frequency-marginal plot?
  2. Choose Crossing seas and increase the separation. Which plot best identifies the two distinct wave systems?
  3. Increase σ. Why does the peak colour weaken even though the total spectral variance and Hm0 are fixed?

What is a spectral partition?

A partition groups a spectral system by its frequency and direction. Here swell and wind sea are prescribed Gaussian frequency shapes, with variances 0.30 and 0.18 m². Those names identify teaching components rather than an automatic classification: assigning observed wind sea normally also requires wind information. Overlapping systems are not always uniquely separable from spectral shape alone.

From bearings to the plot, and from “toward” to “from”

A compass bearing θ increases clockwise from north. Canvas coordinates are x = cₓ + r sin θ and y = cᵧ − r cos θ. The corresponding Cartesian angle measured counterclockwise from east is α = 90° − θ, modulo 360°. A coming-from bearing is (θ + 180°) modulo 360°: waves travelling toward 320° come from 140°.

The heatmap has 72 cells covering [0°, 360°), centred at 2.5°, 7.5°, …, 357.5°. Its bottom 360° edge joins its top 0° edge; there is no extra 360° cell. A lobe crossing north appears split only because a circular coordinate has been laid flat. The rose joins the same two edges naturally.

Directional density is not directional variance

G(θ) is variance per radian; its integral is elevation variance in m². It is not the statistical variance of the angles themselves. Bearings are circular: 359° and 1° are close neighbours, so ordinary arithmetic averages and variances can mislead. The width σ here sets a wrapped-Gaussian shape. The model normalizes the sampled density so ΣDjΔθ = 1.

A density per degree is a different numerical quantity: Sdegree = Sradianπ/180. Using a per-radian density with 5 instead of 5π/180 in an integral would overestimate variance by 180/π.

Why one elevation sensor cannot identify direction

At one location, waves with the same frequency can produce the same elevation history regardless of bearing. A directional buoy also measures horizontal motion or slopes; arrays measure spatial phase differences. Cross-spectra between such channels constrain directional moments. Reconstructing a full directional distribution still involves assumptions, finite angular resolution and possible ambiguities, especially for crossing or reflected systems.

Go deeper: how the directional spectrum works

Conservation of variance when spreading over direction
S(f,θ)=∑Sp(f)Dp(θ)∫02πDp(θ)dθ=1,S(f)=∫02πS(f,θ)dθ

Each partition p has its own frequency distribution Sp(f) and normalized directional spreading Dp(θ). Because ∫Dpdθ = 1, spreading changes the location of variance in direction, not its total. Integrating over frequency and direction gives m0 = 0.48 m² in every case.

Directional bins, units and coordinate conventions
Δm0≈S(f,θ)ΔfΔθ

Here Δf = 0.005 Hz and Δθ = 5π/180 rad. Therefore S(f, θ) has units m²/(Hz·rad). Plotted angular labels are in degrees for accessibility; a radian conversion is needed to integrate densities expressed per radian. In real applications document whether a directional angle indicates travel toward or coming from. This chapter uses travel toward, with degrees increasing clockwise from north, while the density and integrals still use radians internally.

What this model does and does not predict

The two partition shapes are idealized and prescribed rather than inferred from a wave buoy. The plots do not include geographical propagation, finite-depth dispersion, nonlinear interaction, refraction, diffraction, currents or source/sink terms. A genuine directional spectral wave model solves an energy or action balance and evolves the spectrum through space and time; this is a controlled visualization of spectral representation alone.

CHAPTER 13 / REALIZATIONS, GROUPS & SUBSURFACE RESPONSE

The same spectrum
can make different records.

A spectrum specifies how variance is distributed, but it does not fix when the largest crest will occur. Build different phase realizations of one target spectrum, then follow the same components below the surface using linear pressure and velocity transfer functions.

A · Surface elevation and pressure-equivalent elevation

Navy: η(t). Blue: p′(t)/(ρg), in metres. The plot shows the first 128 s of a 256 s periodic numerical realization. Pressure is the oscillating part only; hydrostatic pressure is excluded.

B · Horizontal orbital velocity

C · Runs of larger waves

The threshold is this realization's H1/3. Adjacent coral bars form a run. A run is a sequence statistic, not a separate wave component or a prediction of group speed.

From variance bins to amplitudes

A frequency bin containing variance SᵢΔf receives a sinusoidal amplitude aᵢ = √(2SᵢΔf). Choose a phase φᵢ uniformly over 0–2π and sum aᵢ cos(2πfᵢt+φᵢ). A component's crest-to-trough height is 2aᵢ; using that height as the cosine coefficient would double the elevation and quadruple the variance.

Fixed amplitudes versus Gaussian coefficients

η(t) = Σ√(2SᵢΔf) cos(2πfᵢt+φᵢ)

Alternative: η(t) = Σ[Aᵢ cos(2πfᵢt) + Bᵢ sin(2πfᵢt)]

Aᵢ, Bᵢ independent, each N(0, SᵢΔf)

The experiment uses fixed amplitudes and pseudo-random phases. With many components its elevation distribution can approach Gaussian, but exact Gaussianity is not guaranteed. Independent Gaussian coefficients give a Gaussian process even for a finite sum, although individual realizations have random component amplitudes and variances. In both constructions the expected variance is ΣSᵢΔf.

Uniformly spaced frequencies make a simulated record periodic with repetition interval 1/Δf. For this 256 s record and integer Fourier bins, the sampled mean-square elevation equals the target m₀ = 0.25 m² to numerical precision. That deliberately convenient orthogonality does not apply to every finite observation.

Grouping depends on order.

A height histogram ignores which waves occur next to one another. Runs above a threshold, lagged height correlation, and a smoothed squared-elevation history offer complementary descriptions of grouping. A narrow frequency band often produces a slowly changing envelope, but a spectrum alone does not prescribe the exact sequence of large waves in a particular realization.

A transparent grouping measure

The plot counts consecutive heights above the sample H1/3 and reports the longest such run. This is a threshold-dependent descriptor. It is not the standardized SIWEH groupiness factor, and its short-record variability should not be interpreted as a fitted ocean statistic. A correlation of consecutive heights tests a different aspect of grouping and must state its estimator and sample size.

Different frequencies penetrate differently.

At depth z ≤ 0, linear theory attenuates pressure with Kp = cosh[k(z+d)]/cosh(kd). Each frequency has its own k, so the pressure record is not generally a single scaled copy of the surface record. Horizontal orbital velocity uses Ku = ω cosh[k(z+d)]/sinh(kd), with units s⁻¹.

Transfer functions and response spectra

p′(t)/(ρg) = ΣKpᵢ aᵢ cos(2πfᵢt+φᵢ)

u(t) = ΣKuᵢ aᵢ cos(2πfᵢt+φᵢ)

Sp′(f) = (ρg Kp)² Sη(f)

Su(f) = Ku² Sη(f)

These are co-located, unidirectional, linear responses. Pressure density has units Pa²/Hz; velocity density has units (m/s)²/Hz. In a directional sea, an x-velocity component also needs a directional projection. Multiplying one representative wave height by a single factor cannot recover all frequency-dependent response statistics.

At z = 0, pressure-equivalent elevation matches η in the linearized boundary condition. Moving deeper preferentially removes shorter-period pressure fluctuations. Recovering surface variance from a deeply submerged pressure sensor can amplify noise where Kp is small.

Neither the target spectrum nor this superposition predicts breaking, phase-coupled nonlinear waves, loads on a specific structure, or wave-driven currents. A realization illustrates variability; it is not a forecast of future crest timing.

Follow the wave below the surface: orbital velocity and the near-bed layer

Linear theory predicts the outer-flow horizontal and vertical orbital-velocity amplitudes at height zb above the bed, with 0 ≤ zb ≤ h:

Û(zb) = (H/2)ω cosh(kzb)/sinh(kh);   Ŵ(zb) = (H/2)ω sinh(kzb)/sinh(kh)

ω² = gk tanh(kh),   ω = 2π/T;   Ub = Û(0) = (H/2)ω/sinh(kh)

Here H is regular-wave height (m), T period (s), h depth (m), k wavenumber (m−1) and amplitudes Û, Ŵ and Ub are in m s−1. The irrotational extrapolation Ub is the free-stream amplitude just outside the thin bed boundary layer, not the actual velocity exactly at a no-slip bed. Vertical orbital velocity vanishes at the bed in this ideal linear model.

Blue: horizontal orbital amplitude Û; coral: vertical orbital amplitude Ŵ. Plot is linear, unidirectional, non-breaking wave kinematics; it is not a velocity time history.

Inset resolves the very thin ideal smooth-bed laminar oscillatory layer on a separate millimetre-scale axis. It is not drawn to scale against the water-column profile above.

Bed outer-flow amplitude Ub—
Orbital excursion amplitude Ub/ω—
Ideal laminar layer scale δν—

For an ideal smooth, no-slip bed with laminar oscillatory motion, δν = √(2ν/ω) is a viscous penetration scale (m), where ν is molecular kinematic viscosity (m² s−1). The ideal harmonic boundary-layer amplitude ratio is |1 − e−(1+i)y/δν| for vertical distance y ≥ 0 above the bed. Its near-bed velocity reaches zero at the wall. We use ν = 1.1 × 10−6 m² s−1 as an illustrative constant; real turbulent and rough-bed wave layers are not described by this curve.

Engineering limitation: Linear potential flow does not predict actual bed shear stress. The ideal laminar model is for understanding scale separation—not for estimating environmental drag, wave-driven currents or design loads. Turbulence, roughness, directionality and combined forcing require an independently validated boundary-layer formulation. No transport or seabed-evolution prediction is made.

Scientific sources: Bosboom & Stive, Coastal Dynamics (2023), §§5.4.1 and 5.4.3 (orbital velocity, boundary layer); Holthuijsen (2007), §5.4. These are new original computational figures and an idealized Stokes-layer demonstration.

CHAPTER 14 / SPECTRA ON THE MOVE

How an offshore spectrum changes
as it approaches the coast.

Keep the offshore spectrum fixed, then move a virtual station into shallower water. Each frequency component feels the changing depth differently because its group velocity changes differently. This chapter isolates spectral shoaling under ideal, shore-normal, non-breaking conditions.

01 Start offshore02 Move the station shoreward03 Compare S0(f) and Sh(f)04 Track Hm0 and group velocity
ONE EXPERIMENT / THREE LINKED VIEWS

A · Offshore and local spectra

Dark line = prescribed offshore spectrum at 50 m depth. Blue fill = local spectrum at the selected station depth. In this idealized model, each frequency changes as Sh(f) = S0(f) Cg0(f) / Cgh(f).

Offshore S0(f)Local Sh(f)Peak frequency fp

B · Group velocity and shoaling factor by frequency

Higher and lower frequencies do not respond identically because finite-depth dispersion changes Cg differently at each frequency. The blue and navy group velocities use the left axis in m/s; the coral shoaling factor uses the right, dimensionless axis. The local shoaling factor is Ks(f) = √(Cg0/Cgh).

C · Depth transect and station location

OFFSHORE Hm02.40 m
LOCAL Hm02.66 m
PEAK PERIOD Tp10.0 s
ENERGY-FLUX CHECK≈ conserved

What this chapter is teaching

The spectrum is not a single wave. It contains many frequencies, and each frequency has its own wavelength and group velocity. As depth decreases, some parts of the spectrum slow down more than others, so the spectral shape can change even before breaking begins.

When does Hm0 rise or fall? Energy flux is conserved, so density rises when group velocity falls and falls when group velocity rises. Group velocity can first increase as a wave leaves deep water, then decrease in shallower water. The resulting height change need not be monotonic. Integrating the transformed spectrum determines Hm0.

What stays unchanged? The frequency axis itself. No wind generation or nonlinear transfer is acting here, so the chapter isolates the depth effect alone. In a full coastal wave model, spectral peaks may also shift or broaden because additional processes are active.

Questions to test the model

  1. Compare the swell-dominated and wind-sea-dominated presets at the same depth. Which one shoals more strongly, and why?
  2. Move the station from 50 m to 5 m. Which frequencies show the largest increase in Ks(f)?
  3. Why is this not yet a surf-zone model?

Go deeper: the transformation logic

Frequency-by-frequency shoaling relation
S(f,h)=S0(f)Cg0(f)Cgh(f)Ks(f)=Cg0Cgh

This is the spectral version of regular-wave shoaling: each narrow frequency band behaves like a small wave component with its own group velocity. When Cg decreases, the corresponding variance density rises.

How Hm0 follows from the transformed spectrum
m0(h)=∫0∞S(f,h)df,Hm0=4m0

Once the new S(f, h) is known, the updated spectral variance and Hm0 follow exactly as in Chapter 10. This chapter therefore links regular-wave shoaling to spectral descriptions without changing the statistical definitions.

What is still missing

Real nearshore transformation also involves depth-induced breaking, bottom friction, refraction, diffraction, wave–wave interactions, currents and sometimes wind input. Those mechanisms can alter not only the amplitude of spectral bands but also their shape and direction. Here we intentionally stop at the cleanest non-breaking baseline.

CHAPTER 15 / BEYOND THE SINUSOID

When a crest grows sharper.

Linear theory describes a wave as a symmetric oscillation about the mean surface. Finite-amplitude waves can have higher, narrower crests and broader, shallower troughs. The change in shape matters even when the crest-to-trough height stays the same.

Observe the shapeMeasure the correctionRecognise the limits

Small compared with what?

Wave height alone cannot tell whether linear theory is adequate. A one-metre wave can be relatively gentle in deep water yet occupy a substantial fraction of a shallow water column. The relevant comparisons are the height with wavelength, H/L, and with depth, H/d. Relative depth d/L identifies how strongly the bed influences the motion.

The steepness parameter ka = πH/L measures the slope scale of the leading sinusoid, where a = H/2. In shallow water, weak dispersion can make nonlinear effects important even when H/L is small. The Ursell number UR = HL²/d³ combines relative height and relative depth: UR = (H/d)/(d/L)². This lesson uses that definition throughout; other normalizations include different constant factors.

A harmonic that belongs to the wave.

A second-order Stokes profile adds a smaller oscillation at twice the fundamental frequency and wavenumber. Its phase is locked to the fundamental: it raises the crest and also raises the trough toward the mean surface. It is a bound harmonic, not a separately chosen free wave.

This distinction connects directly to the FFT lesson. A peak at twice the frequency may reflect nonlinear wave shape. It does not, by itself, prove that a separate wave system is present. Randomizing this harmonic's phase would destroy the particular Stokes shape.

Compare one wavelength.

Change H, T and d, then move the phase cursor. Both curves use the same wavelength from linear dispersion and the same H. This is a regular progressive wave over a horizontal bed, not a shoaling calculation.

Blue: linear profile. Coral: second-order profile. Vertical scale adjusts to H; the horizontal axis is x/L. This is not an equal-scale drawing.

Read the correction, not just the curve.

Write η = a cosψ + b cos2ψ, with ψ = kx − ωt. The ratio b/a measures the size of the second harmonic relative to the first. While the retained profile has its crest at ψ = 0 and trough at ψ = π, their elevations are a + b and −a + b. Their difference remains 2a = H. A higher crest therefore does not imply a larger crest-to-trough height in this comparison.

The second-order formula and its limits

ω = 2π/T,   ω² = gk tanh(kd),   L = 2π/k,   a = H/2

σ = tanh(kd),   b = ka²(3 − σ²)/(4σ³)

η(x,t) = a cos(kx − ωt) + b cos[2(kx − ωt)]

The equivalent coefficient is b = (πH²/8L) cosh(kd)[2 + cosh(2kd)]/sinh³(kd). In deep water σ tends to one and b tends to ka²/2. In shallow water the coefficient grows strongly, signalling the failure of this expansion if its correction is no longer small. Linear dispersion supplies k at this order; higher-order amplitude-dependent corrections to phase speed are not calculated here.

The display uses conservative teaching screens ka ≤ 0.2, H/d ≤ 0.3 and b/a ≤ 0.25, plus a Miche steepness screen. These are transparent interface choices, not validated error bounds. Passing them does not certify accuracy. The b/a = 0.25 boundary also marks the onset of additional stationary points in this two-harmonic profile: beyond it, an artificial secondary crest can develop in the trough.

The Miche comparison is H/L ≈ 0.142 tanh(kd), a limiting-steepness estimate, not a universal surf-zone breaking law. A case near that limit needs a different assessment even if a curve can be drawn. When any display screen fails, only the linear reference remains; it too must not be interpreted as a reliable prediction for that case.

Three experiments to try.

  1. Reduce H while holding T and d fixed. The first harmonic scales with H, but the second scales with H². The relative correction therefore shrinks.
  2. Return to H = 2 m and T = 8 s, then reduce d. Watch the Ursell number and b/a increase. A small H/L alone does not guarantee a useful Stokes expansion.
  3. Move the phase cursor. The two harmonics translate together; their relative phase stays fixed. The shape changes position, not form.

Height skewness versus wave-front asymmetry: not the same quantity

A sharply peaked crest with a broad trough has nonzero vertical (crest–trough) skewness. A wave leaning or pitching to one side has front–back asymmetry. The two need not increase together. We can illustrate the difference with a controlled two-harmonic signal, η(φ) = A cos φ + B cos(2φ + δ), where A and B (m) are harmonic amplitudes, φ is phase (rad) increasing in time at a fixed station, and δ is harmonic phase offset (rad).

Blue: prescribed elevation η; dashed grey: first harmonic A cos φ. Coral dotted curve: temporal phase slope dη/dφ, shown on its own normalized scale (not metres). Axes and legends identify both quantities.

Elevation skewness Sk(η)—
Temporal slope-skewness indicator—
Actual crest-to-trough height—

The displayed elevation skewness is the standardized central third moment, Sk(η) = ⟨η³⟩/⟨η²⟩3/2 (zero-mean signal). For this two-harmonic example: ⟨η²⟩ = (A² + B²)/2 and ⟨η³⟩ = (3A²B/4) cos δ. To expose phase asymmetry, the second readout uses a clearly defined temporal phase-slope skewness, Sk(η′) = ⟨(dη/dφ)³⟩/⟨(dη/dφ)²⟩3/2. This is a teaching diagnostic, not the standardized Hilbert-transform asymmetry coefficient used in some field studies. Its sign depends on whether phase is oriented forward in time or space.

Interpretation: δ = 0 gives a narrow high crest but zero phase-slope skewness; δ = 90° gives zero elevation skewness but nonzero phase-slope skewness. This arbitrary harmonic superposition is not a hydrodynamically solved Stokes or Boussinesq wave for every δ. The amplitude ratio is restricted for diagram clarity, not by a universal physical validity threshold. Crest-to-trough height is recalculated as δ changes; it is not assumed to remain 2A.

Source and terminology: Bosboom & Stive, Coastal Dynamics (2023), §5.3 (open-text rendering); Holthuijsen (2007), §5.6 for the limitations of finite-amplitude approximations. Equations and diagrams here are independently formulated for teaching.

CHAPTER 16 / CHOOSING THE DESCRIPTION

Choose the theory
for the question.

A wave theory is a set of assumptions and equations. Its usefulness depends on the conditions and on what needs to be predicted: a steady profile, local pressure, an evolving wave train or the statistics of an irregular sea.

The previous chapter showed one correction to a sinusoid. Adding terms is not always the best next step. Long waves in shallow water require a different balance of nonlinearity and dispersion, while an evolving coastal wave field requires equations that can follow changes through space and time.

Different descriptions answer different questions
DescriptionUseful starting pointWhat to check
Linear / AirySmall-amplitude waves; dispersion, first-order kinematics and superposition.Small corrections for the quantity of interest. A good period prediction does not ensure accurate crest pressure.
Stokes expansionPeriodic finite-amplitude waves, especially outside strongly nonlinear shallow-water conditions.Order and convergence; successive corrections should decrease. Second order is a shape illustration, not a universal substitute for higher orders.
CnoidalPeriodic, long, nonlinear shallow-water waves; pronounced crests separated by long troughs.Shallow-water assumptions and the specific approximation order. Its elliptic-function profile is not a cosine with an arbitrary crest sharpened.
SolitaryAn isolated long elevation pulse in shallow water.No repeating wave period or finite periodic wavelength. A solitary pulse is a limiting idealization, not a whole irregular sea state.
Stream-function / Fourier solutionA numerical steady periodic wave satisfying nonlinear free-surface conditions over a flat bed.Resolution, convergence and residual boundary-condition errors. A converged steady solution does not model overturning or an evolving breaking wave.
Boussinesq-type equationsEvolution of long waves, including interactions and propagation over varying bathymetry.Dispersion and nonlinearity range of the chosen formulation. Breaking and dissipation require additional treatment; the family has no single universal validity range.

Locate a wave in parameter space.

The diagram plots relative depth d/L against relative height H/d on logarithmic axes. The point comes from H, T and d, using linear dispersion to estimate L for initial screening. The lines are contours of UR, not borders between certified theories.

Each equal step on a logarithmic axis multiplies the value. Coral dashed line: Miche limiting-steepness estimate expressed as H/d. A point below it is not automatically safe or accurately described.

Editable range: H 0.05–8 m, T 2–25 s, d 1–100 m. Invalid or empty inputs pause the diagram.

How the guide produces its advice

First it checks whether the proposed wave reaches the Miche estimate or H/d ≥ 0.8. These are warning screens, not a simulation of breaking. It then identifies long shallow waves using d/L < 1/8 and UR > 20 as a broad educational cue toward cnoidal or Boussinesq descriptions. CEM discusses a cnoidal range of this kind, but published criteria vary with formulation and error tolerance.

Otherwise, b/a ≤ 0.05 together with ka ≤ 0.05 and H/d ≤ 0.1 indicates a small shape correction in this guide. Remaining cases prompt investigation of higher-order Stokes or a converged numerical steady-wave solution. These rules explain the displayed advice; they do not replace a validated selection chart, a convergence study or a check of the output quantity that matters. A numerical wavelength may differ from this initial linear estimate.

Periodicity is a physical choice.

Stokes, cnoidal and numerical steady-wave solutions describe repeating waves. A solitary-wave solution instead approaches still water on either side of one pulse. A leading shallow-water expression is η = H sech²[√(3H/4d³)(x − ct)], with c ≈ √(gd)(1 + H/2d), where H/d must be small enough for the approximation. No arbitrary repeating period should be attached to this pulse.

Cnoidal waves form a periodic long-wave family. As crests become widely separated, their form approaches an isolated-pulse limit. This is a conceptual connection between theories, not permission to apply a solitary profile to every large coastal wave.

Steady profiles and evolving seas.

A stream-function or Fourier solution solves a nonlinear steady-wave problem. The Fourier coefficients are determined together with the free-surface conditions; this is different from taking an FFT of measured elevations.

Boussinesq-type models evolve the wave motion and retain phase information. Phase-averaged spectral models instead evolve a statistical distribution in frequency and direction. Neither description is simply a more detailed version of the other: they retain different information and suit different scales and questions.

A practical sequence for choosing.

  1. Define the desired quantity and allowable error: surface crest, velocity, pressure, transformation or sea-state statistics.
  2. Check the physical setting: periodic or irregular, steady or evolving, uniform or varying depth, nonbreaking or breaking.
  3. Estimate d/L, H/L, H/d and UR using a consistent convention. Treat them as screening information.
  4. Compare suitable formulations, verify numerical convergence and check against an independent solution or measurements where available.
  5. Keep input uncertainty separate from theory and discretization errors. More equations cannot recover missing information about the waves.
CHAPTER 17 / DIRECTIONAL SEAS NEAR SHORE

Each component takes
its own path.

An offshore spectrum contains waves with different periods and directions. As the water becomes shallower, each component changes speed and turns according to its own wavelength and approach angle. The nearshore spectrum is therefore more than an offshore spectrum multiplied by one height factor.

Turning toward the depth-contour normal.

Over straight, parallel depth contours, a component obeys sinα/C = constant. Here α is its signed angle from the shoreward normal, and C is phase speed. A decrease in C reduces |α|: oblique waves turn toward the normal. Group velocity Cg controls energy transport, while phase speed C controls this refraction relation.

The example places north toward shore, east to the right and depth contours east–west. All bearings describe propagation toward: 0° is north and angles increase clockwise. Thus 320° is α = −40°, and 70° is α = +70°. Refraction can move both systems toward north without making them identical.

Conserve the flux crossing a contour.

For a stationary, laterally uniform, nonbreaking component, its shoreward variance flux is v Cg cosα, where v is the component variance. With no sources or losses, that flux stays constant. Variance itself can rise or fall as group speed and direction change. Multiplying by ρg gives the corresponding energy flux per unit crest-parallel length.

Conservation applies component by component. Refraction can also compress a directional band. A higher directional density need not indicate local wave generation: the same contribution can occupy a narrower interval of direction.

Two systems over parallel contours.

Two frequency lines, 0.10 and 0.16 Hz, carry offshore variances 0.30 and 0.18 m². Each is represented by five weighted rays, separated by 5°. This deliberately coarse directional quadrature isolates the mechanism; it is not a continuous forecast spectrum.

Plan view: distances in metres, east right and north toward shore. The bed slopes uniformly from 50 m offshore to the selected depth over 1 km. Central rays begin at different eastings for clarity; lateral uniformity makes their starting positions arbitrary.

Bar heights are integrated line variances in m², not density per hertz. Blue: offshore. Coral: selected depth. Ray directions and line contributions come from the same component calculation.

Offshore central bearings: 320° and 70°. Each band has offsets −10°, −5°, 0°, 5°, 10° and normalized weights 1:4:6:4:1.

Component calculation and directional density

ω² = gk tanh(kd),   C = ω/k,   Cg = ½[1 + 2kd/sinh(2kd)]C

α(d) = asin{[C(d)/C₀] sinα₀}

v(d) = v₀ Cg₀ cosα₀ / [Cg(d) cosα(d)]

Hm0(d) = 4√Σv(d)

For a continuous distribution, a mapped interval must carry the transformed variance: S(f,θ)dθ follows the interval, including the angular Jacobian and flux factor. It is incorrect to move density values into new directions and leave their angular widths unchanged. Angular integrals use radians. This experiment retains weighted integrated contributions rather than reconstructing a finely resolved directional density.

The displayed paths integrate dx/dy = tanα along a linear depth profile using 100 trapezoidal steps. Local directions and flux factors use the dispersion solution directly. Frequency stays constant because the bed is stationary and there is no ambient flow. Reflection, diffraction, breaking and dissipation are excluded; the shallowest settings expose the limit of this idealization rather than predicting surf-zone heights.

CHAPTER 18 / GAINS, LOSSES AND TRANSFERS

Why a spectrum changes.

Propagation moves wave energy from one place to another. Source terms describe local input, dissipation and exchanges between spectral components. Keeping these roles separate makes the spectrum easier to interpret—and the model easier to check.

The local spectral budget
ProcessPhysical interpretationEffect on the spectrum
Wind inputAtmospheric pressure and stress transfer energy to waves.Depends on wind relative to wave direction and speed; a fixed wind does not amplify every bin equally.
WhitecappingBreaking of wind waves dissipates organized wave energy.A modelled loss with formulation-dependent spectral distribution.
Bottom frictionOscillatory near-bed motion loses energy through boundary-layer processes.Depends on depth, wave motion and seabed properties; often important for components that feel the bed.
Depth-induced breakingFinite depth limits the waves that can propagate without breaking.A sea-state dissipation closure; clipping every wave to one height is not an equivalent spectral model.
Nonlinear interactionsWave components exchange energy through nonlinear dynamics.Quadruplet interactions are central in deep water; triad interactions become important in shallow water. They redistribute energy rather than act as an external energy supply.

Inspect a transparent bookkeeping budget.

Toggle prescribed source terms and compare the initial spectrum with one short explicit update. Rates are illustrative fixed numbers, chosen to expose signs and conservation. They are not wind-growth, friction or breaking parameterizations and cannot predict a real sea state.

Six bins of width Δf = 0.02 Hz. Bars show variance density S = v/Δf in m²/Hz. Blue: initial; coral: updated. Source rates remain fixed during this single update.

Include source terms
Equations, units and the conservative transfer

∂S/∂t = Qwind + Qwhite + Qbottom + Qbreak + Qnl

vᵢ,new = vᵢ,old + Δt qᵢ,   qᵢ = Qᵢ Δf

Δm₀ = Δt Σqᵢ,   E = ρg m₀

S has units m²/Hz and Q has units m²/(Hz s). The table reports bin-integrated q in m²/s. Negative rates are losses. Integrating a physical energy-source density instead would require the factor ρg. This lesson uses variance consistently.

The redistribution example adds 0.00016 m²/s at 0.08 Hz and 0.00024 m²/s at 0.12 Hz while removing 0.00040 m²/s at 0.10 Hz. Their sum is zero. Their sum after division by ω is also zero, so the discrete variance-based action proxy Σvᵢ/ωᵢ is conserved. This is a designed accounting example, not a resonant quartet solution: no wavenumber resonance or interaction coefficient is evaluated. Do not infer that all nonlinear processes conserve wave action; the distinction between triad and quartet interactions matters.

All selectable combinations remain nonnegative over the offered time range. A production model must also control stability, positivity and source-term stiffness. Reducing a time step does not correct an inappropriate physical parameterization.

A budget can balance without being accurate.

Conservation tests identify numerical and bookkeeping errors. They cannot establish that a wind-input or dissipation closure represents the observed sea. Different terms may compensate for one another and produce a plausible significant wave height while the peak period, direction or high-frequency tail is wrong. Compare the spectrum and its evolution, not only one bulk number.

CHAPTER 19 / PUTTING THE TERMS TOGETHER

A spectrum on the move.

A spectral-wave model follows a distribution over position, frequency and direction. Its balance equation accounts for what enters, what leaves and what changes locally. It does not follow the phase of every crest.

Start with one box.

Consider one geographic cell and one frequency–direction bin. The amount in the box increases when transport brings in more than it carries out, or when local sources supply a net gain. Refraction moves the distribution between directional bins while propagation moves it between geographic cells.

Shoaling is part of transport through a spatially varying group velocity. It should not be added again as an independent energy source. Likewise, turning between directions is not the same process as nonlinear exchange between frequencies.

Why introduce action?

Wave action is energy divided by intrinsic angular frequency σ. Intrinsic frequency is measured relative to the ambient water. For the stationary water and stationary bed used here, σ = 2πf remains constant along a ray, so energy and action conservation give equivalent transport results for each frequency.

Action provides the natural extension when intrinsic frequency changes in more general environments. That motivates the notation, but no current interaction or wave-driven current calculation is included in this lesson.

Read the balance equation term by term

∂N/∂t + ∂(cx N)/∂x + ∂(cy N)/∂y + ∂(cθ N)/∂θ = QE,σ/σ

cx = Cg sinθ,   cy = Cg cosθ

The equation shown is for stationary bathymetry and no ambient flow, with x eastward, y northward and clockwise propagation-toward bearings. The terms represent local change, eastward transport, northward transport and directional turning. cθ = dθ/dt must use the same angular convention as θ, with radians in derivatives and integration. It is not a user-chosen constant for arbitrary bathymetry.

Transport is written in conservative form: ∂(cx N)/∂x includes spatial changes in cx. Replacing it with cx ∂N/∂x would omit that effect. General formulations can add ∂(cσ N)/∂σ for intrinsic-frequency transport. That term is zero in the stationary, no-flow setting used here; nonlinear frequency redistribution still belongs in the source terms.

Keep frequency coordinates and units consistent

Let Sf(f,θ) be variance density per hertz and per radian of direction. Under σ = 2πf, Sσ = Sf/(2π), so Sfdf dθ = Sσdσ dθ. Define energy density Eσ = ρg Sσ and action density N = Eσ/σ. The energy source QE,σ must use the same angular-frequency coordinate before dividing by σ.

Integrating Eσ over dσ dθ gives energy per horizontal area, in J/m². Integrating N gives action per area, in J s/m². With radians treated as dimensionless, Eσ has units J s/m² and N has units J s²/m². Calling Sf/σ an angular-frequency action density without the 1/(2π) conversion would be inconsistent.

Move a packet through a row of cells.

This experiment isolates constant-speed, one-dimensional transport of one spectral bin. A smooth packet moves east at 10 m/s through 40 cells of width 50 m. Periodic boundaries wrap the packet around a 2 km domain; this is a numerical test, not a coastline.

Blue: numerical cell averages of N/Nref. Coral: exact translated cell averages. The reference profile is 0.1 + 0.9[(1 + cos(2π(x − 500)/2000))/2]⁸. No sources, refraction or frequency change occur in this isolated test.

Changing the Courant number resets the packet. Δt = Courant × Δx/Cg.

The finite-volume update

Nᵢⁿ⁺¹ = Nᵢⁿ − λ(Nᵢⁿ − Nᵢ₋₁ⁿ),   λ = Cg Δt/Δx

For positive constant speed, the upwind flux uses the upstream cell. With 0 ≤ λ ≤ 1, the update is a convex combination of two nonnegative values. Summing over the periodic domain cancels all internal fluxes, conserving integrated action. The exact comparison uses cell averages, not point samples, so it compares the same quantity.

At λ = 1 this particular scheme shifts the cell averages exactly by one cell each step. At smaller λ, repeated averaging spreads the packet and reduces its peak while preserving its total. This numerical diffusion is not whitecapping: no physical energy loss has been specified. The exact shift at λ = 1 is a special constant-speed, uniform-grid result, not a general recommendation for a full spectral model.

Connect the three lessons.

Chapter 17 supplied the physical change in propagation direction and shoreward flux. Chapter 18 separated local gains, losses and redistribution. The balance equation combines those processes while retaining a conservation account for every spectral bin. A complete model additionally needs bathymetry, boundary and initial conditions, physical source formulations and a verified numerical method.

CHAPTER 20 / DISCRETIZATION AND CONVERGENCE

Turn equations
into a calculation.

A model stores a finite set of numbers. Geographic cells describe where waves are; frequency and direction bins describe which waves are present. Time steps advance those numbers. Each choice can change the answer, even when the physical equations stay the same.

Four resolutions, different jobs.

A geographic grid must represent the bathymetry and the spatial scales of propagation. Frequency bins must resolve spectral peaks and tails. Directional bins must distinguish spreading and crossing systems. Time steps must resolve evolution and satisfy the requirements of the numerical method. Refining one coordinate does not repair inadequate resolution in another.

For a structured calculation, a rough storage count is Nx Ny Nf Nθ values per field. Doubling both horizontal dimensions multiplies that count by four; doubling the frequency and directional counts as well multiplies it by sixteen. More time steps and more source iterations add work without necessarily adding stored fields. These counts describe scaling, not a hardware benchmark.

Stable does not mean accurate.

A stable scheme can still smear a narrow swell peak or diffuse a packet in geographic space. Convergence means that a sequence of appropriately refined calculations approaches a consistent solution. It is not established by one fine-looking plot or by a small iteration residual on one grid.

Compare at the same physical time and with the same forcing and geometry. Record the outputs that matter: wave height, period, directional spread, energy flux and local extremes. A mesh that resolves significant height adequately may still miss a narrow directional feature.

Separate spatial and spectral resolution.

The upper plot repeats Chapter 19's periodic packet test at t = 200 s (one complete circuit of the domain), using Courant number 0.5 on each grid. The lower plot samples a narrow Gaussian teaching spectrum with known variance 0.36 m². Its peak is deliberately not aligned with most bin centres. No numerical renormalization hides a missed peak.

Blue: upwind cell averages; coral: analytical translated cell averages. Compare the same 2 km domain, speed 10 m/s and elapsed time.

Blue: midpoint-bin samples joined for visibility; coral: continuous reference. The frequency interval is 0.04–0.40 Hz. Peak frequency is 0.123 Hz and standard deviation is 0.006 Hz.

These controls change different calculations. Spatial refinement reduces packet diffusion; spectral refinement resolves the narrow peak. Neither is a physical growth or dissipation process.

What is being measured?

The packet error is the domain mean of the absolute difference between numerical and exact cell averages. The table compares all four grids after 200 s. Each grid uses Δx = 2000/N and Δt = 0.5Δx/10, hence 2N updates. A smaller time step accompanies a smaller cell to keep the Courant number fixed.

The spectrum is S(f) = 0.36 exp[−(f − 0.123)²/(2 × 0.006²)]/(0.006√(2π)). Its full-real-line integral is 0.36 m²; the omitted negative-frequency and displayed-band tails are negligible at these parameters. The numerical estimate is ΣS(fi)Δf. Peak period uses the largest sampled value and can change in steps as the grid changes. Zero-padding an FFT would not supply this missing physical or observational resolution.

Boundaries are part of the problem.

Specify incoming spectra at open boundaries, with units and direction conventions checked explicitly. Land boundaries, reflections and outgoing conditions depend on the model and configuration. Place artificial lateral boundaries far enough from the area of interest, and test their influence. Finer interior cells cannot repair incorrect boundary forcing.

A stationary solver seeks a balance under fixed conditions; a nonstationary solver also needs an initial state and an appropriate spin-up or initialization procedure. Iterative convergence, time-step sensitivity and grid convergence are separate checks. Source terms can impose additional restrictions beyond simple propagation.

CHAPTER 21 / EVIDENCE AND UNCERTAINTY

Agreement needs
more than one number.

A credible model result needs evidence about its equations, implementation, inputs and performance. A conservation check answers a different question from a comparison with a buoy. Both are useful, and neither replaces the other.

Verification

Does the implementation solve the intended mathematical problem sufficiently accurately? Use analytical solutions, conservation identities, limiting cases and convergence studies. Chapter 20 is an example.

Calibration

Estimate uncertain parameters using a designated dataset and a stated objective. Document the choices and plausible parameter range. A good fit to that dataset is not an independent performance test.

Validation

Evaluate performance against independent evidence for the intended application. Hold out complete events, locations or time periods where appropriate; nearby samples from one storm are not independent replications.

Diagnose four synthetic comparisons.

These ten reference values and their model counterparts are invented teaching data, not buoy observations. The second plot compares representative spectra normalized to the same m0 = 0.36 m²; it is a separate shape diagnostic, not a spectrum reconstructed from the ten height values.

Coral: synthetic reference. Blue: model example. Sample number has no assigned real date or sampling interval.

Both spectra have Hm0 = 2.4 m. Equal area does not require equal peak frequency, width or mean period.

None of these scores is a universal pass/fail threshold. Acceptable errors depend on the use, observed uncertainty and conditions being tested.

Metric definitions

ei = Mi − Oi;   bias = mean(ei);   RMSE = √mean(ei²)

Centered RMSE = √mean[(ei − bias)²]

Scatter index here = centered RMSE / mean(Oi)

Pearson correlation describes linear co-variation, not equality. A consistently scaled and biased model can have correlation one. Definitions of scatter index vary, so state the chosen convention. RMSE² = bias² + centered RMSE² for the same paired sample and population-mean normalization used here. These metrics retain their units except correlation and scatter index.

A ten-point demonstration cannot establish uncertainty bounds or rare-event skill. Correlated errors reduce the effective amount of information; resampling should respect storm or temporal dependence. A confidence interval derived as though every nearby sample were independent can be misleadingly narrow.

Compare like with like.

Before computing a score, align time stamps, locations, water levels, averaging intervals and quality-control flags. Confirm whether directions are from or toward, whether spectra use hertz or angular frequency, and whether reported heights are spectral Hm0 or wave-by-wave H1/3. Instrument response, sampling limits and processing choices can contribute to disagreement.

Use several diagnostics: time series, paired errors, spectral shapes, frequency and directional marginals, partitions and performance across sea-state ranges. Direction errors require circular differences, for example wrap(M − O) into [−180°,180°), rather than ordinary subtraction across north. A single mean direction can conceal two crossing systems.

Keep an evidence record.

Record the model version, bathymetry, forcing, boundary spectra, numerical settings, calibration dataset, independent evaluation dataset and acceptance criteria. Test sensitivity to plausible changes in uncertain inputs. Numerical error, observational uncertainty and physical model inadequacy should be discussed separately, even when their effects overlap.

CHAPTER 22 / A TRANSPARENT COASTAL CASE

Follow a sea state
toward the coast.

Bring the spectral description, transformation equations and conservation checks together in one bounded calculation. The aim is to explain every step from a prescribed offshore spectrum to a local result—not to imitate the capabilities of a full forecasting system.

1 · Specify the boundary2 · Transform every contribution3 · Inspect the budget4 · Export the evidence

The physical setting.

A straight coast lies north of an offshore boundary. Depth decreases linearly from 50 m to the selected local depth over 2 km; contours run east–west. The sea state is stationary and uniform along the contours. The calculation uses linear dispersion, Snell refraction and conservation of shoreward variance flux for each represented component.

The offshore spectrum contains a swell partition and a shorter-period partition with central toward bearings 320° and 50°. In the single-system case only the swell remains. Each partition uses a normalized Gaussian frequency shape and five weighted directional contributions. These are prescribed teaching spectra, not wind-generated predictions.

One optional loss experiment.

A prescribed attenuation coefficient μ reduces every component's shoreward flux by exp(−μy/2000), where y is the northward distance in metres. μ is a dimensionless cumulative attenuation over the whole transect. It is an accounting surrogate, not a bottom-friction or breaking formulation, and is applied per shoreward distance rather than distance along a ray.

With μ = 0, every component conserves flux. With μ > 0, the total boundary-to-local flux ratio must equal exp(−μ). This identity gives an independent check even when the spectral shape and height change. No wind input, nonlinear redistribution or breaking closure is included.

Run the idealized transect.

Hm0 at 21 stations along a 2 km straight transect. Station spacing is for displaying the analytic component transformation, not a finite-volume propagation grid.

Blue: offshore frequency marginal. Coral: local marginal from the same transformed directional contributions. Frequency bins are 0.005 Hz wide, centred from 0.05 to 0.25 Hz; all angular weights are integrated contributions.

Calculation recipe and checks
  1. Set target offshore variance m0 = (Hm0/4)². For two systems, allocate 65% to swell and 35% to the shorter-period partition; a single swell receives all the variance.
  2. Normalize the sampled Gaussian frequency weights on the stated finite grid. Swell standard deviation is 0.012 Hz; the shorter-period partition peaks at 0.16 Hz with standard deviation 0.018 Hz.
  3. Distribute each frequency-bin variance over direction offsets −10°, −5°, 0°, 5°, 10°, with normalized weights 1:4:6:4:1. No per-degree density is integrated as though it were per-radian density.
  4. At each station solve dispersion, turn each component using Snell's law and apply v = v0 Cg0 cosα0/[Cg cosα] × exp(−μy/2000).
  5. Sum local contributions to obtain m0, Hm0 and m1. Compute Tm01 = m0/m1 and the variance-weighted circular mean bearing. The mean bearing summarizes the mixture; it is not either partition's peak direction.
  6. Check the summed shoreward variance flux against the prescribed exponential ratio. Read the applicability notice before interpreting shallow-water results.

What this case can—and cannot—establish.

The calculation verifies an idealized stationary transformation and its flux account. It does not solve an arbitrary two-dimensional coastal domain, predict breaking or validate a real location. At large Hm0/depth, the nonbreaking extrapolation may be physically inappropriate even though the conservation equations still balance.

A site study would additionally need surveyed bathymetry, defensible boundary forcing, appropriate source formulations, spatial and spectral resolution tests, and independent observations. The exported station table records this example's inputs and outputs so its assumptions and results remain traceable.

CHAPTER 23 / DIFFRACTION AND WAVE SHELTERING

Behind a breakwater,
the sea still moves.

A breakwater interrupts an approaching wave crest, but its geometrical shadow is not a region of perfectly still water. Waves spread around the tip, changing both their height and the shape of their crests. This redistribution is called diffraction.

Three processes, three questions.

Refraction changes propagation direction when wave speed varies, for example over changing depth. Reflection sends part of the wave field back from a boundary. Diffraction redistributes the field laterally around obstacles and across variations along a crest. These processes can occur together; a sheltered point may receive waves even when a straight incident ray cannot reach it.

The experiment below isolates diffraction around a single tip in constant depth. It includes the contribution associated with reflection from an ideal rigid barrier, although the map shows only the water on its lee side. There is no shoaling or depth-induced refraction in this setting.

A coefficient describes local height.

The diffraction coefficient is Kd = Hd/Hi, the local regular-wave height divided by the incident regular-wave height. For this periodic field, Kd² is the ratio of local surface-elevation variance to incident variance, and also the ratio of mean gravitational potential energy. In a coherent interference field, total local energy additionally depends on particle velocities. Kd² is not, by itself, the fraction of power transmitted through an entire harbour entrance.

Kd can exceed one locally because interference concentrates the field. This is redistribution, not energy generation. In the shadow, the height generally diminishes away from the edge; beyond the shadow, constructive and destructive interference produce oscillations.

EXPERIMENT / ONE TIP, ONE FREQUENCY

Look into the wave shadow.

A thin, fully reflecting barrier extends south from the origin. Incident waves travel east, toward 90°, normal to the barrier. The map covers a fixed 300 m × 300 m lee-side region. Changing period or depth changes wavelength without moving the barrier or gauge.

A · Lee-side wave field
North is up; east is right. The dashed line is the geometrical shadow boundary. The dot is the movable gauge.
The gauge result below reports the local diffraction coefficient.

Paused

B · Across the lee at the gauge's eastward distance
Blue: Kd. Dashed: straight-ray prediction, with an abrupt shadow. Coral dot: gauge.

Height ratio is time-independent. Run switches to the moving surface. Colours are quantities in the legend, not bathymetry. aᵢ = Hᵢ/2.

Predict, then experiment.

  1. Keep the gauge south of the shadow boundary. Increase the period and compare Kd at the same physical location; the ratios x/L and y/L change.
  2. Move the gauge through y = 0. The wave field makes a smooth transition, unlike the abrupt straight-ray shadow.
  3. Run the surface view. Curved crests extend into the shadow. Switch back to height ratio to separate an oscillating surface from its stationary amplitude pattern.

A small calculation with a useful warning.

If Kd = 0.30 and the incident regular-wave height is 1.0 m, the local regular-wave height is 0.30 m and the surface-elevation variance ratio is 0.30² = 0.09. This does not mean that 91% of all incoming energy has been dissipated: most of the wave field lies elsewhere. For a rigid barrier, no energy passes through the material itself.

Mathematics: the constant-depth half-plane solution

Let x point east, y north, and the barrier occupy x = 0, y ≤ 0. We show x ≥ 0. Solve ω² = gk tanh(kd), with ω = 2π/T and L = 2π/k. A complex spatial coefficient Z describes the surface: η(x,y,t) = aᵢ Re[Z exp(−iωt)], and Kd = |Z|.

Radial distance
r=x2+y2
Diffraction coordinates
u1=sign(y)4(r−x)L,u2=−4(r+x)L
Fresnel cosine integral
C(u)=∫0ucos(πs22)ds
Fresnel sine integral
S(u)=∫0usin(πs22)ds
Complex Fresnel function
F(u)=1+C(u)+S(u)2+i2[S(u)−C(u)]
Complex superposition
Z=exp⁡(ikx)F(u1)+exp⁡(−ikx)F(u2)

This is the normal-incidence Sommerfeld–Penny–Price half-plane solution in the coordinates above. Both terms are retained; dropping the second would omit the diffracted contribution from the reflected wave. It solves the linear constant-depth Helmholtz problem with zero normal velocity at the rigid barrier. Fresnel integrals are evaluated from a numerical table; the map samples the resulting analytical field rather than advancing a coastal model on a mesh.

The mathematical tip has singular spatial gradients. The image samples cell centres and does not resolve local tip loads. A value near Kd = 0.5 along the distant shadow boundary is a useful limit, not an exact coefficient everywhere on that line.

From one wave to a spectrum and a harbour entrance

Different periods and incident directions have different diffraction patterns. For independent random components, transform each component's variance by its own |Z(f,θ)|² and then integrate. Adding wave heights instead of variances is incorrect for that statistical calculation; coherent components require their complex phases. A shorter component and a longer component therefore need not be sheltered by the same amount.

A finite entrance has two edges, with interacting diffracted fields. Multiple reflections, entrance geometry, varying depth and harbour resonance can all matter. Two copies of this height-ratio map cannot simply be added to produce a gap solution. CEM discusses separate gap and directional-wave analyses; those are extensions beyond this experiment.

Explore the linked 1D–2D–3D reflection studio in Wave Lab.

CHAPTER 24 / REFLECTION, TRANSMISSION AND DISSIPATION

New in this chapter: Explore incident and reflected waves together using linked 1D, 2D and 3D views. The older two-view reflection/transmission/loss experiment is preserved in the supplementary section below.

One incoming wave.
Three destinations for its power.

When waves meet a coastal structure, some energy may return seaward, some may reach the sheltered side, and some may be converted into turbulence and heat. A quieter lee does not tell us which process dominates. The reflected field and the energy budget must also be examined.

Height coefficients are not percentages of power.

Define the amplitude coefficients Kr = Hr/Hi and Kt = Ht/Hi using the heights of the separate incident, reflected and transmitted waves. For regular linear waves, mean energy per unit horizontal area is E = ρgH²/8. Power per metre of crest is P = E Cg, where Cg is group velocity.

At equal depth, equal frequency and normal incidence, reflected and transmitted power fractions are Kr² and Kt². A reflected height half the incident height therefore carries one quarter of its power. The dissipated fraction D is defined by the power balance, not by subtracting heights: D = 1 − Kr² − Kt².

Ask what the structure actually does.

An ideal impermeable vertical wall reflects strongly and transmits no waves through its material. A permeable structure can transmit part of the field while dissipating energy internally. Waves can also reach the lee by overtopping or by travelling around an end. These pathways involve different physics and cannot be inferred from a single coefficient alone.

The experiment prescribes a passive energy partition. It does not predict coefficients from a structure's material, geometry or crest elevation. CEM provides separate methods for particular structure types; those methods require additional inputs and applicability checks.

MAIN REFLECTION EXPERIMENT · THREE LINKED VIEWS

Reflection: see the standing-wave field develop

Move the virtual gauge and vary the amplitude reflection coefficient, period and angle. The 1D gauge record, 2D interference map and 3D perspective wave surface use one linear incident–reflected superposition calculation.

The interactive reflection experiment loads when this section is in view. Alternatively explore it in Wave Lab.

Important distinction: This main studio treats reflection and interference at a long, straight boundary in uniform depth. It does not independently calculate transmitted wave conditions, structural dissipation or impact pressures. Expand the supplementary reflection–transmission–loss budget experiment below for those separate bookkeeping concepts.

The Wave Lab reflection studio remains available for independent exploration.

Supplementary experiment — reflection, transmission and dissipation budget

This original two-view teaching tool introduces a prescribed passive power allocation among reflected, transmitted and dissipated components. It complements, but does not replace, the linked reflection-wave field above.

Follow the same wave through two views.

Regular waves travel east, toward 90°, in 10 m depth on both sides. Incident height is 0.40 m and period is 6 s. The grey band is an unresolved interaction region, not a model of a wall or porous medium.

A · Surface elevation and propagation
Navy: total seaward surface. Blue: incident; coral: reflected. Green: transmitted. Dashed: seaward amplitude envelope.

Paused

B · Time-averaged power budget
The three shares sum to 100%. They remain constant while the surface moves.

The loss slider applies only to the power that was not reflected. Transmission is calculated from the remainder, so the example cannot create energy. Phase shifts move interference nodes without changing the power fractions.

Three controlled experiments.

  1. Set Kr = 1 and reflection phase = 0°. All incident power returns seaward; the seaward surface is a standing wave. The loss slider has no effect because no power remains to allocate.
  2. Set Kr = 0 and loss = 0%. The incident wave is entirely transmitted in this idealization.
  3. Set Kr = 0.5 and loss = 40%. Reflection takes 25%; dissipation takes 40% of the remaining 75%, or 30%; transmission takes 45%. Kt = √0.45 ≈ 0.671, not 0.45.

A standing pattern does not generate power.

For perfect reflection with zero phase shift, the surface is η = 2a cos(kx) cos(ωt). Antinodes have twice the incident amplitude while nodes have no surface oscillation. The net seaward-side power toward the structure is zero. Local surface variance alone does not describe total local energy: particle motion contributes kinetic energy, and local kinetic and potential energies need not be equal in an interference pattern.

With partial reflection, the total seaward surface is the coherent sum of two travelling components. Do not treat its local crest-to-trough height as the incident height when calculating Kr. Incident/reflected separation generally needs suitable spatial measurements and a wave model; one elevation record cannot uniquely separate two arbitrary opposing waves.

Equations and exactly what the controls prescribe

Let a = Hi/2 = 0.20 m, ω = 2π/6 s⁻¹, and solve ω² = gk tanh(kd) at d = 10 m. Group velocity is Cg = ½[1 + 2kd/sinh(2kd)]ω/k. Use ρ = 1025 kg/m³ and g = 9.81 m/s².

Energy fractions
R=Kr2,D=λ(1−R),T=(1−λ)(1−R)
Amplitude transmission
Kt=T
Incident flux per crest length
Pi=ρgHi2cg8
Reflected, transmitted and dissipated
Pr=RPi,Pt=TPi,Pd=DPi
Incident wave
ηi=acos⁡(kx−ωt)
Reflected wave
ηr=aKrcos⁡(−kx−ωt+δ)
Transmitted wave
ηt=aKtcos⁡(kx−ωt)
Standing-pattern amplitude
A(x)=a1+Kr2+2Krcos⁡(2kx−δ)

λ is the slider's fraction of unreflected power dissipated; it is not D itself. R, T and D above denote power fractions; the period is written explicitly as 6 s in this experiment to avoid confusing it with the fraction T. δ is the prescribed reflection phase. The transmitted phase is set to zero as a teaching convention. No surface or velocity matching condition is imposed across the unresolved interaction region, and these chosen coefficients and phases are not a solution for a specified structure.

The plotted coordinate is x/L. Incident and reflected fields occupy x < 0; the transmitted field occupies x > 0. A small band around zero hides the unresolved interaction. The envelope is a positive amplitude magnitude; both ±A are drawn. For Kr = 1, δ = 0, the first seaward node is at x = −L/4.

Different depths, angles and random seas

If transmitted and incident group velocities differ, a height ratio alone no longer gives a power ratio. At normal incidence, Pt/Pi = Kt² Cg,t/Cg,i. For power crossing a fixed interface under oblique incidence, also include the ratio of the normal-direction factors cosαt/cosαi, with α measured from that interface's normal. The reflected branch on the incident side has the same group speed and equal angle magnitude in the usual stationary, uniform-medium case.

For a directional spectrum, reflected and transmitted coefficients may depend on frequency and angle. Integrate the component power contributions with their group velocities and normal-direction factors. An integrated significant-height ratio summarizes variance; it need not equal the square root of the integrated power ratio when spectra, depths or directions differ.

A measured missing-power residual is not automatically all dissipation. An incomplete control volume, unresolved directions, temporary energy storage and measurement errors can also affect the balance. The three-way budget here assumes steady conditions and that reflection, transmission and dissipation account for every pathway.

CHAPTER 25 / SWASH, RUN-UP AND OVERTOPPING

Where does the
water reach?

A wave arriving at a beach or sloping defence does not stop at the still-water shoreline. Water moves up the slope and returns downslope. A high crest may be reached only occasionally; a lower crest may allow repeated overtopping. Elevation, exceedance probability and discharge describe different parts of this behaviour.

Separate the water levels.

Still water level is the reference level for the calculation, including the selected tide and surge contribution. Wave setup is a wave-induced change in the mean level. Swash is the fluctuating shoreline motion about that mean. Run-up describes the elevation reached relative to a stated reference, so the reference and any setup contribution must be made explicit.

For an idealized beach, instantaneous shoreline elevation may be written z(t) = zSWL + setup + swash(t). That accounting identity is not a run-up predictor. The structural formula below predicts Ru2% relative to its specified SWL; do not add a separate setup estimate to its output without checking the formulation and reference level.

A high percentile is not a maximum.

Ru2% is the run-up height exceeded by 2% of incoming waves under the specified sea state. It is neither the largest possible run-up nor a 2% annual flood probability. Crest freeboard Rc = zcrest − zSWL measures the height available above the still water level.

Mean overtopping discharge q is the volume crossing the crest per unit time and per metre of structure length. It has units m³/s/m, equivalently m²/s; multiplying by 1,000 gives L/s/m. A small average can conceal intermittent large events. Ru2% below the crest does not imply zero overtopping, and q is not a flow velocity.

EXPERIMENT / A SMOOTH SLOPING DEFENCE

Raise the water. Keep the crest fixed.

This calculator uses EurOtop (2018) mean-value relationships for a straight, smooth, impermeable slope under normal wave attack. All influence factors equal one. Inputs Hm0 and Tm−1,0 are the incident sea-state parameters at the structure toe, not untransformed offshore values or a peak period.

A · Slope and reference levels
Blue: SWL. Coral: Ru2% elevation on a hypothetical continued slope. Navy: crest. Heights are relative to one local datum; geometry is schematic.

B · Freeboard sensitivity
Blue: predicted mean q while varying Rc alone. Coral dot: current freeboard. Vertical scale is logarithmic; small values are not zero.

Toe bed is fixed at −10 m datum, so toe depth = 10 + SWL. Changing SWL here holds the entered toe wave conditions fixed. A real tide or surge can also change wave transformation before the toe.

Try three comparisons.

  1. Raise SWL without moving the crest. Freeboard falls. Compare the corresponding increase in overtopping.
  2. Increase the crest elevation. Ru2% remains unchanged in this formulation, while q decreases. Run-up on an uninterrupted slope and discharge across a finite crest answer different questions.
  3. Change the energy period while keeping Hm0 fixed. The surf-similarity parameter changes. Try an input outside the teaching range and read why the quantitative prediction is withheld.

From a rate to a volume.

A hypothetical mean q = 0.010 m³/s/m equals 10 L/s/m. If that rate were uniform along a 20 m crest and steady for ten minutes, the total overtopped volume would be 0.010 × 20 × 600 = 120 m³. It does not tell us the peak volume of one overtopping wave or the depth of flooding behind the structure.

Equations, units and the selected branches

Write H = Hm0, T = Tm−1,0 = m−1/m0, tanα = 1/n and g = 9.81 m/s². The reference deep-water wavelength L0 = gT²/(2π) is used to define the surf-similarity parameter even though the wave parameters are specified at the toe; it is not a claim that the toe is in deep water.

Deep-water steepness and surf similarity
s=HL0,ξ=tan⁡αs
Illustrative runup relation
Ru2%=Hmin(1.65ξ, 4−1.5ξ)
Breaking branch
Qb=0.023ξtan⁡αexp[−(2.7RcξH)1.3]
Non-breaking branch
Qn=0.09exp[−(1.5RcH)1.3]
Mean discharge (R_c ≥ 0)
q=gH3min(Qb,Qn),Rc≥0

These are EurOtop (2018) Eqs. 5.1–5.2 and 5.10–5.11 specialized to all influence factors equal to one. “With a maximum of” is implemented as the smaller of the two branch values, not their sum. The fractional power applies to the full positive quantity inside its parentheses; the minus sign stays outside that power. The chart and numerical results use the same function.

The run-up equations give a mean prediction of the 2% exceedance level; the overtopping equations give a mean prediction of the time-averaged discharge. Neither is a conservative design assessment. Empirical scatter and uncertainty in wave conditions, water levels and geometry remain. The displayed digits support comparisons, not an accuracy claim.

Applicability: when this calculator withholds a result

The implementation is deliberately narrower than the complete manual: slope 1:2.5 to 1:6; wave steepness 0.01 ≤ H/L0 ≤ 0.06; 0.5 ≤ ξ ≤ 4; toe depth/H ≥ 4; and nonnegative freeboard. These are conservative teaching screens for this simplified experiment, not a universal validity envelope or a guarantee of accuracy. The actual structure and sea state must also satisfy the stated assumptions.

Predicted Ru2% and q are withheld outside these screens. A negative freeboard introduces overflow, requiring different treatment. Very shallow foreshores, substantial infragravity energy, rough or permeable armour, berms, vertical walls, oblique attack and compound slopes need the appropriate additional formulations. Natural-beach run-up also requires a formulation developed for that setting; this structural calculator should not be relabelled as a beach predictor.

The dashed continued slope is a way to display a run-up statistic above a low crest. It does not model a sheet of water travelling through the air. No overtopping jets, individual-wave volumes, landward flooding, erosion, structural loads, safety thresholds or wave-driven currents are calculated.

CHAPTER 26 / INFRAGRAVITY WAVES AND SURF BEAT

Small waves arrive in groups.
A slower motion follows.

Watch a beach for several minutes and the shoreline may advance and retreat on a rhythm much slower than the individual incoming waves. Infragravity waves occupy this lower-frequency part of the surface motion. Understanding them starts with a distinction: the envelope of a short-wave group is not itself an extra wave in the elevation spectrum.

Grouping and a long wave are different.

Two nearby frequencies add to produce alternating groups of larger and smaller waves. Their linear sum still contains only those two frequencies. Its amplitude envelope varies slowly, but that envelope is a derived quantity, not a third sinusoid in the surface elevation.

A real low-frequency elevation component requires additional physics or an independently present long wave. Nonlinear interactions and spatial variations in the momentum carried by short waves can produce such a response. Calling every visible group an infragravity wave confuses the forcing pattern with the resulting water-level motion.

Bound and free describe different behaviour.

A bound long wave remains coupled to its forcing group. In the classical weakly nonlinear, uniform-depth picture, the slowly varying surface is depressed beneath the larger waves and elevated beneath the smaller waves. Its propagation is tied to the group, so treating it as an independent free wave with its own linear dispersion relation would be inappropriate.

A free long wave can propagate independently of the short-wave group. Changes in depth and breaking can alter the coupling and contribute to the generation or release of free infragravity waves. These processes do not reduce to simply switching off the short waves at one point.

Separate the envelope from the elevation.

This experiment uses an exact two-frequency short-wave signal and an optional prescribed slow elevation component. The slow component has the phase relationship of a bound-wave illustration. Its amplitude is chosen with a slider, not predicted from the short-wave amplitude, water depth or a nonlinear wave model.

A · Short waves and their envelope
Blue: ηshort. Dashed: ± envelope. The dotted line marks the selected instant.
B · Slow elevation component
Coral: ηslow, with a depression beneath the centre of a large group when enabled. This panel uses its own vertical scale.
C · The combined elevation
Navy: ηshort + ηslow. All three panels share the same time axis.
D · Component variance, not spectral density
Each stem has units m². Blue: the two short-wave components. Coral: the optional slow component. Stem width has no physical meaning.

Each short-wave component has amplitude a = 0.25 m. Their mean frequency is 0.125 Hz (8 s). The view spans two group periods and changes length with Δf. This is a time-series experiment, not a spatial beach simulation.

Three comparisons that reveal the mechanism.

  1. Select “Short waves only”. Large and small groups remain, but the elevation signal has no component at Δf. A slow envelope alone is not evidence of an infragravity elevation wave.
  2. Enable the prescribed component. Its variance appears at Δf, and the combined elevation shifts downward under large groups. No energy-transfer calculation has taken place: the experiment adds a specified signal.
  3. Halve Δf. The group period doubles. The periods of the individual short waves remain close to 8 s; there is no universal rule that the group period must be a fixed multiple of the short-wave period.

A 100-second rhythm from waves near eight seconds.

With f1 = 0.120 Hz and f2 = 0.130 Hz, individual periods are 8.333 s and 7.692 s. The repeating group envelope has period 1/(f2 − f1) = 100 s. Each 0.25 m amplitude component contributes 0.25²/2 = 0.03125 m² variance. The linear short-wave sum therefore contributes 0.06250 m².

If the separately prescribed slow amplitude is B = 0.06 m, its variance is 0.06²/2 = 0.00180 m². Total component variance is 0.06430 m². The component budget is exact for a common-period average; a short arbitrary observation window need not give exactly that sample variance.

What changes near a beach?

Infragravity motion can involve more than one generation mechanism. Short-wave groups can force bound low-frequency motion outside the breaking region. The breaking location and the distribution of breaking can also vary with the arriving groups, producing low-frequency forcing. Reflection at the shoreline and interactions with the bathymetry then modify the observed signal. Surf beat describes slow surf-zone fluctuations; it is not a single universally applicable formula.

Periods of tens to hundreds of seconds are useful introductory scales, but the boundary between infragravity and sea-swell bands depends on the application and the incident spectrum. State the frequency limits whenever reporting band-integrated variance or height. The experiment's 50–200 s slow periods are teaching choices, not a definition covering all environments.

Exact signal mathematics and the difference-frequency distinction
Two nearby frequencies
f1=f0−Δf2,f2=f0+Δf2
Example central frequency
f0=0.125 Hz
Short-wave superposition
ηshort(t)=acos⁡(2πf1t)+acos⁡(2πf2t)
Product form
ηshort(t)=2acos⁡(πΔft)cos⁡(2πf0t)
Group envelope and recurrence
A(t)=2a|cos⁡(πΔft)|,Tg=1Δf
Prescribed long-period component
ηslow(t)=−Bcos⁡(2πΔft)
Total elevation
ηtotal=ηshort+ηslow
Component variances
V1=V2=a22,Vslow=B22

The signed modulation cos(πΔf t) has period 2/Δf, but its absolute amplitude envelope repeats after 1/Δf. This explains a common factor-of-two mistake. Squaring the short-wave signal creates a difference-frequency term through a product identity. A low-frequency peak in the squared signal or envelope therefore does not prove that the original elevation record contains that component.

The prescribed ηslow is anticorrelated with A²: A² = 2a²[1 + cos(2πΔf t)]. This establishes the displayed phase relationship only. It does not derive the amplitude of a physically generated bound wave. When the slow wave is disabled, its effective B is zero; the two short-wave components are unchanged.

For these selected frequencies, an 800 s record contains an integer number of every component's cycles. Independent verification uses that common duration to compare sample variance and Fourier-bin amplitudes with the analytical component values. The displayed two-group window serves visual interpretation and is not used to estimate the plotted stem variances.

Measurements, modelling and what the experiment leaves out

To observe slow motions, a record must be long enough to contain many of their cycles, with instrument response and processing suitable for the low-frequency band. Detrending and high-pass filtering can remove part of the desired signal. Tides, surge, instrument drift and nonstationarity must be distinguished from wave-band variability rather than absorbed into a generic “long wave”.

Separating bound and free motions usually requires information about propagation and coupling, such as spatial measurements and phase relationships. A single elevation spectrum cannot uniquely identify the generation mechanism. A phase-averaged model configured only for sea-swell frequencies does not automatically resolve infragravity wave generation or its phase relationships. Appropriate long-wave coupling or a suitable phase-resolving model requires separate verification.

This illustration does not solve radiation-stress forcing, nonlinear transfer, depth transformation, breaking, release, reflection, shoreline run-up or resonance. No wave-driven currents are calculated. Increasing B adds prescribed variance without reducing short-wave variance; it must not be interpreted as a conservative simulation of energy transfer. No site forecast or design value is supplied.

CHAPTER 27 / SEICHES AND HARBOUR RESONANCE

A basin has its own rhythm.

Water disturbed in a basin can continue to oscillate after the disturbance has passed. These standing oscillations are called seiches. Their natural periods depend on the basin's dimensions, depth and boundaries. Repeated forcing near a natural frequency can build a much larger response, but the forcing pattern and the ways energy escapes or dissipates also matter.

A standing pattern, not a travelling crest.

In a single standing mode, some locations rise while others fall. A surface-elevation node stays at the mean level throughout the cycle; an antinode has the largest vertical excursion. A flat surface at one instant does not mean the oscillation has stopped. Different parts of the basin exchange potential and kinetic energy as the cycle continues.

The fundamental mode is the slowest nonzero oscillation. Higher modes fit more spatial structure into the same basin and have shorter periods. A real disturbance can excite several modes at once; the resulting motion need not resemble one simple back-and-forth pattern.

The opening changes the ideal pattern.

A rigid end is an elevation antinode in this uniform-depth, one-dimensional model. A perfectly open end is idealized as an elevation node. A basin closed at both ends fits half a wavelength in its fundamental mode; a basin closed at one end and open at the other fits a quarter wavelength. With identical length and depth, the latter therefore has twice the fundamental period.

A real harbour mouth is not generally a perfect node. Entrance geometry, exchange with the exterior and wave radiation alter the oscillation periods and damping. The quarter-wave estimate is a useful starting point, not a complete harbour model.

Explore a mode, then explore its response.

Panel A shows an undamped free mode with a fixed 0.10 m antinode amplitude. Panel B uses the same natural period in a separate, normalized forced-oscillator experiment. Changing damping or forcing frequency changes Panel B only; it does not scale or damp Panel A.

A · Free standing mode
Blue: instantaneous elevation. Dashed: ± amplitude envelope. Coral dots: permanent elevation nodes. Depth and elevation use different scales; this is a schematic longitudinal section.

B · Idealized single-mode response
Navy: steady response divided by static modal displacement. Coral: chosen frequency ratio. Dashed vertical line: undamped natural frequency, r = 1. This is not harbour/outside wave-height amplification.

Run uses an accelerated eight-second display cycle, not real-time basin evolution. Adjusting the free-mode cycle pauses playback. No currents, vessel motions or loads are calculated.

Three experiments to try.

  1. Keep the fundamental mode and change the right end from closed to open. Its natural period doubles, and the right-hand antinode becomes a node. The basin length has not changed; the wavelength that fits has changed.
  2. Double the length: the period doubles. Increase depth from 2.5 m to 10 m: the period halves because long-wave speed varies as √h. Higher modes are faster, not larger by definition.
  3. At r = 1, increase damping. The normalized response falls and the resonance curve broadens. Move to r > 1: the displacement lags the forcing by more than a quarter cycle. The blue free-mode illustration remains unchanged because it is a separate experiment.

A kilometre-long basin, ten metres deep.

With g = 9.81 m/s², long-wave speed c = √(gh) = 9.9045 m/s. A closed basin's fundamental period is 2 × 1000/9.9045 = 201.93 s, or 3.37 minutes. Its fundamental wavelength is 2000 m. The ideal open-ended counterpart has wavelength 4000 m and period 403.86 s, or 6.73 minutes.

In the separate oscillator example, ζ = 0.10 and r = 1 give response factor 1/(2ζ) = 5 and phase lag 90°. The displacement-response maximum occurs slightly below r = 1: rpeak = √(1 − 2ζ²) = 0.98995, with peak factor 5.0252. Neither number specifies a real harbour wave height because the modal forcing strength and coupling have not been supplied.

Why a small surface motion can matter.

Long-period oscillations can disrupt harbour operations even when the visible rise and fall seems modest. A moored vessel and its mooring system have their own dynamic response, which must be assessed separately. An elevation node is not a place where all water motion vanishes, so placing a gauge or berth at a node does not establish safe conditions. This lesson does not estimate vessel movement or mooring loads.

Infragravity waves can supply forcing in a relevant period range, linking this chapter to Chapter 26. Wind and atmospheric-pressure disturbances can also excite basin oscillations. Matching a forcing period to a natural period indicates potential resonance; substantial response additionally requires the spatial forcing to couple to the mode. A free seiche left after a disturbance and a continuously forced resonant response are distinct situations.

Deriving the mode periods and identifying nodes

For small surface displacement in a narrow basin with uniform depth, the linear long-wave equation is ηtt = gh ηxx. Let c = √(gh), and write η(x,t) = A cos(kn x) cos(ωn t), with x = 0 at the closed left end, ωn = ckn and A = 0.10 m. A rigid end imposes zero normal motion, equivalent here to ηx = 0. The ideal open right boundary imposes η = 0.

Closed–closed natural periods
kn=nπℓ,Tn=2ℓngh
Closed–open natural periods
kn=(2n−1)π2ℓ,Tn=4ℓ(2n−1)gh
Mode index and kinematics
n=1,2,3,…,λn=2πkn,Tn=2πωn
Elevation nodes (within basin)
xj=(2j+1)π2kn,0≤xj≤ℓ
Elevation antinodes (within basin)
xj=jπkn,0≤xj≤ℓ

The closed-basin uniform mode n = 0 is excluded: it represents a mean-level change, not a free oscillation with a finite period. Each displayed closed mode has zero spatially integrated surface displacement. The open-ended mode can exchange water with the exterior and need not satisfy that closed-basin volume constraint.

The shallow-water approximation requires knh ≪ 1. The readout reports knh and compares the period with the finite-depth linear relation ω² = gkn tanh(knh), retaining the same ideal boundary wavenumber. This checks the long-wave approximation only; it does not correct a real harbour entrance. The displayed parameter limits keep knh below 0.283.

For a rectangular closed basin with appreciable width W, additional transverse modes have Tnm = 2/{√(gh)√[(n/ℓ)² + (m/W)²]}, where n and m are nonnegative integers and not both zero. These two-dimensional modes are not plotted here. Irregular bathymetry, connected basins and entrances usually require more detailed analysis.

What the resonance curve actually calculates

The curve is the exact steady harmonic solution of a linear single-degree-of-freedom modal oscillator with constant positive damping. Let q be a modal displacement coordinate and F0 the forcing amplitude per unit modal mass. F0 is not an offshore water elevation. The static displacement for constant forcing is qstatic = F0/ωn².

Forced modal oscillator
q¨+2ζωnq˙+ωn2q=F0cos⁡(ωt)
Frequency ratio
r=ωωn=TnTforcing
Normalized response amplitude
M=|q|qstatic=1(1−r2)2+(2ζr)2
Phase lag and response
δ=atan2(2ζr,1−r2),qqstatic=Mcos⁡(ωt−δ)
Displacement response peak
rpeak=1−2ζ2,ζ<12
Peak response magnitude
Mpeak=12ζ1−ζ2

At r = 1 the phase lag is 90°, but the displacement-response peak lies slightly below this frequency when damping is nonzero. Frequency ratio and period ratio are reciprocals, not interchangeable axes. The plot holds its vertical scale fixed at 0–11 to make damping comparisons honest; all allowed peak values fit.

This is a steady forced response after transients have decayed. It does not show startup or ring-down. Damping ζ is dimensionless and prescribed; it is not a calibrated bottom-friction coefficient or an energy-loss percentage. In a harbour, radiation through the entrance and several dissipative mechanisms contribute to an effective damping that may depend on frequency. This curve cannot predict them.

CHAPTER 28 / MEASURING WAVES

From a sensor signal
to a wave estimate.

An instrument records a physical quantity: distance, pressure, acceleration or position. Turning that signal into wave height, period and direction requires a measurement model as well as careful processing. A smooth record is not necessarily an accurate record, and a precise sensor can still miss part of the wave spectrum.

Choose the quantity before the instrument.

A fixed surface gauge measures the changing elevation at one location. A submerged pressure sensor measures pressure fluctuations that depend on both the surface motion and the sensor's depth. A wave buoy measures its own motion, which follows the surface only within the limits of its mechanical and electronic response.

The required frequency band matters. An instrument and processing chain suitable for sea and swell may suppress the slower infragravity and seiche motions in Chapter 26 and Chapter 27. Record length, calibration, sampling and the site all need to match the question being asked.

One elevation record does not reveal direction.

A single elevation record supports a frequency spectrum and wave-by-wave statistics. It cannot uniquely determine the directional distribution. Directional observations need additional information, such as suitably spaced sensors or multiple buoy-motion components with a known orientation.

Reported directions also require a convention. This website uses clockwise angles from north describing propagation toward. When a data provider reports where waves come from, convert explicitly: θtoward = (θfrom + 180°) modulo 360°. Instrument axes and magnetic versus true north must be resolved before comparing observations.

What common instruments observe
InstrumentMeasured quantityImportant checks
Resistance or capacitance gaugeWater contact along a fixed probe, calibrated to elevationCalibration, wetting, fouling, mounting disturbance and measurement range
Downward-looking radar or acoustic gaugeDistance to the surface from a reference positionFootprint, obstructions, spray, loss of returns and reference-platform motion
Submerged pressure sensorTotal or gauge pressure at a known submerged elevationAtmospheric reference, density, depth, frequency-dependent pressure response and noise
Wave buoyMotion inferred from accelerometers, attitude sensors or positioning signals, depending on the systemHull and mooring response, orientation, processing filters and the supported frequency band

What reaches a submerged pressure sensor?

A small, regular wave has surface amplitude a = 0.30 m (height H = 0.60 m). Move the sensor, change the depth or change the period. The same linear-wave solution generates the diagram, record and response curve. The pressure record below is noise-free; the separate noise calculation illustrates the sensitivity of the inverse correction.

A · Sensor position and coordinates
z = 0 is the mean surface; the bed is z = −h. Sensor height above the bed is b, so zs = b − h. The wave outline is schematic.
B · Two quantities, one common scale
Blue: surface elevation η. Coral: pressure-equivalent elevation p′/(ρg). This is wave-induced pressure after removal of atmospheric and mean hydrostatic contributions.
C · Pressure amplitude response by frequency
Navy: Kp. Coral dot: selected wave. The vertical axis is logarithmic. Shading below Kp = 0.10 marks the demonstration's excluded inversion range; it is not a universal field cutoff.

ρ = 1025 kg/m³; g = 9.81 m/s². Noise is a prescribed RMS pressure in a narrow band around the chosen frequency, not a sensor specification or a generated random time series. Fixed scales support direct comparisons.

Try the measurement before trusting the correction.

  1. Shorten the period while keeping depth and sensor position fixed. The pressure amplitude falls more strongly than the surface amplitude. The surface wave has not lost energy; the instrument is observing a different quantity below the surface.
  2. Raise the sensor toward the surface. Kp increases and the required inverse gain decreases. The pressure-response benefit must be balanced against practical exposure, mounting and possible emergence, which are not simulated.
  3. Increase the prescribed pressure noise. The clean records do not change, but the corrected noise rises. Dividing by a small Kp amplifies the desired signal and the noise together; it does not improve their signal-to-noise ratio at that frequency.

A transparent reference calculation.

For h = 10 m, b = 1 m and T = 8 s, solve (2π/T)² = gk tanh(kh) for k. This gives k ≈ 0.08862 m⁻¹ and Kp ≈ 0.70745. A 0.30 m surface amplitude produces about 0.21224 m of pressure-equivalent amplitude, or 2134 Pa of wave-pressure amplitude. The inverse amplitude gain is about 1.4135.

Surface variance is a²/2 = 0.045 m². Pressure-equivalent variance is Kp²a²/2 ≈ 0.02252 m². Recovering the surface variance requires division by Kp², not Kp. For 10 Pa RMS noise concentrated around this frequency, the equivalent surface-noise RMS is about 0.001406 m after correction. These are idealized linear calculations, not a claim about a particular instrument's accuracy.

Sampling and processing are part of the instrument.

A sample rate fs has a Nyquist frequency fs/2. An analogue or instrument-level anti-alias response is needed before sampling; a later digital filter cannot recover frequencies already folded into the record. A record of duration D has Fourier-bin spacing 1/D, while windowing and averaging affect the ability to distinguish peaks and the uncertainty of estimated spectral density. Chapter 10 develops these distinctions.

Keep the raw observations and document units, timestamps, sensor elevation, water depth, calibration and processing settings. Inspect spikes, missing samples, clipping, drift, motion of the mounting platform and changes in operating conditions. Removing a mean or trend can be appropriate, but an aggressive high-pass filter may also remove a real long-period signal.

For a 100 s oscillation, a 20-minute record contains only 12 cycles. A four-second wave contributes 300 cycles over the same duration. Neither count alone specifies uncertainty, but the contrast explains why slow processes require a different observation strategy. Avoid filling long gaps with smooth interpolation and then treating the result as measured wave information.

Pressure transfer, spectra and inverse-noise amplification

Take z positive upward from mean water level, h > 0 and 0 ≤ b ≤ h. For a fixed sensor at zs = b − h, linear theory gives the wave-pressure departure p′ from the mean hydrostatic state. The experiment assumes a uniform-depth, nonbreaking, small-amplitude wave with no ambient flow.

Angular frequency and dispersion
ω=2πT,ω2=gktanh⁡(kh)
Pressure-response coefficient
Kp(f,zs)=cosh⁡[k(zs+h)]cosh⁡(kh)=cosh⁡(kb)cosh⁡(kh)
Elevation and wave-induced pressure
η(t)=acos⁡(ωt),p′(t)=ρgKpacos⁡(ωt)
Pressure-equivalent elevation
q(t)=p′(t)ρg
Pressure-equivalent spectrum
Sq(f)=Kp2Sη(f)
Pressure spectrum
Sp′(f)=(ρg)2Kp2Sη(f)
Corrected surface-elevation spectrum
Sη,estimated(f)=Sp′,measured(f)(ρg)2Kp2
Narrowband measurement-noise amplification
σnoise,η≈σnoise,pρgKp

Kp is dimensionless and positive: Kp = 1 at the mean surface, and Kp = 1/cosh(kh) at the bed. As frequency approaches zero, Kp approaches one. Sq and Sη have units m²/Hz; Sp′ has units Pa²/Hz. The pressure transfer represents a depth-dependent response, not dissipation of wave energy.

For additive pressure noise uncorrelated with the wave signal, the estimated surface spectrum contains the true surface spectrum plus the noise spectrum divided by (ρgKp)². Inverting the response alone cannot distinguish them. Broadband noise must be treated frequency by frequency and integrated across the retained band; the single-band RMS calculation here is not a broadband correction.

The demonstration reports corrected signal amplitude only when Kp ≥ 0.10. This illustrative cap of ten on amplitude gain is not an accepted universal quality standard. A practical cutoff depends on noise, calibration, depth uncertainty, signal strength and the purpose of the measurement. Frequencies discarded as unreliable contain unknown wave variance; replacing them by zero is a processing choice, not proof that no waves exist there.

Buoy response and the limits of directional reconstruction

In an ideal heave measurement, vertical acceleration and displacement Fourier coefficients obey âz = −ω²η̂. Consequently Saz = ω⁴Sη and the inverse displacement spectrum divides by ω⁴. The negative sign describes a 180° phase difference; it disappears in the spectrum. Low-frequency errors become especially important when integrating acceleration, so orientation correction and well-characterized filtering are essential. This ideal identity is not a complete transfer function for any buoy.

Real hulls, moorings and electronics have frequency-dependent responses. Position-based and acceleration-based systems therefore need their own calibration and processing descriptions; an arbitrary generic buoy correction should not be applied to all observations.

Directional estimates from multiple motion components or arrays provide finite directional information. Reconstructing a smooth directional spectrum requires assumptions and does not uniquely reveal every crossing-wave component. Check array geometry, timing, compass alignment and the provider's conventions. Compare data and models over compatible frequency bands and averaging periods, as discussed in Chapter 21.

CHAPTER 29 / WAVE-DATA QUALITY CONTROL

Keep the evidence.
Question the result.

Quality control makes observations interpretable and their limitations visible. It combines instrument knowledge, reproducible tests and scientific review. The objective is to identify unreliable information without erasing real extremes, unusual sea states or slow motions merely because they are unexpected.

Quality assurance prevents; quality control evaluates.

Calibration, deployment design, maintenance and configuration establish the conditions for reliable measurement. QC then evaluates the observations and derived products. A check cannot reconstruct a clipped crest, establish an unknown sensor elevation or compensate for an undocumented clock error merely by assigning a colour.

A passing test means that its particular criterion was satisfied. It does not prove that a record is correct. A suspect flag means that further attention is needed; an unusual storm wave may deserve that attention just as much as a sensor fault.

Evaluate the right level.

Separate individual samples, complete bursts, spectral bins and successive sea-state summaries. A missing sample is not the same as a missing sea state. An unreliable high-frequency bin need not invalidate every lower-frequency component, but it may prevent release of a parameter that depends on the full band.

Preserve the raw observations, individual test results and processing history. Real-time screening and later scientific review answer different questions with different information. Keep both versions rather than rewriting the original evidence.

01 · AcquireCalibration, geometry, clock, raw packets
02 · Inspect samplesIntegrity, missing values, waveform faults
03 · Inspect productsResponse, spectra, moments, direction
04 · Review and releaseContext, uncertainty, flags, audit trail

A reference framework, with an explicit scope.

The principal wave-specific crosswalk below follows IOOS QARTOD In-Situ Surface Wave Data, version 2.1 (2019). Its hierarchy applies within that framework; it is not a declaration that every instrument everywhere must run every test. Tests 17–18 concern supplied spectra. Instrument-specific tests need their corresponding channels.

Reference roles: standards, guidance and educational extensions
ReferenceRole in this chapterBoundary of the claim
IOOS QARTOD waves v2.1, 2019Wave-test index and flag vocabularyOperator-defined thresholds and sensor applicability remain essential.
IOC Manuals and Guides 26, 1993Historical international oceanographic validation guidance; wave-data appendixExamples require adaptation to present instrumentation and local conditions.
WMO-No. 8, 2024 editionObservation and quality-management framework; calibration and intercomparisonThe 2026 material identified as provisional is not treated as a finalized replacement here.
ISO/IEC 17025:2017International standard concerning testing/calibration laboratory competenceIt does not supply a universal wave-spike threshold. This website claims no accreditation or clause-by-clause compliance.
CF conventions 1.10, §§3.4–3.5Published data-exchange conventions for ancillary data and flagsMetadata conformance does not establish physical accuracy; the chosen convention version must be declared.

The 32-method guide is an original engineering synthesis, with practical implementation notes and cautions. It extends beyond the QARTOD test index to spectral mathematics, directional consistency, processing reproducibility and delayed review. The interactive algorithms are deliberately small teaching examples, not a certified implementation of all referenced procedures.

All 21 QARTOD wave tests: reference crosswalk

Numbers and priority labels refer to v2.1, Table 3-2 and §§3.3.1–3.3.3. Names below are abbreviated. Follow each link to the original implementation guidance in this lesson. ST means a short sampled record; LT means a series of wave products.

TestTopicManual hierarchy
1Echo strengthStrongly recommended
2Echo correlationStrongly recommended
3Acoustic noiseStrongly recommended
4Signal/noiseStrongly recommended
5Pressure/surface trackingStrongly recommended
6Acoustic velocity rangeStrongly recommended
7Acoustic velocity statisticsStrongly recommended
8Usable sample countStrongly recommended
9Short-record gapsStrongly recommended
10Short-record spikesStrongly recommended
11Short-record rangeStrongly recommended
12Segment shiftsSuggested
13AccelerationStrongly recommended
14Check factorsStrongly recommended
15Long-term mean/SDStrongly recommended
16Long-term flat linesRequired
17Operational frequency bandRequired
18Low-frequency energyRequired
19Bulk-value rangeRequired
20Bulk-value changeRequired
21Neighbour comparisonSuggested

Tests 1–8 are acoustic-instrument diagnostics. Their presence here does not add a wave-driven-current lesson. The laboratory below implements its own limited sample tests, not this entire crosswalk.

Fault laboratory: a flag is evidence, not a repair.

The clean reference is a prescribed two-component signal sampled at 2 Hz for 128 s. Choose an injected fault and inspect the unchanged received record, diagnostic residuals and spectral consequences. The reference is known only because this is a synthetic experiment; field QC rarely has such a truth signal.

A · Received record and known reference
Navy: received values. Dashed grey: clean reference. Red crosses: hard range failures; orange circles: local-test suspects; grey bands: missing samples. Markers are not replacement values.
B · Local residual and threshold
ei = |xi − (xi−1 + xi+1)/2|. Dashed coral: selected threshold. Missing neighbours and record endpoints are not evaluated.
C · Variance-density consequence
Blue: clean reference. Coral: received record or an explicitly labelled gap-filled copy. Both are demeaned, rectangular-window, one-sided periodograms. No flagged sample is silently deleted.

Teaching rules: |η| > 1.5 m → fail; local residual > threshold → suspect; a run of ≥8 samples with span ≤10⁻6 m → suspect. A half-record mean difference >0.30 m triggers a separate burst-level review. These limits are not published operational standards.

Exactly what this laboratory computes—and what it misses
Synthetic reference record
ηref(t)=0.40cos⁡(2π⋅0.125t)+0.15cos⁡(2π⋅0.1875t+0.7)
Sampling and record duration
Δt=0.5 s,N=256,D=NΔt=128 s
Spectral-bin resolution
Δf=1D
Reference variance
Vref=0.402+0.1522=0.09125 m2
Discrete Fourier transform
Xk=∑j=0N−1(xj−x―)exp(−i2πjkN)
One-sided variance-density periodogram
Sk=ck|Xk|2fsN
Endpoint convention
ck=1 (DC and Nyquist),ck=2 (elsewhere)
Variance conservation
∑kSkΔf=1N∑j(xj−x―)2

Samples occur from 0 to 127.5 s; the DFT record duration is NΔt = 128 s. The displayed spectrum uses the complete record, not only passing points. This makes contamination visible without compressing the time axis by deleting observations. A direct DFT is used for this small example; an FFT computes the same coefficients more efficiently.

Missing samples receive 9; a hard range failure receives 4; otherwise a detected local anomaly receives 3. A valid point with an unavailable three-point test receives 2; other points receive 1 for this limited test set. A segment shift is a separate record-level review, not automatically assigned to all samples. This is an explicit teaching aggregation policy, not a complete operational flag algorithm.

The flat-run detector flags an entire run retrospectively. The centred residual needs a future neighbour. These are delayed-inspection methods, unlike a strictly causal sample-by-sample service. The optional linear fill is allowed only for bounded gaps of at most eight samples; there is no extrapolation. Raw values and flags stay unchanged. A four-second interpolation can erase much of a short wave even though the curve looks plausible.

Clipping can pass all three local tests. A mean offset may also escape a spike threshold. Lowering a threshold can flag ordinary wave curvature. These are deliberate demonstrations of missed faults and false alarms, not defects hidden by the lesson. The spectral panel is omitted when a complete analysis copy is unavailable.

Read indicators in their scientific context.

A spike raises broadband variance; clipping can lower it. A shift can add low-frequency content. A gap cannot be made harmless simply by removing its missing timestamps. Compare the two spectra and the variance budget, then return to the received waveform to understand the cause. The largest change in a plot is not necessarily the most consequential error for a particular product.

A comprehensive method guide.

Each method below identifies its inputs, an implementation approach and a reason to avoid automatic overconfidence. Together they cover the observation lifecycle. They are a planning and review guide, not evidence that all methods have been run on an uploaded field dataset.

01 · Purpose, coverage and acceptance plan

Input level: Before deployment

Method. Define the quantity, frequency band, location, averaging interval, latency and acceptable uncertainty. Specify which tests operate on samples, bursts, spectral bins or successive sea states. Assign an owner to each exception and a deadline for review.

Interpretation and limitations. Document a test as not applicable when its inputs do not exist; do not silently call it passed. Requirements for storm detection and routine climatology may differ. A single checklist cannot establish fitness for every use.

02 · Calibration and traceability

Input level: Before and after deployment

Method. Keep calibration coefficients, dates, sensor serial numbers, reference instruments and uncertainties. Check zero, span and response across the intended operating range; compare post-recovery results with the original calibration before proposing a correction.

Interpretation and limitations. A new offset does not prove that a constant correction applies throughout a deployment. Drift may be gradual, abrupt or caused by fouling that changes frequency response. Preserve the original measurements and the correction model separately.

03 · Deployment geometry and health

Input level: Every deployment and service event

Method. Record position, water depth and datum, sensor elevation, orientation, mooring configuration, firmware and sampling settings. Review tilt, battery, temperature, diagnostics, fouling and service logs alongside wave products.

Interpretation and limitations. A moved sensor can remain within a numerical range while representing a different water depth or location. A pass flag from software cannot replace inspection of mounting, orientation and environmental exposure.

04 · Message integrity and provenance

Input level: Every received message

Method. Validate packet length, checksum where provided, schema, encoding, station identity, units and sequence numbers. Detect duplicate or truncated messages before joining them into an observation. Retain receipt time separately from measurement time.

Interpretation and limitations. A checksum supports transmission integrity, not physical accuracy. Retransmitted observations should be reconciled by identifier and version rather than counted as additional independent samples.

05 · Time stamps, order and clock drift

Input level: Raw time axis

Method. Inspect Δti=ti−ti−1 against the declared sample interval, including duplicates, reversals and clock resets. For bursts, verify start/end times, time zone and the relationship between sample count and duration. For arrays, check relative clock alignment.

Interpretation and limitations. A small timing bias can change phase and inferred direction without noticeably changing a single-channel spectrum. Sorting a corrupted series does not explain why its time axis failed.

06 · Gaps, completeness and longest outage

Input level: Samples and bursts

Method. Create an expected sample grid from the documented sampling plan. Count missing samples, separate isolated losses from consecutive runs, and report both missing fraction and longest gap in seconds. Compare usable duration with the analysis requirements.

Interpretation and limitations. Equal missing percentages can have very different effects: ten isolated losses and one long outage are not equivalent. An absent value remains missing, never zero. A failed derived product can coexist with many individually usable samples.

07 · Hard range and contextual range

Input level: Calibrated channels and bulk values

Method. Use separate limits for sensor capability and environmental plausibility. Check units and offsets before applying them. Mark impossible or saturated values distinctly from unusual values that merely exceed a site history.

Interpretation and limitations. A storm can exceed a historical envelope while remaining physically real. Ranges need a documented basis and version. Neither a universal maximum wave height nor a single height-to-depth ratio is appropriate to all sensors and settings.

08 · Isolated spikes and robust outliers

Input level: Raw channel values

Method. Compare a point with its local neighbourhood and with a robust distribution estimate. A simple illustrative residual is ei=|xi−(xi−1+xi+1)/2|. Inspect adjacent points, raw diagnostics and waveform continuity before classifying a large residual.

Interpretation and limitations. This centred test needs the next sample and is delayed by one sample. A spike can contaminate its neighbours’ residuals. Large crests, breaking and rapid physical change can also trigger a detector; clipping all statistical outliers biases extremes downward.

09 · Flat lines and repeated values

Input level: Raw samples and successive products

Method. Search for runs whose maximum minus minimum is within a tolerance tied to resolution. State the run length in samples and seconds. Compare the apparent freeze with other channels and telemetry counters.

Interpretation and limitations. Quantized low-energy data or rounded peak periods can legitimately repeat. The raw-sample illustration here is distinct from the QARTOD long-term bulk-parameter flat-line test. A repeated zero should be investigated rather than exempted automatically.

10 · Clipping and quantization

Input level: Instrument output

Method. Look for repeated occupancy at upper/lower rails, long horizontal plateaus at crests or troughs, and unusually coarse increments. Compare observed rails with electronic and configured ranges. Inspect the histogram as well as the waveform.

Interpretation and limitations. Clipping often reduces variance and creates harmonics while leaving every recorded sample inside a nominal range. Increasing a software range limit cannot recover the lost crest information.

11 · Mean shifts, drift and segment changes

Input level: Within a burst and across deployments

Method. Divide a sufficiently long record into comparable segments and inspect means, variances and their changes. Use a two-sided difference, |μj+1−μj|, with a documented limit. Check service times, depth changes, timing resets and low-frequency content.

Interpretation and limitations. Segments must span enough waves to avoid confusing phase or grouping with an offset. A genuine long wave or water-level change may alter the mean; removing it is a scientific processing decision, not an automatic QC repair.

12 · Acceleration and impulse checks

Input level: Buoys with appropriate raw channels

Method. Compare acceleration with calibrated sensor limits after establishing axes and gravity treatment. Use recorded diagnostics to distinguish impacts, attitude problems and saturation. Record whether the channel contains gravity or has already been compensated.

Interpretation and limitations. A finite difference of sampled elevation is a noisy proxy, not an interchangeable accelerometer observation. Manufacturer limits and filtering affect what an acceleration threshold means.

13 · Acoustic amplitude, correlation and noise

Input level: Acoustic instruments only

Method. Review signal strength, echo correlation and acoustic noise for each contributing beam/bin and surface-tracking return. Confirm whether values are linear power, amplitude, counts or decibels before computing a ratio or applying a threshold.

Interpretation and limitations. SNR in dB is not formed by dividing two dB values. Low correlation or poor tracking can invalidate a wave estimate even when its final height looks plausible. Instrument-specific diagnostics must follow the relevant manufacturer definitions.

14 · Pressure and surface tracking plausibility

Input level: Pressure or acoustic surface return

Method. Compare mean level and variability with depth, atmospheric referencing, sensor elevation and plausible water-level excursions. Check for lost returns, bottom/structure echoes, emergence and sudden changes of reference distance.

Interpretation and limitations. Do not assume pressure is hydrostatic at every wave frequency. The frequency response and correction limits in Chapter 28 still apply after these basic checks.

15 · Wave-channel velocity diagnostics

Input level: Acoustic wave-estimation channels

Method. For instruments deriving wave information from orbital-velocity channels, inspect each selected beam/bin for range errors and anomalous mean or variability, using the documented coordinate frame and valid-sample mask. Count surviving samples after diagnostic rejection.

Interpretation and limitations. These are instrument-channel checks for wave estimation, not a treatment of wave-driven currents. Do not apply one instrument’s velocity thresholds to a different measurement geometry or convert beam statistics to earth axes without the required transformation.

16 · Sensor response and usable band

Input level: Before inversion or spectral integration

Method. Apply the documented transfer function only over the supported band. Inspect the inverse gain and noise contribution as functions of frequency. Track depth and response changes, and report which frequencies were excluded.

Interpretation and limitations. A correction can amplify noise while producing a smooth-looking spectrum. Unsupported low-frequency buoy response and weak high-frequency pressure response require different investigations; the same cutoff cannot be assumed for every instrument.

17 · Aliasing and sample-rate consistency

Input level: Acquisition and processing

Method. Compare sample rate, instrument response and anti-alias filtering with the highest retained frequency. Check that any downsampling was preceded by suitable filtering. Report acquisition rate separately from output interval.

Interpretation and limitations. A sampled sinusoid at 0.8 Hz and another at 0.2 Hz can coincide at a 1 Hz sample rate. Filtering the already aliased record cannot decide which frequency was originally present.

18 · Spectral validity and variance closure

Input level: Each spectrum

Method. Check finite values, ordered frequencies, positive bin widths and nonnegative variance density within a stated numerical tolerance. Verify one-sided normalization, window-energy correction and the correspondence between integrated density and the processed record variance.

Interpretation and limitations. With a rectangular, demeaned complete record, Parseval provides an exact numerical check. A windowed periodogram closes against its window-weighted quantity, not automatically the unwindowed variance. Frequency density and angular-frequency density require a Jacobian.

19 · Low/high-frequency contamination and noise floor

Input level: Spectral shape

Method. Inspect energy outside the validated instrument band, unusually large low-frequency contributions, narrow electrical/mechanical peaks and high-frequency plateaus. Compare changes against instrument diagnostics and the unfiltered record.

Interpretation and limitations. A low-frequency peak can be a real seiche or infragravity process. Classify it as outside a sea-swell product when appropriate, rather than declaring that the physical observation is false. Negative noise-subtracted bins need explicit handling, not silent clipping.

20 · Stationarity and segmentation

Input level: Spectral-analysis windows

Method. Compare segment means, variances and spectral shapes before pooling a record. Record storms or transitions that make a stationary sea-state approximation weak. Choose duration to balance enough cycles against changing conditions.

Interpretation and limitations. Shorter segments reduce frequency resolution; longer ones can mix different states. Nonstationarity is often a limitation of the summary product rather than a malfunctioning sensor.

21 · Moments and derived-parameter consistency

Input level: Bulk wave products

Method. Recompute mn=ΣfiⁿSiΔfi over the declared valid band. Check Hm0=4√m0, Tm01=m0/m1 and Tm02=√(m0/m2) where denominators are positive. Identify Tp from the stated peak method, and keep H1/3 distinct from Hm0.

Interpretation and limitations. Exclude unsupported products rather than divide by zero in calm or empty-band records. A fitted peak and the largest discrete spectral bin can give different Tp values; neither algorithm should be left undocumented.

22 · Rate of change and plausible evolution

Input level: Successive sea-state products

Method. Compute changes using the actual interval: |Xj−Xj−1|/Δt. Compare with limits developed for that variable and location, and inspect storms, rapidly changing winds, telemetry gaps and processing changes.

Interpretation and limitations. Peak period and peak direction can jump when two valid spectral peaks exchange dominance. Do not apply a smoothness rule for integrated wave height blindly to these discrete peak descriptors.

23 · Climatological and robust context

Input level: Long-term records

Method. Compare observations with seasonally and directionally relevant historical distributions, using a sufficiently representative baseline. Inspect changes in instrument, depth and processing before merging records. Track flags during known storms separately.

Interpretation and limitations. An extreme outside previous experience is not automatically wrong. A contaminated baseline can normalize faults, while a narrow fair-weather baseline can reject the very events the observing programme needs.

24 · Neighbour, duplicate and model comparisons

Input level: Independent observations

Method. Match timestamps, averaging periods, frequency bands and direction conventions. Prefer collocated independent instruments for direct comparisons; allow for propagation, sheltering and depth transformation when comparing sites. Compare residuals and uncertainty, not only correlation.

Interpretation and limitations. A model is supporting evidence, not truth. Two sensors can share the same bias, and nearby sites may observe different seas. Disagreement should normally trigger investigation before automatic rejection.

25 · Directional check factors and channel consistency

Input level: Directional instruments

Method. Use the provider’s definition of the check factor, channel scaling and depth dependence. Inspect its behaviour across the valid band and over time; compare deviations with fouling, mooring changes and sensor health.

Interpretation and limitations. Do not combine unscaled heave spectra in m²/Hz with raw pitch/roll spectra in rad²/Hz and expect a universal dimensionless ratio. A single copied formula without response and channel definitions can be dimensionally invalid.

26 · Directional range, wraparound and convention

Input level: Directions and directional products

Method. Validate the encoded angle before normalizing it. Resolve true/magnetic reference and from/toward meaning. Use the signed circular difference atan2[sin(θ2−θ1),cos(θ2−θ1)] rather than ordinary subtraction.

Interpretation and limitations. 359° to 1° is a two-degree change, not 358°. A numerical wrap can hide an invalid raw value, so record the original. Mean direction becomes poorly defined when directional concentration is small.

27 · Coherence and spectral-matrix feasibility

Input level: Multichannel spectra

Method. For consistent spectral estimates, inspect coherence |Sxy|²/(SxxSyy) and Hermitian symmetry, and check that the spectral matrix is positive semidefinite within numerical tolerance. Confirm channel synchronization and compatible averaging.

Interpretation and limitations. Coherence should lie between zero and one for valid estimates with positive denominators. Near-zero energy makes ratios unstable. Testing each channel separately cannot detect every inconsistent cross-spectral combination.

28 · Directional coefficients and variance conservation

Input level: Frequency–direction spectra

Method. Check coefficient conventions and dimensions, directional normalization and frequency/direction marginals. If a1,b1,a2,b2 are normalized trigonometric moments, r1=√(a1²+b1²) and r2=√(a2²+b2²) cannot exceed one. Integrate angular density in radians.

Interpretation and limitations. These bounds are necessary, not sufficient for joint realizability. A truncated Fourier reconstruction may become negative despite admissible coefficients. Distinguish reconstruction artefacts from corrupted coefficients; do not silently delete negative lobes and change variance.

29 · Per-test flags and product aggregation

Input level: Each processing level

Method. Store results for each test plus an explicitly defined aggregate. Keep sample, burst, bin and product flags separate. Preserve the real-time record and add a versioned reviewed result. Flag a missing value as missing, not failed or passed.

Interpretation and limitations. Flag numbers are categories, not a quality score. Taking their arithmetic mean is meaningless. A good aggregate must not conceal an unevaluated required test or a missing input; document the precedence policy.

30 · Repairs, interpolation and reprocessing

Input level: Reviewed derived copies

Method. Retain original values and mark every replacement with method, bounds, rationale and software version. Compare spectra and moments before and after processing. Withhold products when gaps or corruption exceed the validated repair capability.

Interpretation and limitations. Interpolation invents estimates, not observations. It often suppresses short-period variance; zero-filling adds discontinuities. A repaired copy must never overwrite the evidence used to make the decision.

31 · Uncertainty, false alarms and missed faults

Input level: Acceptance validation

Method. Use controlled faults and independently reviewed real events to measure detection and false-alarm performance. For a known labelled set, report both missed faults and falsely flagged valid data, with the definition of each denominator. Evaluate uncertainty in retained parameters.

Interpretation and limitations. Synthetic faults only cover the failures injected. Passing this chapter’s tests does not validate an operational QC system, and aggressive screening can produce a deceptively quiet but biased archive.

32 · Release, audit trail and metadata

Input level: Distribution and archive

Method. Publish units, coordinates, time convention, valid band, calibration, sampling, window/averaging, thresholds, flag meanings, processing version and relevant uncertainty. Keep raw, provisional and reviewed products distinguishable and link flags to their data variables.

Interpretation and limitations. Metadata formats support exchange; they do not certify scientific quality. Provide a reason for an unavailable product and retain review decisions so another analyst can reproduce the result.

Two valid behaviours that can look like faults.

D · Direction crosses north
Clockwise, propagation toward. Blue: first direction; coral: second. The short arc and signed result use circular arithmetic.

E · Two spectral peaks exchange dominance
Two equal-width, idealized spectral bins keep total variance fixed. Moving energy between them can make the selected peak period jump.

The compass demonstrates arithmetic, not retrieval uncertainty. The peak example represents two discrete bins, not a fitted ocean spectrum. At an exact tie, this lesson reports both peak periods instead of pretending that one is uniquely defined.

Flags, publication and a reproducible decision.

1 · PassTest criterion satisfied2 · Not evaluatedRequired evidence absent3 · SuspectReview needed4 · FailCritical criterion failed9 · MissingNo observation

Keep a per-test result even when a summary flag is supplied. Separate flag categories from uncertainty estimates and from an analyst's final acceptance decision. Record why a test was skipped, whether a product was withheld, and which replacements were introduced. Never average flag codes or interpret a higher numerical flag as a larger measurement error.

Minimum reproducibility record for a QC decision
RecordInformation to preserve
Observation identityStation, sensor serial, deployment, location, depth/datum, sample times, time zone, units and source-file identifier
Measurement chainCalibration, response function, sampling/anti-alias settings, orientation, firmware and diagnostic channels
AnalysisBand limits, detrending, window, segment length/overlap, averaging, spectral normalization and parameter definitions
Test configurationTest name/version, input level, units, thresholds, tolerance, inequality boundary and not-applicable reason
Result and actionPer-test and aggregate flags, suspect/fail/missing fractions, longest gap, withheld parameters and uncertainty
Review historyOriginal values, replacement mask, method, reviewer, rationale, timestamps and versioned outputs
An illustrative machine-readable flag description

For a netCDF variable called wave_height, an ancillary_variables attribute can name its QC variables. An enumerated QC variable can declare flag_values = 1, 2, 3, 4, 9 and corresponding flag_meanings = "pass not_evaluated suspect fail missing". Attribute types must match the selected format/convention rules. Store thresholds, software version and individual test variables as well; a five-value list alone does not explain how quality was assessed.

A CSV export can carry the same scientific information through explicitly named columns and a versioned metadata file. Do not assume a spreadsheet colour will survive export. The definitions and processing history must travel with the values.

Before releasing a dataset, answer four questions.

  1. What was actually measured? Establish the quantity, geometry, calibration, time axis and usable band.
  2. Which evidence supports each decision? Inspect raw values, diagnostics, spectra and relevant independent observations.
  3. What information was changed or lost? Identify missing, rejected, interpolated, filtered and unsupported parts of the record.
  4. Can another analyst reproduce the release? Supply configuration, code version, flags, uncertainty and review history.
WAVE LAB / QUICK ESTIMATES · CONTROLLED EXPERIMENTS

Ask a wave.
Test an idea.

Explore synchronized 1D–2D–3D wind-wave generation, shoaling, refraction and reflection laboratories. Each experiment distinguishes its analytical physics, rendering assumptions and limits.

Educational estimates, not forecasts or design values. These tools use idealized conditions. They do not ingest weather observations, surveyed bathymetry or calibration data. No wave-driven currents are included. All calculations run locally; external references are optional links.

01 / LINKED WIND–WAVE LAB · 1D · 2D · 3D

From wind to waves: three synchronized views.

Explore sustained over-water wind at 10 m above the water surface, effective fetch and wind duration using a historical empirical wave-growth estimate. Compare the estimated Hm0 and Tp with one synthetic irregular sea represented by a 1D virtual gauge, a 2D surface-elevation map and a rotatable 3D surface. Match 2D aligns the spatial wave patterns for direct inspection.

Statistical versus instantaneous: The bulk height and period are properties of the idealized estimated sea state. The animated waves are a deterministic finite-component realization, not predicted or measured individual waves. The growth equations assume a steady spatially uniform wind, no incoming swell and an appropriate deep-water validity range.

The wind-wave simulator loads when this Wave Lab section is in view.

Practical interpretation: Check wind averaging period, height convention, fetch exposure, storm evolution, depth applicability and measurement representativeness before using bulk estimates downstream. No wind-field grid, tropical cyclone, squall model or measured time series is simulated here.

02 / LINKED WAVE LAB · 1D · 2D · 3D

One wave. Three ways to understand it.

Investigate a regular wave propagating across an idealized gentle seabed slope. The 3D surface, 2D transect and 1D gauge record are calculated from the same linear-wave dispersion and lossless energy-flux relationships, and share one time and phase. Unlike a three-dimensional hydrodynamic model, the 3D scene is a visualization of an idealized wave field.

Interpretation: The wave height shown is the regular-wave crest-to-trough height H, not spectral significant wave height Hm0. The simplified depth-limited breaking indicator is an applicability screen, not a numerical wave-breaking solver. No diffraction, current field, turbulence or bed friction is calculated.

03 / LINKED REFRACTION LAB · 1D · 2D · 3D

When the seabed changes, wave crests turn.

Follow one prescribed, regular wave approaching straight and parallel depth contours through a synchronized 1D virtual gauge, 2D wavefront and ray map, and 3D travelling-wave surface. Compare the Match 2D (top) view with the oblique surface before changing wave height, period or approach angle.

Reading the views: The white plan-view phase contours correspond to pale 3D crest ridges; amber curves indicate wave rays, not water-particle paths. The 3D rendering is a height-exaggerated visualization of a linear regular-wave solution, not a 3D hydrodynamic solver. No breaking, currents or wave dissipation are calculated.

The interactive refraction studio loads when you reach this section. The same verified code is used in Chapter 4.

04 / LINKED REFLECTION LAB · 1D · 2D · 3D

Reflection, connected across three views.

Explore an idealized incident and reflected regular-wave field with a 3D perspective surface, 2D wave-interference map and a 1D virtual gauge. The views share one analytical solution, one wave phase and one control set. Compare perfect, partial and oblique reflection, and move the gauge to find nodes and antinodes.

Read the physics: The 3D surface is a perspective rendering of a linear solution with an infinite straight boundary and uniform water depth—not a three-dimensional fluid solver. Amplitude reflection R is distinct from the reflected normal energy-flux fraction R2. For transmission and dissipation concepts, see Chapter 24.

METHODS / PROVENANCE / NEXT STEPS

Grounded in published science.

The Wave Lab’s explanatory text, interface, JavaScript and diagrams were created for Earth in Motion. Published scientific relationships are credited below. No textbook pages, publisher illustrations, external calculator source code, logos or Celeris code are bundled with these tools. Attribution identifies intellectual sources; it does not imply endorsement or replace permission to reuse protected material.

  1. Sorensen, R. M. (2006), Basic Coastal Engineering, 3rd ed., Springer. §6.7, pp. 183–185, Eqs. 6.39–6.42: SPM–JONSWAP deep-water growth; §2.7, pp. 35–38: standing waves and reflection. Equations are independently implemented; prose and illustrations are not reproduced.
  2. U.S. Army Coastal Engineering Research Center (1984), Shore Protection Manual, Vol. I. Historical source of the growth method. USACE catalogue and manual ↗. This release uses the SPM–JONSWAP formulation with fractional exponents, as disclosed above; it is not labelled as an implementation of the newer Coastal Engineering Manual.

Explore further, outside this website

These links are optional references, not embedded services or dependencies. Their implementations, conventions, availability and terms are controlled by their respective authors.

Scope of linked shoaling: The new standalone calculation supports an idealized gentle slope and linear wave transformation only; it does not provide a variable-depth, structure-resolving reflection solver. Bars, complex wave breaking, fully coupled refraction–diffraction and arbitrary structures need additional governing physics, numerical verification and documented boundary conditions.

Wave Lab presents idealised educational demonstrations of wind-generated waves, shoaling, refraction, diffraction and reflection. These simplified visualisations are not site-specific forecasts or design calculations. Linked source materials retain their respective rights.

Return to the learning journey
THE READING ROOM / FREE TO EXPLORE

Follow the evidence.

Begin with a clear explanation. Go further with public guidance, scientific assessments and technical handbooks.

These resources have different purposes: education, technical guidance, assessment and statistical reference. They are not all formal international standards. Links point to the named organisations; availability can change. Attribution does not imply endorsement.

A lifelong learning project

A world of processes.
A lifetime of curiosity.

An independent educational project by Stanley Chuah, bringing coastal engineering and oceanography into an accessible space for exploration. Built to grow, one carefully explained process at a time.

31 available chapters: wave foundations, irregular seas, nonlinear theory, spectral modelling, diffraction, run-up and overtopping. The separate Wave Lab contains four linked 1D–2D–3D studios for wind-wave generation, shoaling, refraction and reflection. Wave-driven currents remain outside the current scope. Chapters 26–30 explore infragravity waves, basin resonance, wave measurements, data quality and uncertainty.

Review edition 35 · Wave modelling pathway · Not published
CHAPTER 30 / UNCERTAINTY & CONFIDENCE

How certain is
the wave estimate?

A spectrum is an estimate of a changing physical process, not an error-free picture of the sea. Separate sampling variability, instrument and processing uncertainty, and natural sea-state variability before making engineering decisions.

30.1 · What is uncertain?

Let η(t) denote sea-surface elevation (m), with one-sided variance-density spectrum Sη(f) (m² Hz⁻¹), for positive frequency f (Hz). A measured record x(t) may be modelled schematically as x(t)=g[η(t)]+b(t)+ε(t), where g represents the sensor and its response, b a systematic contribution, and ε random measurement noise. In practice, the observation model may be frequency dependent, nonlinear and affected by orientation, depth, drift and environmental conditions.

Three distinct questions. (1) How much would a spectral estimator vary across repeated realizations of the same stationary process? (2) How uncertain is the measurement after calibration, sensor response and processing? (3) How representative is this record of another hour, storm or design condition? A χ² interval addresses primarily the first question, not the other two.

Type A evaluations use statistical analysis of observations; Type B evaluations use other information such as calibration certificates, manufacturer specifications, prior characterization and physical models. These classifications describe the evaluation method, not a simple division into random and systematic errors. A known significant bias should be corrected where justified, with uncertainty retained for the correction.

30.2 · Distribution and confidence interval of a spectral estimator

For a locally stationary, approximately Gaussian process, suitably normalized independent periodogram ordinates are asymptotically exponentially distributed. Averaging independent estimates gives the useful approximation ν Ŝ(f)/S(f) ∼ χ²ν, with ν effective degrees of freedom (DOF). For K independent, equally weighted, unsmoothed periodograms, ν≈2K. This is a sampling model, not an exact finite-record guarantee.

Pointwise spectral confidence interval
Pr[νS^(f)χ1−α/2,ν2≤S(f)≤νS^(f)χα/2,ν2]≈1−α

Here Ŝ is the estimated one-sided spectral density (m² Hz⁻¹), χ²p,ν is the p-quantile of the chi-square distribution, α is the two-sided error probability, and ν is the estimator's justified effective DOF. The interval is asymmetric because χ² is skewed. The commonly quoted relative standard deviation is approximately √(2/ν), valid under this estimator model.

Interactive experiment A · Sampling confidence interval

Observe how more independent information narrows the interval. The displayed interval is pointwise and conditional on a stationary Gaussian spectral-estimation model.

30.3 · Welch averaging, leakage and effective information

Welch's method divides N samples (sample interval Δt in s) into K segments of L samples, applies a window w[n], computes modified periodograms and averages them. With hop D samples, K=1+⌊(N−L)/D⌋ when the record is fully populated. The nominal Fourier spacing is Δf=fs/L (Hz), with fs=1/Δt. Window main-lobe width and equivalent noise bandwidth (ENBW) affect practical resolving power: ENBW = fs Σw[n]²/(Σw[n])², in Hz. Zero-padding interpolates displayed frequencies; it does not create independent resolution.

Overlapping, windowed segments are generally correlated. An approximate Gaussian-process correction is νeff≈2K/[1+2 Σj=1…K−1(1−j/K)ρj²], where ρj=Σnw[n]w[n+jD]/Σnw[n]² (zero outside the window). This approximation addresses overlap covariance for a given window and is not universally valid for all spectral smoothing, processes or frequencies. More general weighted averages require a covariance-based DOF estimate; do not infer ν from K alone.

Interactive experiment B · Resolution versus stability

Illustration assumes N=4096 samples, fₛ=2 Hz, Hann window, and the stated approximate overlap-DOF model. Greater overlap does not produce proportionally more independent estimates.

30.4 · Pointwise intervals are not simultaneous bands

A 95% pointwise interval has approximately 95% coverage at one preselected frequency. Plotting 95% intervals at many frequencies does not yield a 95% probability that the entire true spectrum lies inside all intervals. For M independent tests, the probability of simultaneous coverage would be 0.95M; correlated spectral bins invalidate this simple product.

A conservative simultaneous construction applies Bonferroni: use individual two-sided significance α/M at each of M predeclared frequencies, yielding familywise coverage at least 1−α under valid marginal intervals, regardless of dependence. This may be wide. Resampling, parametric simulation or covariance-aware methods can provide less conservative bands if the data-generating model and dependence are defensible. Frequency selection after inspecting the data needs special care.

Interactive experiment C · Multiple-frequency coverage

30.5 · Spectral moments and derived wave parameters

For integration bounds fa and fb (Hz), define mn=∫fₐfᵦ fnSη(f)df, with units m² s−n. In discrete form, m̂n=ΣifinŜiΔfi. Thus Hm0=4√m0 (m), Tm01=m0/m1 (s), and Tm02=√(m0/m2) (s). The factor four defines the conventional spectral significant wave height; it is not an assertion that every sea-state distribution is Rayleigh.

Write Cij=Cov(Ŝi,Ŝj) (m⁴ Hz⁻²). Then Cov(m̂p,m̂q)=Σij(fipΔfi)(fjqΔfj)Cij. Omitting off-diagonal terms is justified only if their contribution is negligible. For g(m), the first-order delta method gives u²(g)≈∇gᵀCm∇g. For Hm0, ∂H/∂m0=2/√m0; therefore u(H)≈(2/√m0)u(m0) and relative uncertainty u(H)/H≈½u(m0)/m0.

For Tm01, the gradient with respect to (m0,m1) is (1/m1, −m0/m1²). For Tm02, the gradient with respect to (m0,m2) is (Tm02/(2m0), −Tm02/(2m2)). Their uncertainties must include moment covariance. The peak period Tp=1/fp is especially unstable for flat or multi-peaked spectra; a first-order derivative cannot capture peak switching or bin-selection effects. Use realizations, spectral peak fitting with validated assumptions, or a Monte Carlo analysis where appropriate.

Interactive experiment D · Propagating uncertainty into Hm0

First-order propagation only; a ±u range is not automatically a 68% confidence interval or an expanded measurement uncertainty.

30.6 · Traceable measurement uncertainty budget

Build a measurement model from the actual sensor, acquisition chain, response correction and spectral-processing pipeline. Potential contributions include calibration scale and offset; sensor frequency response and transfer-function inversion; tilt/orientation and pressure-to-surface conversion; clock drift, synchronization and sampling jitter; digitization, resolution and clipping; mooring motion, depth and atmospheric pressure; acoustic contamination; missing data and QC choices; window leakage, finite bandwidth and high-frequency truncation. Avoid counting the same noise source twice when a statistical repeatability estimate already includes it.

Illustrative Hm₀ standard-uncertainty budget; teaching values, not universal specifications
ContributionStandard uncertainty (m)Evaluation / dependency
Repeatability of Hm₀ estimator0.09Type A; conditional on stationary record
Calibration and sensor response0.06Type B; response characterization
Processing and bandwidth0.04Type B or A; sensitivity experiments
Environmental correction0.03Type B; site-specific model

For this illustrative independent-component budget, uc=√(0.09²+0.06²+0.04²+0.03²)=0.119 m. For Hm0=3.00 m, a nominal k=2 expanded uncertainty is U≈0.238 m. Coverage probability cannot be asserted as exactly 95% without distributional and effective-DOF justification. Correlated terms require uc²=Σui²+2Σi<jrijuiuj, with rij correlation coefficients.

Monte Carlo propagation: define joint probability distributions for uncertain inputs, including correlations; sample input vectors; pass each sample through the same measurement and spectral-processing model; summarize the resulting output distribution and its quantiles. Verify convergence and sensitivity, document distribution choices and distinguish propagated input uncertainty from stochastic variability of the sea state. A time-series bootstrap must respect serial dependence (e.g. blocks), and may fail under nonstationarity.

30.7 · Nonstationarity, decisions and reporting

A storm ramp, evolving swell system, tidal modulation or abrupt sensor change violates the constant-spectrum premise over a record. Splitting a nonstationary record into shorter windows may reduce bias from changing conditions but increases estimator variability and worsens frequency resolution. Diagnostics include time-frequency displays, segment-by-segment moments, changes in variance and spectral shape, and contextual environmental records. A stationary χ² interval must not be used as a confidence interval for a changing physical spectrum without qualification.

Report the sampling rate and duration; sensor and calibration provenance; instrument response and uncertainty corrections; detrending and QC flags; spectral window, segment length, overlap, ENBW, normalization and integration limits; estimated DOF and its method; confidence level and whether pointwise or simultaneous; spectral moment definitions; treatment of covariance and nonstationarity; and any Type A/Type B budget, coverage factor and assumptions. For design or model validation, distinguish instrument uncertainty, sampling uncertainty, spatial representativeness, model discrepancy and longer-term climate variability.

Common misconceptions. A narrow spectral confidence interval does not establish sensor accuracy. Increasing zero-padding does not improve independent frequency resolution. Overlapping segments are not independent replicates. A 95% pointwise interval is not a 95% simultaneous envelope. A k=2 expanded uncertainty is not invariably a 95% interval. A large peak-period jump can result from two competing spectral peaks, not a defective sensor.

30.8 · Methods and reference provenance

Metrology guidance: JCGM 100:2008, Evaluation of measurement data — Guide to the expression of uncertainty in measurement (GUM); JCGM 101:2008, Supplement 1 — Propagation of distributions using a Monte Carlo method. These govern uncertainty evaluation concepts, not wave-specific numerical thresholds. Spectral methodology: 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 Transactions on Audio and Electroacoustics 15(2), 70–73, DOI 10.1109/TAU.1967.1161901; Percival, D. B. & Walden, A. T. (1993), Spectral Analysis for Physical Applications, Cambridge University Press. Wave science: Holthuijsen, L. H. (2007), Waves in Oceanic and Coastal Waters, Cambridge University Press, especially the spectral-analysis appendix; Kamphuis, J. W. (2000), Introduction to Coastal Engineering and Management, World Scientific, short-term wave analysis. Data context: IOOS QARTOD Manual for Real-Time Quality Control of In-Situ Surface Wave Data, version 2.1 (2019), DOI 10.25923/7yc5-vs69; CF Conventions 1.10 for exchange metadata. These sources have distinct roles; no claim of compliance or complete clause-by-clause review is made here.

Coastal modelling hub / Wave pathway available

Start with an engineering question.
Leave with traceable evidence.

A growing reference hub for coastal engineers, metocean analysts and numerical modellers. The currently available pathway focuses on waves: use the tasks below to reach Chapter 31: Numerical wave modelling and the supporting lessons on wave physics, observations, data quality and uncertainty.

Available now: numerical wave modelling. Planned, not yet published: hydrodynamic modelling, sediment transport and morphodynamics, and coupled coastal modelling. This hub can accommodate each as a separate module without renumbering the existing wave chapters.

Model suitability is not software popularity. Phase-averaged spectral and phase-resolving wave models answer different questions. A refined mesh cannot repair an inappropriate equation family, missing physical process or unrepresentative boundary forcing. The required quantity of interest and acceptable uncertainty must lead the choice.

Wave modelling: choose your task

A six-gate checklist for an auditable modelling study

  1. Purpose & acceptance criteriaState the required output, spatial/temporal scale, decision use, tolerances and failure consequences; see 31.1.
  2. Physics & numerical designDocument governing equations, parameterisations, grid/spectral resolution, stability criteria and limitations; see 31.2, 31.7 and 31.16.
  3. Traceable external inputsRetain bathymetry and datum sources, forcing fields, directional conventions, spectral density units, boundary/nesting logic and timestamps; see 31.21 and 31.12.
  4. Verification before calibrationRun limiting/analytical cases, mesh/time-step refinement, balance checks, sensitivity screening and reproducibility tests; see 31.10 and 31.16.
  5. Independent observational evidencePreserve QC flags, calibration/validation separation, time-window matching, skill by sea-state regime and uncertainty; see 29, 30 and 31.23.
  6. Decision-ready reportingProvide configuration/version, boundary provenance, alternatives tested, limits of applicability, unresolved errors and a conclusion specific to the decision; see 31.17 and 31.25.

Wave modelling: specialised cases

Offshore-to-nearshore wave transformation

Wave-action balance, spectral nesting, directional conversions, source terms and source-free benchmark cases. The linked demonstrations are not full two-dimensional solver implementations.

Action balance → · Nesting → · Shoaling lab →

High-wind and tropical-cyclone forcing

Spatial/temporal sampling, wind-field reconstruction, storm-core resolution, squalls and Holland-family applicability. Cyclone examples reconstruct idealized forcing, not validated operational storm forecasts.

Spatial wind → · Holland framework → · Wind labs →

Wave-resolving processes and structures

Assess what is lost by phase-averaging when the question involves individual crests, near-field reflection, diffraction, basin resonance or runup.

Dispersive models → · Diffraction → · Wave Lab →

Observations, spectral statistics and uncertainty

Connect instrument response and flagging to spectral confidence, measurement uncertainty and fit-for-purpose comparisons with model output.

Measurements → · QC → · Uncertainty →

Published guidance and what it supports

IOOS / QARTOD wave-data QC — documented real-time checks and observational quality practice, not universal local thresholds.

JCGM measurement-uncertainty guides — uncertainty evaluation and propagation; application depends on the measurement model.

ASME V&V 20 — a formal verification/validation framework for specified quantities and validation points; not an automatic accreditation of this website or of a coastal study.

SWAN technical documentation — one recognised implementation of phase-averaged spectral modelling. Its code, settings and calibration are not part of these educational widgets.

Evidence boundary: Every linked figure or laboratory is an idealised or synthetic instructional example unless explicitly identified otherwise. The hub provides an engineering workflow and routes to existing lessons; it does not run a verified, calibrated or validated site-specific numerical model. No wave-driven currents are added by this release.

Read Chapter 31: Numerical wave modelling Browse all 31 learning chapters ↑For project work, apply the relevant project specification, model-version manual, independent observations and peer review.
CHAPTER 31 / NUMERICAL WAVE MODELLING · ADVANCED ENGINEERING

Numerical wave modelling
From offshore spectra to nearshore wave-by-wave physics

The right model is the simplest defensible mathematical representation of the processes that control the required engineering output. Model inputs, temporal sea-state definitions, evidence of skill against measurements and downstream design consequences are equally fundamental. This chapter explains the governing physics, model-selection trade-offs, numerical errors, forcing and boundary conditions, validation evidence and practical limits—without treating a software product as an answer.

31.1 · The modelling hierarchy is a hierarchy of questions

Start with the quantity of interest (QoI): offshore wave climate; nearshore Hm0, directional spectra or energy flux; harbour agitation and long-period resonance; wave-by-wave runup; coastal flooding; local force, impact pressure or overtopping jet. Identify the minimum physical scales and mechanisms needed to determine that quantity. Grid size, software, parameterisations and runtime follow from the physics.

Three levels of numerical wave physicsAn original conceptual diagram showing how offshore frequency-direction spectra provide coastal boundaries, how wave-phase-resolving domains operate at smaller scales, and when local three-dimensional flow modelling is needed. Arrows are possible data transfers, not a mandatory modelling chain. Spectral / phase-averagedRegional → shelf → coastalE(f, θ), Hm0, T, θ, flux Spectralboundary Phase-resolving, 2DH/3DDispersive or non-hydrostaticη(x,y,t), velocities, runup Localforcing 3D free-surface flowNavier–Stokes / VOFPressures, jets, local turbulence Domain extent usually decreases and required physical detail usually increases →Not every project needs all three stages. Each transfer adds assumptions and an uncertainty pathway.
Figure 31.1. Original conceptual model hierarchy. Spatial dimensions (1D/2DH/3D), spectral dimensions (frequency and direction), and phase treatment are independent descriptors. The arrow does not prescribe a commercial workflow.

Terminology: “2D spectral” may refer to a two-horizontal-dimensional computation, or to a frequency–direction spectrum E(f,θ). A 2DH spectral model may carry two spectral coordinates at every map cell. A 2DH Boussinesq calculation resolves η(x,y,t) at individual wave time scales. These meanings must never be confused.

31.2 · Distinguish equation families without marketing

Mathematical familyState variables and typical scalesWhy an engineer may use itWhat must not be assumed
Phase-averaged spectral action/energy balanceE(f,θ;x,y,t), N; regional to coastal domainsWave climate, wind sea/swell evolution, depth refraction, directional statistics and integrated wave design conditionsNo coherent crest-by-crest interference, crest shape, instantaneous impact pressure or direct individual-wave runup
Linear mild-slope / phase-resolving potential-flow approximationsComplex amplitude/phase, possibly at individual frequencies; harbour/coastal domainLinear diffraction, refraction, reflection and harbour agitation when weak nonlinearity is adequateNot automatically a nonlinear breaking or inundation model
Boussinesq-type, depth-integrated dispersiveη and horizontal transport/velocity; coastal to nearshoreIndividual-wave shoaling, phase-coherent diffraction/reflection, nonlinear interactions, suitable long waves and runupFinite validity in kh and nonlinearity; breaking and wetting–drying are closure/scheme dependent
Non-hydrostatic free-surface, one or multiple vertical layersη, velocity layers, non-hydrostatic pressure; nearshoreDispersive transformation, bores and surf-zone dynamics with tunable vertical resolutionOne vertical layer and multiple layers do not have identical dispersion, pressure or vertical flow fidelity
Wave-group / surfbeat formulationsShort-wave envelope/action, long-wave water level and transportInfragravity response, storm erosion and long-period shoreline motions without every short-wave crestShort-wave crest phases are not explicitly resolved in the group-resolving mode
3D Navier–Stokes, interface capturing (VOF or equivalent)3D velocity, pressure and air–water interfaceLocal breaking jets, overtopping, impact and complex structure hydraulics when verifiedNot feasible as a default for decades of large-area wave climates; detailed turbulence and air physics remain modelled

Examples of implementations only to aid literature lookup: WAM, WAVEWATCH III, SWAN and MIKE 21 SW are used for spectral approaches; FUNWAVE-TVD and MIKE 21 BW for Boussinesq-type approaches; SWASH for a non-hydrostatic approach. XBeach has stationary, group-resolving and non-hydrostatic modes; its name alone does not identify the time scales resolved. Releases and available modules matter. No ranking or equivalence is intended. See the primary scientific and versioned documentation sources.

31.3 · Phase-averaged spectral models: action conservation, not individual waves

For short-crested random waves, define the variance density Eσ(σ,θ;x,y,t) in m² per (rad s⁻¹) per radian of direction. Let intrinsic angular frequency σ (rad s⁻¹) be frequency relative to a mean current, θ wave-propagation azimuth (radian), and N=Eσ/σ the action density under this variance convention. In horizontally Cartesian coordinates a standard conservative form is:

Wave-action conservation
∂N∂t+∂(cxN)∂x+∂(cyN)∂y+∂(cσN)∂σ+∂(cθN)∂θ=Stotσ

Here (x,y) are horizontal coordinates (m), t time (s), cx, cy propagation velocities (m s⁻¹), cσ frequency-space propagation rate (rad s⁻²), cθ angular turning rate (rad s⁻¹), and Stot the source/sink rate in the same density convention as Eσ. Angles expressed in degrees require consistent density and Jacobian conversion. Physical wave-energy density per sea-surface area is ρgEσ; the corresponding physical action density includes the same factor. The phase-averaged balance predicts statistical wave properties—not crest arrival times.

Net spectral source
Stot=Sin+Snl4+Snl3+Swc+Sbf+Sbr+⋯

Source terms represent wind input, quadruplet transfer, triad transfer, whitecapping, bottom-friction dissipation and depth-induced breaking (with convention-dependent signs). A term being available is not proof that it is applicable. For example, nearshore triad effects can redistribute energy among harmonics whereas a breaking closure controls bulk dissipation; tuning breaking to compensate for incorrect triad physics is not a defensible calibration.

Why action rather than wave energy? Derivation and units

For slowly varying, weakly nonlinear wave packets, σ²=gk tanh(kh) with wave number k (m⁻¹), depth h (m), and g (m s⁻²). In stationary, current-free media, σ does not change along a ray; energy balance follows. With time-varying media or mean current, intrinsic frequency and absolute frequency differ by the Doppler term ω = σ + k·U, and action E/σ is the suitable adiabatic conserved quantity in the source-free approximation. This formulation describes wave propagation with a prescribed mean current; it does not calculate the generation of wave-driven mean currents. The approximation requires a separation between rapidly oscillating wave phase and slowly varying medium properties.

Conversion between Ef (per Hz per radian direction) and Eσ (per rad s⁻¹ per radian direction): Eσ=Ef/(2π) because dσ=2π df. Spectral moments based on Hz are mn=∫∫ fnEf df dθ, with units m² Hzn. Do not silently mix frequency conventions.

Parameters to check: frequency limits and tail treatment (missed energy changes m0 and periods); direction sector width (spreading/refraction); bathymetric resolution (false blocking/focusing); wind and surface stress conventions; breaking and bottom-friction closures; shallow-water nonlinear interactions; boundary consistency; solver iteration/convergence and stationarity assumptions. Assess full E(f,θ) skill—not only Hm0.

31.4 · Wave-resolving physics: dispersion, nonlinearity and breaking

Depth-integrated Boussinesq approaches start from mass and horizontal momentum conservation, retaining higher-order terms associated with finite wavelength relative to depth. A useful two-horizontal-dimensional (2DH) structural representation is given below, with h(x,y) the undisturbed water depth, η the free-surface displacement, u = (u,v) the horizontal depth-averaged velocity (m s⁻¹), and ∇ = (∂/∂x, ∂/∂y):

Depth-integrated continuity
∂η∂t+∇⋅[(h+η)u]=0
Dispersive momentum schematic
∂u∂t+(u⋅∇)u+g∇η+D[u,η,h]=F

D represents the formulation-specific dispersive differential operators (units m s⁻²), while F denotes separately specified friction, breaking or other momentum-source terms (m s⁻²). This is an equation-family schematic, not one universally valid Boussinesq system: the derivatives, bathymetric-gradient terms and velocity level differ between formulations. Setting D and F to zero recovers the depth-averaged nonlinear shallow-water skeleton, not the full dispersive wave physics. For constant depth h and depth-mean horizontal velocity u, the following one-dimensional, illustrative classical form makes dispersion explicit:

One-dimensional continuity
∂η∂t+∂[(h+η)u]∂x=0
One-dimensional momentum
∂u∂t+u∂u∂x+g∂η∂x−h23∂3u∂x2∂t=0

η (m) is surface elevation from the reference still-water level, u (m s⁻¹) depth-mean velocity and x distance (m). The second expression is a selected weakly dispersive schematic, not the general variable-depth equations implemented in modern solvers. The dispersive term is absent from the nondispersive nonlinear shallow-water equations. Modern extended Boussinesq formulations modify the operators to improve dispersion and nonlinear fidelity; some use reference-level rather than depth-averaged velocity.

Derive the classical linear dispersion relation

Linearise about η = 0 and u = 0 at constant depth. Substitute η = η̂ exp[i(kx−ωt)] and u = û exp[i(kx−ωt)] into ηt + hux=0 and ut+gηx−h²uxxt/3=0. Continuity gives û = ωη̂/(kh). Momentum gives ω(1+k²h²/3)û = gkη̂. Eliminating û yields:

Classical Boussinesq dispersion
ω2=ghk21+(kh)2/3
Normalized phase speed
cBgh=[1+(kh)23]−1/2

Compare exact linear potential-flow ω²=gk tanh(kh), which gives cexact/√(gh)=√[tanh(kh)/(kh)]. Both recover √(gh) as kh→0. At larger kh the classical Boussinesq approximation departs from exact dispersion; an extended model may differ substantially. This graph compares formulations, not software implementations or the additional numerical dispersion of a finite mesh.

Validity questions: (i) Is kh sufficiently small for the chosen formulation's documented error tolerance? (ii) Are nonlinearities such as H/h and wave steepness H/L in the model's tested envelope? (iii) Is breaking modelled through bore capture, eddy viscosity or a switch between equations? (iv) Are moving shoreline and reflected waves treated correctly? (v) Is wave generation compatible with second-order bound harmonics and infragravity energy? (vi) Does a depth-integrated variable suffice, or are vertical structure and local pressures needed?

Dimensionless diagnosticDefinitionReason to inspectLimitation
Relative depthkh = 2πh/LDispersion accuracy and model regimeNot a universal single cutoff
SteepnessH/LNonlinearity, onset of steep/breaking behaviourDepends on wave definition and sea state
Relative wave heightH/hDepth limitation, free-surface excursionsBreaking criteria and vertical datum matter
Ursell numberUr = H L²/h³ (one convention)Long-wave nonlinearity relative to dispersionConventions vary by coefficient
Grid samplingLmin/Δx and Lmin/ΔyNumerical phase and amplitude fidelityMethod, stencil, mesh orientation dependent
Courant numberC = crefΔt/ΔxExplicit time-integration stability and accuracyUse solver-specific multidimensional criterion; not a universal threshold

For non-hydrostatic approaches, the non-hydrostatic pressure correction and number/placement of vertical layers affect dispersion and vertical velocity representation. For surfbeat approaches, verify which short-wave and long-wave time scales are actually advanced. For local 3D CFD, check mesh interface resolution, turbulence/air entrainment, boundary reflections and impact-pressure convergence. Do not automatically escalate to 3D if a simpler verified model already answers the QoI.

31.5 · Physical-process capability matrix

Read “resolved” as represented by the model equations at an appropriate validated resolution, not automatically accurate. The entries deliberately avoid a simplistic yes/no comparison.

Process or requested deliverablePhase-averaged spectralPhase-resolving depth-integrated / non-hydrostaticLocal 3D CFD / interface resolving
Offshore wind-wave generation and decades of wave climateAppropriate with validated source terms and forcingImpractical for large-domain climate; external waves typically suppliedImpractical for climate scales
Depth refraction, shoaling and bulk Hm0Primary use; bathymetry and closure dependentWave-by-wave if within dispersion and mesh envelopePossible but often excessive
Coherent diffraction / interference around gapsNot phase-resolved; optional approximate treatments in some implementationsExplicit, subject to grid, reflection and dissipationExplicit locally
Harbour standing waves / resonanceDirectional energy/reflection approximations only; no deterministic phaseExplicit basin modes and long-period motions if excited and adequately simulatedLocally possible, costly
Breaking and average surf-zone Hm0Statistical breaking closureBreaking/bore schemes; wave shapes approximatedOverturning possible with interface and turbulence treatment
Infragravity response and wave groupsUsually no explicit phase-resolved long-wave forcingCan resolve if forcing, nonlinear physics, boundary and simulation duration suffice; surfbeat is an intermediate optionCan resolve but seldom economical at field scale
Individual-wave runup and swashNot directly from bulk spectraPotentially resolved with moving shoreline and breaking schemePossible at smaller scales
Impact pressure, air entrainment and local turbulenceNot resolvedDepth-averaged pressures insufficient for detailed impulsive loadsCandidate; verify multiphase and subgrid assumptions

31.6 · Interactive model-selection reasoning

The decision tree below screens equation families, not products. A model with more equations is not necessarily the right model; distinguish observables from mechanism requirements. Choices are prompts for engineering review rather than universal decision rules.

Lab A · Choose the required prediction, then declare indispensable physics

Processes demonstrably significant for the project (optional refinements)
Select requirements to see an engineering screening argument.

Scope limitation: no software recommendation is produced. Validate domain, kh, wave nonlinearity, boundaries, computational cost and field evidence before selection.

31.7 · Critical setup parameters: what changes and why

Prioritise parameters by the sensitivity of your target output, not by how many parameters a program exposes. A parameter can be physically influential yet unidentifiable from available data. Calibration is not a licence to compensate for an invalid equation regime or incorrect bathymetry.

Parameter / decisionPhysical / numerical mechanismTypical failure if neglectedEvidence required
Domain extent and placement of open boundariesAvailable fetch, shoaling/focusing before target, reflected-wave travel timeArtificial boundary influence or missing wave-growth distanceBoundary-distance sensitivity; reflection diagnostics
Vertical datum and bathymetric interpolationControls depth, dispersion, breaking, exposed/shallow cellsFalse shoals/channels, altered kh and energy focusingSurvey metadata, grid-versus-survey plots, datum audit
Frequency minimum/maximum and tailCaptures wind sea, swell, relevant long waves, spectral momentsHm0 and periods biased by missing bandsm0, m−1, m1, m2 stability on refined spectral domain
Directional bins / convention / headingAngular concentration, refraction and boundary incidence180° reversal, 90° rotation, false directional spreadingKnown-bearing synthetic test and circular angular checks
Wind input, whitecapping, quadruplet/triad transferSpectral growth and redistributionPeak and high-frequency tail errors masked by Hm0 matchFull spectra across wind regimes and source-term sensitivity
Bottom friction / depth-induced breakingDissipation along shallow pathwaysExcess or missing wave variance in nearshoreObserved nearshore profiles, bottom type, closure sensitivity
Spatial Δx, Δy / mesh distortion and grid orientationResolves bathymetry and phase gradientsSpurious refraction or phase lag; false reflectionGrid-convergence study; model-specific dispersion analysis
Time step, CFL and iteration toleranceExplicit stability; implicit convergence and temporal accuracyFalse stability, numerical damping, delayed arrivalsΔt/solver tolerance convergence, conserved-quantity checks
Wave generation and lateral/sponge boundariesIncident phase spectrum and absorbed outgoing wavesSpurious standing waves, nonphysical energy buildupIncident-only, reflection and propagation tests
Wave breaking / wetting–drying thresholdsSurf-zone bores and shoreline motionUnphysical runup or discontinuous surface behaviourMeasured runup distribution and breaking-location checks
Phase-resolving record duration and spin-upSample stability, low-frequency resolution and basin adjustmentMissing resonance, unstable Hm0 and biased long wavesTime-window independence, stationarity diagnostics

Spatial and time resolution

Determine the shortest wavelength Lmin that materially affects the output, then calculate Lmin/Δx and Lmin/Δy across the domain. A teaching starting point such as 10–20 cells per shortest relevant wavelength is not a portable acceptance criterion. High-order schemes, large kh, nonuniform meshes, nonlinear wave shapes, reflection and wetting–drying can demand different resolutions. Test convergence of the actual QoI.

Spectral and boundary resolution

Frequency bins determine whether narrow peaks, low-frequency wave energy and spectral moments are represented. Directional bins control turning and shadowing. Use frequency–direction spectra at boundaries where important; Hm0, a single period and one direction cannot uniquely reconstruct swell-plus-wind-sea partitions.

Physical parameter sensitivity

Vary one closure at a time only for diagnosis; interactions require multi-parameter designs. Record physical ranges, site conditions and whether calibration remains identifiable. Avoid transferring a bottom-friction coefficient, breaker index or sponge coefficient between distinct equations without justification.

Uncertainty and data quality

Start from Chapter 29 QC flags and Chapter 30 measurement/sampling uncertainties. A confidence interval on observed Hm0 does not quantify model structural error. Maintain separate contributions from forcing, bathymetry, numerical truncation, parameterisation and observation uncertainty.

31.8 · Why identical spectra do not determine individual crests

For discrete variance density bins EjΔfj (m²), a physically consistent random-phase synthesis uses aj=√(2EjΔfj) and η(t)=Σ ajcos(2πfjt+φj). With distinct frequencies and a sufficiently long orthogonal record, variance is Σaj²/2=m0 irrespective of the assigned phase. Short finite records can have cross terms. Phase-averaged E(f) does not retain φj; it therefore cannot predict a particular crest sequence.

Lab B · Same spectral variance, different sea-surface realisation

Panel 1 · η(t) in m; both component amplitudes fixed.

Panel 2 · Component variance in m², not density per Hz.

Original two-frequency deterministic illustration. The finite plotted window is not used to estimate the theoretical variance. Directional fields require additional wave-vector components.

31.9 · Physical dispersion and the limits of Boussinesq approximations

Let z=kh. Compare exact linear dispersion, shallow-water (nondispersive) and the classical constant-depth Boussinesq approximation. The exact phase and group speeds are c=ω/k and cg=½c[1+2z/sinh(2z)]. In the shallow-water limit z→0 both approach √(gh). Group speed is the transport velocity in the linear wave-energy flux calculation when currents are absent.

Lab C · Compare phase-speed approximations at the selected T and depth

Exact linearClassical BoussinesqShallow waterSelected kh

Figure 31.2. Governing-equation (physical) dispersion error only; discrete spatial/temporal solver error is addressed separately. All curves are normalised by √(gh); plotted kh range 0–3.

31.10 · Numerical dispersion is different from physical dispersion

Even if the governing equation is suitable, discretisation can alter phase speed. For the linear nondispersive wave equation ηtt=c²ηxx, a centred, second-order spatial difference with continuous time gives a particular numerical dispersion relation:

Semi-discrete wave frequency
ωnum=2cΔxsin(kΔx2)
Numerical phase-speed ratio
cnumc=sin⁡(kΔx/2)kΔx/2

This derivation assumes a uniform grid, sinusoidal Fourier mode and centred three-point second derivative; it is not the dispersion curve of a named spectral, Boussinesq or CFD solver. Error accumulates as a phase shift over many wave periods even when a plot looks smooth.

Lab D · Phase lag in one explicit spatial discretisation

Analytical waveCentred-difference semi-discrete wave

Figure 31.3. Idealised Fourier-mode phase drift on the same grid, not a numerical simulation of a real coastal model. 2DH cell count scales approximately as (1/Δx²) for a fixed area; stable explicit time steps often add computational cost.

31.11 · Offshore-to-coastal nesting: preserving physics across representations

One-way spectral nesting transmits a discretised E(f,θ) (and compatible water level, forcing/metadata when needed), interpolated consistently in frequency, direction, location and time. A spectral-to-phase-resolving boundary requires both energy-preserving component amplitudes and a choice of phases; the phase information cannot be recovered from a phase-averaged spectrum. Ensemble simulations with multiple seeds may be required for stable statistics.

  1. Assess overlap and fetch: outer boundary should precede major transformations required in the inner domain; provide adequate space for spin-up and wave generation.
  2. Check coordinate systems: geographic vs Cartesian, x east/y north or local grid, azimuth clockwise from north versus counterclockwise from east, and incoming “from” versus propagating “to.” Convert intentionally, not heuristically.
  3. Conserve discrete variance: m0 = Σ Eij Δfi Δθj (radians) for directional variance density in m² Hz⁻¹ rad⁻¹. Check spectrum tail and energy clipping after regridding.
  4. Set physical phases: for a travelling component ηij=aijcos(kxx+kyy−2πfit+φij), choose aij=√(2EijΔf Δθ), and k components from the propagating direction and dispersion relation. This is a linear boundary synthesis; second-order bound waves may need explicit treatment.
  5. Protect the boundaries: use wavemakers, radiation/open conditions or absorbing sponge zones appropriate to the target formulation; compare incident and reflected components quantitatively.
  6. Verify at a control transect: compare frequency–direction spectra, m0, Hm0, periods and directional moments in the overlap. Note that two phase realisations need not reproduce the same short-record statistics.

For a propagation-to heading θ measured clockwise from north with x east and y north, kx=k sinθ and ky=k cosθ. If input gives meteorological from bearing, θto = (θfrom+180°) mod 360°. Beware software implementations that define θ counterclockwise from the positive x axis. A 180° error can reverse an incoming offshore wave field.

31.12 · Directional-spectrum energy and boundary-convention laboratory

The synthetic distribution below is numerically normalised to a prescribed Hm0, then integrated over frequency and direction. The angular spread is illustrative, not a wind-input parameterisation. Changing the number of directional sectors updates the discrete representation while conserving the prescribed m0 by explicit normalisation.

Lab E · Check variance, density units and “from”/“to” azimuth

Panel 1 · E(f) after integrating E(f,θ) over direction, in m² Hz⁻¹.

Panel 2 · Directional distribution with geographic N/E/S/W conventions.

Synthetic and energy-normalised, not a field spectrum. The variance spectrum alone cannot specify the associated wave phases or complete incident bound-wave conditions.

31.13 · Energy-flux conservation and the pitfalls of bulk shoaling

In a stationary, current-free, one-dimensional, normally incident, nonbreaking and nondissipative linear wave field, wave-energy flux per unit crest width is constant. For a narrowband wave component the relation is:

Energy flux conservation
F=ρgm0cg=constant
Wave-height transformation
Hm0,2Hm0,1=cg,1cg,2

ρ (kg m⁻³) is water density, g (m s⁻²) gravity, m0 (m²) zeroth variance moment, and cg (m s⁻¹) linear group speed. For broadband spectra, apply group speed by frequency before integration; a single representative period is only a narrowband teaching approximation. Breaking, refraction, directional spreading, currents, friction or geometric convergence invalidate this simple one-dimensional relation.

Lab F · Source-free narrowband transformation

Figure 31.4. The trajectory is an analytical diagnostic of energy-flux conservation. It deliberately excludes breaking; large predicted H/h requires checking this assumption, not accepting the value as a safe nearshore design condition.

31.14 · Verification, calibration and independent validation

These procedures are distinct. Code verification checks whether the intended equations are correctly implemented, using analytic/manufactured solutions or published benchmarks. Solution verification estimates error from grid, time, spectral and iterative discretisation. Calibration estimates uncertain inputs against designated observations. Validation assesses predictive capability against separate evidence, for variables and conditions relevant to use. Uncertainty quantification describes uncertainty in outputs from data, parameters, numerics and model-form limitations. See ASME V&V 20 for an established domain-specific terminology and philosophy; it is not presented as a coastal-model certification.

Mean error
Bias=1n∑i=1n(Pi−Oi)
Root-mean-square error
RMSE=1n∑i=1n(Pi−Oi)2

Pi and Oi are paired predictions and observations (same units), after matching time, location and QC criteria. Bias and RMSE alone do not establish skill for extremes, spectra, direction, harbour modes or overtopping; use residuals, quantile behaviour, event and direction regimes, uncertainty bounds and transparent sample counts.

Lab G · Over-tuning to a subset can harm independent predictions

The observations and pre-calibration predictions below are deliberately synthetic. The scalar parameter multiplies the predictions; lower-severity cases are a training subset and higher-severity cases are independent holdout observations. Their different biases expose why a single calibrated parameter can conceal regime-dependent errors.

Figure 31.5. Synthetic data only. This is a method demonstration, not validation of any wave-model software or a claim that any particular observed site exhibits these residual patterns.

Convergence design: run at least three suitably refined spatial/time/spectral resolutions when feasible; check that the observed convergence regime supports extrapolation, and record numerical errors in QoIs rather than only a stable solver residual. Change physics parameters separately from grid resolution. In a phase-resolving case inspect phase lag and phase coherence; in a spectral case inspect directional spectra and source-integral balance; for inundation inspect wet/dry front and mass conservation.

31.15 · Engineering cases: choosing against a model

A · Offshore wave hindcast

Output: multiyear E(f,θ), Hm0, Tm−1,0, Tp, extreme statistics. Physics: wind input, whitecapping, nonlinear transfers, propagation. Reasoning: select spectral action modelling; a wave-by-wave domain would not practically resolve the climate horizon. Important checks: forcing winds, source terms, hindcast spin-up, directional spectra and separate seasonal/storm performance. Reject: phase-resolving local CFD as a regional-hindcast substitute.

B · Shallow reef and coastal structure

Output: nearshore Hm0, spectra and wave incidence. Physics: refraction, shoaling, bottom dissipation, depth breaking, possible triad transfers. Reasoning: start with a sufficiently resolved coastal spectral approach if coherent wave phase is not essential. Escalation: if reef gaps, reflections or nonlinear individual-wave shapes control the design variable, screen a phase-resolving formulation. Checks: bathymetry, spectral-bin sensitivity, instrument response and propagation transects.

C · Harbour agitation and long waves

Output: agitation amplitude by basin location, berth and period. Physics: reflection, coherent diffraction and resonance. Reasoning: compare linear phase-resolving mild-slope/potential-flow and dispersive models; choose nonlinear phase-resolving only where required. Reject: interpreting smooth phase-averaged Hm0 alone as a harbour resonance prediction. Checks: low-frequency boundary energy, free-mode periods, reflection and simulation duration.

D · Breaking, runup and overtopping

Output: individual runup extremes, overwash or impact. Physics: breaking, bore motion, wetting–drying, possibly vertical jets. Reasoning: screen dispersive/non-hydrostatic models for wave-by-wave shoreline dynamics; if detailed local impact pressure/air entrainment determines the QoI, investigate local 3D free-surface modelling with its own validation evidence. Reject: extrapolating statistical spectral Hm0 directly to individual overtopping volumes.

Additional forcing-specific cases F–G: For a compact tropical cyclone, verify track, radius of maximum winds, intensity, averaging convention and storm-wave evolution; a spatially uniform 10 m wind or a coarsely sampled wind core may be inadequate even in an otherwise capable spectral solver. For Sumatra/Malacca Strait squalls, distinguish short-duration gust loading from sustained wind-sea generation and check event position/time support. A higher-resolution WRF-type hindcast or reconstructed wind field should be treated as an alternative requiring both atmospheric and wave validation, not a guaranteed upgrade. See Sections 31.26–31.30 for original wind-field experiments and the engineering evidence chain.

E · Storm-induced erosion: a group-resolving surfbeat approach may be better aligned with evolving morphology and infragravity runup than resolving each incident short-wave phase. Do not assert that a group model predicts individual short-wave crests. A morphodynamic prediction also requires sediment and bed-update physics, which lie outside this chapter.

31.16 · Reproducible verification benchmarks before calibration

BenchmarkAnalytical or independent targetDiagnosesFailure signal
Uniform depth, small-amplitude regular waveω²=gk tanh(kh); c and cg from linear theoryPhysical/numerical dispersion, phase propagationWavelength, phase arrival or amplitude drift without physics
Plane-wave propagation on flat bed (no sources)Invariant E(f,θ), or amplitude/variance for a lossless phase modelPropagation numerics and artificial diffusionEnergy loss/gain increasing with distance/grid orientation
1D no-break shoalingE cg conserved in an idealised narrowband caseDepth/dispersion and flux handlingFlux imbalance in a source-free diagnostic
Reflection against a perfectly rigid wallη = 2a cos(kx) cos(ωt) with wall at x=0, in linear idealisationCoherent reflection and boundary treatmentNonphysical transmitted energy or wrong standing-wave phase
Directional convention and wrap testIncoming 350° → propagating 170°; 359°→1° circular separation 2°Azimuth handling and polar interpolation180° incidence reversal or 358° false angular jump
Phase-resolving spectral boundaryΣa²/2 = ΣEΔfΔθ = m0; Hm0=4√m0Fourier amplitude, density units, phase reconstructionMissing/doubled spectral energy or spurious directivity
Convergence test (case-specific)QoI stabilises under spatial/temporal/spectral refinementNumerical error vs calibration effectsPersistent or nonmonotonic QoI change beyond engineering tolerance

Benchmark agreement alone does not establish field validation. For acceptance, define quantitative error tolerances before comparing predictions, taking the engineering decision and observation uncertainty into account. Commercial and open-source implementations should face the same evidence standard.

31.17 · Professional reporting and independent peer review

Document the engineering question, decision threshold, geographic/vertical reference systems and wave convention. Archive the exact code version/configuration, governing formulation, modelled vs parameterised terms, all forcings, quality-controlled observations, bathymetry, mesh, spectral grid, Δt and numerical tolerances, nested spectra and phase seeds, boundary absorption, calibration/validation split, sensitivity and convergence evidence, residual distributions, uncertainty budget and limitations. For a phase-resolving run also report record length, discarded spin-up and achieved statistical stationarity.

Common reviewer objections
  • “The measured and modelled Hm0 match, so the spectrum is validated.” Incorrect: spectral shape, direction and period errors can cancel in m0.
  • “The model runs stably, so it is accurate.” Incorrect: numerical stability does not establish dispersion fidelity or grid convergence.
  • “The Boussinesq model resolves everything nearshore.” Incorrect: depth, nonlinearity, breaking, impact and vertical physics limits still apply.
  • “Hm0 and one period determine the incident wave field.” Incorrect: numerous distinct directional spectral shapes share those bulk parameters.
  • “A phase-averaged model never considers diffraction.” Too strong: some use approximations; they do not resolve wave-coherent interference.
  • “A more complex model removes uncertainty.” Incorrect: complexity adds inputs and calibration/identifiability concerns; uncertainty must be assessed.

31.18 · Sources, evidence level and further reading

Original explanations, tables and diagrams. Peer-reviewed methods support the wave equations and model-family capabilities; official solver documentation supports implementation-specific distinctions; engineering/metrological guidance informs verification and uncertainty reporting. The cited publications are not evidence that a particular site-specific setup has been validated.

  1. Holthuijsen, L. H. (2007), Waves in Oceanic and Coastal Waters, Cambridge University Press, Chapters 3 and 5–9, Appendices A–C. Supplied project textbook (physical and spectral foundations).
  2. Reeve, D., Chadwick, A. & Fleming, C. (2018), Coastal Engineering: Processes, Theory and Design Practice, 3rd ed., CRC Press, Chapters 2–3, 7–8. Supplied project textbook (engineering applications and design uncertainty).
  3. Booij, N., Ris, R. C. & Holthuijsen, L. H. (1999), “A third-generation wave model for coastal regions: 1. Model description and validation,” JGR: Oceans 104(C4), 7649–7666. DOI: 10.1029/98JC02622. Peer-reviewed model-family methods.
  4. Ris, R. C., Holthuijsen, L. H. & Booij, N. (1999), “A third-generation wave model for coastal regions: 2. Verification,” JGR: Oceans 104(C4), 7667–7681. DOI: 10.1029/1998JC900123. Published verification case evidence.
  5. Madsen, P. A., Murray, R. & Sørensen, O. R. (1991), “A new form of the Boussinesq equations with improved linear dispersion characteristics,” Coastal Engineering 15, 371–388. DOI: 10.1016/0378-3839(91)90017-B. Extended dispersion context; not the classical formula plotted above.
  6. Shi, F., Kirby, J. T., Harris, J. C., Geiman, J. D. & Grilli, S. T. (2012), “A high-order adaptive time-stepping TVD solver for Boussinesq modeling of breaking waves and coastal inundation,” Ocean Modelling 43–44, 36–51. DOI: 10.1016/j.ocemod.2011.12.004.
  7. Zijlema, M., Stelling, G. & Smit, P. (2011), “SWASH: An operational public domain code for simulating wave fields and rapidly varied flows in coastal waters,” Coastal Engineering 58, 992–1012. DOI: 10.1016/j.coastaleng.2011.05.015.
  8. SWAN project: official manuals and model documentation; NOAA Wavewatch model description and documentation; FUNWAVE-TVD documentation, equations and boundary treatments; XBeach manual, mode distinctions. Official/software documentation; verify exact version used.
  9. USACE, Coastal Engineering Manual, EM 1110-2-1100, Part II (wave climate, analysis and transformations). Multi-part engineering guidance with revisions; check section/version applicable to the method rather than inferring a universal model preference.
  10. ASME V&V 20–2009 (R2021), Standard for Verification and Validation in Computational Fluid Dynamics and Heat Transfer. Formal standard concerning V&V accuracy quantification in its stated scope, not a certification of coastal wave models.
  11. JCGM 100:2008 with published 2026 amendment, and JCGM 101:2008, measurement uncertainty and propagation of input distributions. Metrology guidance relevant to observation-side uncertainty, not a replacement for structural model-error assessment.

31.19 · Environmental inputs: provenance matters as much as resolution

Wave-model selection is incomplete until the environmental inputs, their uncertainties and their reference conventions have been specified. The computational domain receives data whose space-time support may be coarser than the processes of engineering interest. A phase-averaged offshore model can generate wind sea from atmospheric forcing; a nested nearshore spectral model often relies on incident spectra, local winds and water levels; a phase-resolving Boussinesq or non-hydrostatic model typically needs incident wave components or a consistent spectral-to-time-domain boundary, plus accurate bathymetry and water levels. These are typical configurations, not universal rules for every model.

New advanced forcing audit: Sections 31.26–31.30 extend this input table with independently evaluated 2D atmospheric fields, tropical-cyclone eyewall resolution, Malacca Strait / Sumatra squalls, atmospheric downscaling choices and wave-response timescale limitations. These are not optional accuracy details when intense transients control an engineering quantity of interest.

Input and reference conventionOffshore spectral applicationCoastal spectral / wave-resolving applicationRequired assessment before use
Wind vector and heightTime-dependent speed and direction over the generating domain, typically supplied as equivalent 10 m winds in the form expected by the chosen source-term package.Local wind is relevant when fetch and local generation materially contribute. Incident spectra/wave phases may be the dominant boundary forcing in a small wave-resolving domain.Prove what U10 means: 10 m above which datum, neutral-equivalent or actual wind, averaging period, time stamp, gust versus mean, coastal gradients and cyclone representation. Avoid automatic conversion between gust and mean.
Temporal forcing intervalExamples: hourly or three-hourly wind grids; accuracy depends on storm evolution, wind interpolation and wave response, not on the label alone.Offshore wave-boundary update interval controls incident-sea evolution; phase-resolving time stepping is much shorter and is a separate requirement.Test whether winds and incoming spectra resolve storm intensification, peak and decay. Compare competing time spacings against higher-resolution wind data and wave observations; report any peak attenuation or timing bias.
Offshore spectrum E(f, θ)May be a nested open boundary or a diagnostic output, with intrinsic/absolute frequency carefully defined where currents matter.Boundary forcing should preserve variance, frequency range and directional distribution; a wave-by-wave realization also needs phases and a generation/absorption method.Check frequency and angular-bin widths, m² Hz⁻¹ rad⁻¹ versus per-degree units, nautical “from” versus propagation “toward”, frequency limits, Hm0 and moments before and after interpolation.
Seafloor elevation and horizontal coordinatesControls shoaling/refraction and bottom dissipation; fine coastal topography may be unresolved.Shallow bars, reefs, channels, structures, bed steps and dry/wet boundaries can dominate the result.Check bathymetric survey vintage and survey uncertainty, horizontal projection, vertical datum, interpolation, channel widths and sensitivity to smoothing; IHO S-44 governs hydrographic surveying, not wave-model acceptance.
Still-water level, tides, surge and currentsDepth and current-induced frequency/direction shifts may be important.Depth-controlled breaking, overtopping, reef exposure, wave–current interaction and dry/wet motion can be highly sensitive.State total-water-level composition, datum conversions, whether setup is included or computed, and whether input currents and waves are physically and temporally consistent.
Measurements, QA and calibrationWave buoy, satellite altimetry, independent hindcast and meteorological measurements for calibration/validation.Nearshore gauges, pressure corrections, arrays, video/radar where appropriate, runup observations and harbour measurements.Follow instrument-specific procedures and QC flags, uncertainty budgets, record length and observation time bounds; exclude withheld/rejected data and avoid circular calibration.

Wind convention: ISO 19901-1:2026, Clause 3.21, identifies 10 m as a standard reference elevation and 10 minutes as a common mean-wind averaging period. This is not a prescription that a wave model must receive a wind field every 10 minutes. The elevation reference (e.g. above mean sea level) must be checked; numerical weather products may already provide 10 m winds over a model surface. An illustrative neutral logarithmic profile is:

Neutral logarithmic wind profile
U(z)=u∗κln(zz0)

Here U(z) is wind speed at height z (m), u* friction velocity (m s⁻¹), κ ≈ 0.40, and z0 aerodynamic roughness length (m). This expression omits stability, displacement and marine boundary-layer complications; it is not a validated universal height correction or a substitute for known wind-product metadata. Wind stress is commonly expressed τa = ρaCD|U10|U10 (N m⁻²), where the drag coefficient depends on the adopted parameterisation. Wind-input error therefore changes forcing nonlinearly, but no generic wind-speed-to-wave-height factor is implied.

Lab H · Wind forcing cadence and unresolved storm intensification

Compare a synthetic rapidly intensifying 10 m wind field with instantaneous wind snapshots at selectable cadences. A cubic wind-speed quantity is shown only as a dimensionless input-sensitivity proxy, not a wave source-term formulation or a predicted wave height.

Synthetic wind forcing sampling experimentBlue line shows a high-resolution synthetic storm, coral points show time-sampled instantaneous winds, and vertical grey lines show sample times. X axis is hours, Y axis is wind speed metres per second.

Assumptions: Gaussian peak superposed on constant background, exact instantaneous sampling without interpolation, 24 h window. Reanalysis, interpolation choice, track errors and wave-energy memory are not simulated. Higher-frequency model output cannot recover forcing information absent from the input.

31.20 · Hourly versus three-hourly sea states: what is actually being reported?

Four distinct timescales must appear in every model handover: wind/boundary update Δtforcing (s or h); numerical integration Δtmodel (s); saved output spacing Δtoutput (h); and the chosen sea-state statistical reference duration Tss (h). A phase-resolving model additionally has a high-rate sampled surface-elevation record from which statistics are estimated. None of these quantities can be inferred reliably from the others.

Additional warning for cyclones and squalls: A one-hour wind archive, a one-hour model wave output and a one-hour sea-state definition are three different constructs. The two-dimensional wind field may represent a valid-time analysis rather than a one-hour mean; a gust maximum since the preceding archive is yet another variable. See 31.28 and 31.30 for original moving-storm and response experiments.

Formal standard: ISO 19901-1:2026 §3.32 defines a sea state as a period in which statistics remain approximately stationary and ergodic. Its note describes three hours as often assumed for extratropical conditions and as short as one hour in tropical cyclonic conditions. These are contextual examples, not compulsory universal windows or proof of stationarity. Whether 1 h, 3 h or another window is representative must be established from site conditions, sample support and the intended calculation. IEC 61400-3-1:2019 is a distinct offshore wind-turbine design standard; its load-case context cannot be applied indiscriminately to port structures.

Record or productMeaningWhat it does not establish
3-hourly instantaneous spectral outputA spectral diagnostic at timestamps separated by three hours, as used in some Copernicus Marine wave products.Not necessarily the mean of the intervening three hours; short-lived peaks between timestamps may be missed.
Hourly instantaneous spectral outputAn output at each hour; spectra may vary between reports and over an assumed stationary sea-state interval.Not automatically a one-hour average, a continuously sampled observation, or validation of the sea-state assumption.
Hourly/3-hourly averaged outputA value aggregated over explicitly defined temporal bounds and averaging operation.Not necessarily an adequate single stationary sea state, and not mathematically the same as an instant at its centre/end.
Hourly observation reportMeasurement product published nominally each hour, potentially derived from a shorter acquisition record.Not a continuously observed hour; for example, NDBC documents 20- or 40-minute wave acquisition systems with rounded reporting timestamps.

For equally weighted spectra Ei(f,θ) defined on a consistent frequency–direction grid, the time mean spectrum and its zeroth moment are:

Weighted spectral mean
E―(f,θ)=∑iwiEi(f,θ)∑iwi
Weighted zeroth moment
m―0=∑iwim0,i∑iwi
Height from pooled variance
Hm0,pooled=4m―0

Because m0,i = Hm0,i²/16 for compatible spectral definitions, a variance-equivalent pooled significant height is √(ΣwiHm0,i²/Σwi), not the arithmetic mean of heights. This identity assumes each spectrum represents its assigned interval and weighting is justified. Three evenly spaced instantaneous spectra are only a quadrature approximation to a time mean; they do not by themselves validate the continuously evolving conditions. Mean direction is circular/spectral, not an arithmetic angle; peak period cannot safely be averaged as a scalar during multimodal or changing conditions.

Worked calculation (illustrative): If three equally representative hourly states have Hm0 = 2, 3 and 5 m, their arithmetic mean height is 3.333 m, whereas the variance-equivalent pooled height is √[(2² + 3² + 5²)/3] = 3.559 m. Neither result can be called a verified stationary three-hour sea state without further evidence. If these are merely three instantaneous snapshots, even the pooling assumes unproven temporal representativeness.

Lab I · 1-hour and 3-hour output intervals versus reference-duration averaging

Observe the difference between instantaneous samples (markers), time-block variance-equivalent Hm0 (steps), and a synthetic continuously evolving spectral-state diagnostic (line). Change a storm's width and its timing relative to output samples.

Sampling interval versus sea-state averagingBlue line is a synthetic evolving significant wave-height diagnostic, coral dots are instantaneous samples, and teal steps represent the square root of average squared significant wave height per block.

Demonstrates a synthetic time-varying Hm0 diagnostic (not individual waves or a resolved wave-energy balance). The block aggregation is a mathematical variance-equivalent statistic; it does not establish that the underlying interval is stationary, nor is it a universal correction from 3-hourly to hourly heights. The teaching stationarity index below is descriptive, not a standard acceptance threshold.

Required representativeness assessment: identify product time cell (point/mean/maximum), timestamps and time bounds; review within-window changes in spectral shape, height, periods and direction; test 1 h versus 3 h output/sampling sensitivity; evaluate storm peak and time-of-peak errors against independent measurements; and quantify the effect of record duration on design statistics. An hourly report based on 20 minutes of buoy acquisition must be compared with an appropriately matched model interval and observation uncertainty. Do not treat consecutive wave records as independent samples without assessing serial dependence.

31.21 · Input integrity at nested boundaries and shoreline datums

The model boundary is a transfer of information, not merely a file. A nested regional → coastal → phase-resolving workflow must check that the receiving domain retains the wave variance, directional distribution and environmental state needed to produce the engineering quantity of interest. The ability to reconstruct one time-domain realization from E(f,θ) does not mean that phase coherence or individual crests were present in the source spectrum.

  1. Define the coordinate system: geographical projection, east/north axes, vertical bathymetric datum, tide-gauge datum, sensor elevation, mean sea level and still-water level. State whether model depths already include tide and surge.
  2. Verify spectral dimensions: frequencies in Hz or intrinsic rad s⁻¹; directional density per radian or degree; bin boundaries and integration weights. Reconstruct m0, Hm0, mean periods and directional moments on both sides of interpolation. Separately test wave-action consistency if currents are relevant.
  3. Check directions: meteorological/nautical “from” versus wave-propagation “toward”, 0° at true north clockwise versus mathematical angle counter-clockwise from east. A silent 180° rotation can reverse the incident-wave exposure of a harbour entrance.
  4. Validate time metadata: forecast initialization and lead time, analysis-validity time, instantaneous versus averaged fields, time bounds, interpolation method, calendar and handling of missing boundary updates. Never silently equate an output frequency with a sea-state averaging interval.
  5. Control artificial boundary errors: incident/reflected-wave separation, boundary placement, damping/sponge-layer effectiveness, wave generation method, propagation direction and statistical convergence of synthesized irregular wave trains.
  6. Test coastal inputs: use IHO S-44 survey metadata when appropriate; check under-resolved bars, channels and reefs, grid convergence and morphology changes. Include site-specific bottom roughness and water levels with sources and uncertainty.
Reject or quarantine the boundary transfer when the frequency/direction convention is unknown, time cell methods are ambiguous, wave-energy integration changes without an explained physical/numerical reason, the bathymetric datum is inconsistent, or no independent check of incident-wave fidelity is available.

Minimal machine-readable metadata example: a NetCDF-CF wave dataset should describe units, projection, valid time, coordinate direction, and, for interval statistics, bounds and an appropriate cell_methods attribute such as time: mean. time: point describes a point diagnostic where applicable. CF conventions document a quantity's meaning; they cannot certify the physical accuracy of the model.

31.22 · Report model outputs as engineering variables, not just pictures

The modelling deliverable must explicitly distinguish raw spectral fields, derived bulk parameters, stochastic/extreme parameters and design-basis selections. For a variance-density spectrum E(f,θ) in m² Hz⁻¹ rad⁻¹, with angular integration in radians:

Directional spectral moment
mn=∬fnE(f,θ)dfdθ
Spectral significant wave height
Hm0=4m0
Mean period
Tm01=m0m1
Mean zero-crossing period
Tm02=m0m2

The units of mn are m² Hzn, provided frequency is in Hz. Peak period Tp = 1/fp depends on the identification and smoothing of the spectral peak, so a bimodal wind-sea/swell spectrum may have an unstable bulk peak period even when Hm0 is stable. A reported mean direction requires a defined angular convention and the appropriate spectral vector moments; a resultant direction alone conceals directional bimodality.

Output deliveredRequired accompanying evidenceWhat downstream engineers may reasonably do
Bulk Hm0, Tp, Tm01, Tm02, directionsFrequency limits, temporal support, sea-state duration assumption, treatment of spectra and nearshore processes, model uncertainty.Screen environmental conditions and feed preliminary design only within validated application range.
Full E(f,θ), partitions and directional momentsGrid definition, variable units, wave coming/going convention, wind-sea/swell separation method, variance-integral check.Evaluate multidirectional loading, harbour response, resonance and spectral boundary conditions in appropriate downstream solvers.
Long-term hindcast and storm-event time seriesPeriod of record, meteorological representativeness, homogeneity, gaps, model changes, event segmentation, conditional skill, uncertainty.Prepare extreme-value and joint-probability analyses with explicit serial dependence and climate representativeness assumptions.
Design environmental conditionsApplicable code and design-basis document; return period or exceedance specification; water level and wave direction/period combinations; epistemic and sampling uncertainty; sensitivity cases.Perform structure-specific loads, runup, overtopping, armour stability or operability assessments under the correct project design standard.
Phase-resolving surface elevations, kinematics and pressuresRun length, wave realization/seed, phase conditions, wetting–drying and breaking treatment, numerical convergence, location and reference level.Support suitably validated process-specific computations; do not infer impact pressure or structural force from significant height alone.

No direct leap from model Hm0 to an individual design wave: the maximum wave in a sea state and a structure's extreme response depend on duration, bandwidth, wave nonlinearity and statistical assumptions. Long-term extremes and joint conditions require an additional statistical design process. A 10% error in Hm0 changes linear wave energy by about 21%, but there is no universal 21% change in overtopping, damage or force.

31.23 · Evidence-based model validation with observation time support

Compare only physically comparable quantities. A numerical spectrum saved at 12:00 UTC and a buoy spectrum calculated from a preceding 20-minute acquisition do not automatically represent identical windows. Sensor response and correction, direction convention, tidal stage, location, spectral limits, stationarity and QC flags must be documented. Use Chapter 29 for instrument/process QC and Chapter 30 for statistical versus measurement uncertainty.

Prediction error
ei=Mi−Oi
Mean bias
Bias=1n∑i=1nei
RMSE
RMSE=1n∑i=1nei2

Both are in the units of the compared quantity (e.g. metres for Hm0). Direction error must be evaluated as a signed wrapped angular difference in [−180°,180°). Compare not just aggregate RMSE, but storm peaks, event timing, conditional high-wave subsets, wind-sea/swell regimes, directional sectors, quantiles and spectra. Use uncertainty-aware comparisons where observation uncertainty is material. Skill metrics do not replace an assessment of whether the output is fit for a design decision.

Two-stage hindcast validation: validate meteorological forcing first (U10 vector, wind averaging, event track, peak, passage duration) and wave performance second (height, period, directional spectra, event timing), ideally against independent stations and storms. Comparing wave Hm0 alone permits compensating wind/source-term/bathymetry errors. This principle is developed in 31.29 and its decision diagram.

  1. Observation qualification: record manufacturer calibration, acquisition times, raw-to-derived processing, instrument response, QARTOD or operator flags, uncertainty and representativeness.
  2. Matchups: document temporal support and allowable matching tolerance; compare collocated model predictions, account for spatial depth gradients and do not interpolate across flagged periods without an approved method.
  3. Split development from evaluation: define calibration dataset and parameters before tuning, then use independent times, locations or extreme events as holdouts; avoid storm-by-storm retuning against the validation sample.
  4. Test the right regimes: include storms, peak timing, high quantiles, bimodal seas, shallow-water breaking, swell arrival and potentially model-failure conditions, not just mean climate.
  5. Establish temporal adequacy: repeat comparisons at 30-minute, 1-hour and 3-hour output intervals using adequately sampled reference data; report how resampling changes event peaks and time-of-peak.
  6. Set acceptance before looking at results: project-, output- and consequence-specific criteria, supported by measurement precision and engineering tolerance. No international rule universally makes RMSE less than a fixed fraction of Hm0 sufficient for design.

Lab J · Validation can look good while storm peaks are wrong

Compare a synthetic high-resolution reference (teal) to a prediction with adjustable height bias and timing shift (blue), observed only at a chosen model-output interval (coral circles). Overall RMSE and peak errors respond differently.

Temporal-resolution model validation experimentTime series of synthetic observed and model wave heights with sparse model reporting dots, labelled hours on x axis and metres on y axis.

Synthetic paired experiment, no measured site data, no real solver. Model-at-observation matching is exact at plotted output times; a real deployment requires documented measurement duration and time matching. High-wave subset is pedagogically defined here by Hm0 ≥ 3 m, not an accepted universal design threshold.

31.24 · Consequences for structures, coastal defences and operational decisions

Model uncertainty becomes engineering risk when the prediction controls a decision: crest level, breakwater armour, bridge or deck clearance, line tension, offshore fatigue, installation windows, harbour downtime or inundation exposure. The consequence depends on the governing loading mechanism and joint environmental conditions, not simply on a single wave-height exceedance.

For linear random waves with variance m0, the mean wave energy per unit horizontal area and unidirectional power across a crest-normal section are:

Mean wave energy (J m⁻²)
E=ρgHm0216
Wave power per unit crest (W m⁻¹)
P=Ecg
Normal incident-power component
Pn=Pcos⁡α

Here ρ is water density (kg m⁻³), g gravitational acceleration (m s⁻²), cg linear group velocity (m s⁻¹), and α wave propagation angle relative to the section-normal direction. The normal-flux expression is valid for the single-direction example (and can be integrated over directional spectra for general seas). It does not calculate wave-induced force, breaking pressure or overtopping discharge. For a monochromatic constituent, ω² = gk tanh(kh) and cg = (ω/2k)[1 + 2kh/sinh(2kh)], with water depth h (m).

Lab K · Propagation of wave-height and period error into an energy/flux diagnostic

Change a reference sea state and a hypothetical model error. The tool computes linear wave energy and incident energy flux, not structural loads. Examine how a seemingly modest Hm0 bias produces a nonlinear change in an engineering input.

Wave energy and flux sensitivity experimentTwo original bar comparison panels for energy density and incident wave power per metre section, based on linear wave theory.

Linear single-direction analytical illustration with fixed ρ = 1025 kg m⁻³ and g = 9.81 m s⁻²; period here is a representative component period, not a substitution for an irregular-spectrum integration. No offshore/costal structural-design code accepts this as an overtopping, stability or impact-force calculation.

Coastal protection

Ask whether joint Hm0, period, direction and still-water level reproduce the forcing for runup, overtopping, armour stability and inundation. A design method such as EurOtop must be applied within its own conditions of validity; outputs are not universal functions of Hm0 alone.

Offshore structures

Specify design and abnormal environmental combinations, associated periods/directions/currents, individual-wave/crest statistics where required, response duration and uncertainty. ISO 19901-1 and DNV-RP-C205 address relevant metocean and environmental-loading considerations for their respective scopes.

Harbour and vessel operations

Preserve spectral shape, long-period content, directional exposure and resonance. Define operation-specific wave thresholds and weather-window persistence; instantaneous 3-hourly outputs cannot establish a continuous one-hour safe window without further evidence.

Extreme events and climate

Distinguish sample maxima from return levels and design action effects. Quantify missed short peaks, dependency between observations, track and intensity uncertainty, and whether a hindcast period represents future design conditions.

31.25 · Standards, acceptance evidence and the final metocean-to-design handover

The following are different classes of authority: ISO/IEC requirements, national codes of practice, a classification-society recommended practice, public engineering manuals, a hydrographic survey standard and a metadata convention. Selection is controlled by jurisdiction, asset type, contract and adopted design basis. A wave solver is not “ISO approved” merely because its results appear in a design report following a standard.

Document / evidence levelScope useful for this chapterImportant applicability boundary
ISO 19901-1:2026 — international standardMetocean environmental conditions for offshore structures. Clauses 3.21 and 3.32 provide documented wind and sea-state definitions, including reference elevation and contextual duration.Not an all-structures design code or a universal 1 h/3 h model sampling requirement. Full subscription clauses beyond accessible official preview not audited.
ISO 21650:2007 (confirmed 2022) — international standardWave/current actions on breakwaters, dykes, seawalls and specified cylindrical coastal structures.Does not prescribe a universally acceptable model skill or all harbour layout and sediment design activities.
IEC 61400-3-1:2019 — international standardFixed offshore wind-turbine environmental and design load cases.Not automatically applicable to conventional harbour or coastal-defence works; the standard explicitly is not a complete design specification.
DNV-RP-C205 (2025-04, amended 2026-03) — recommended practiceEnvironmental conditions and loads for marine structures; supplemental industry guidance.Not a certification of a wave model or a site-specific acceptance threshold.
BS 6349-1-1:2025 — British code of practiceMaritime works: planning/design considerations over an asset lifecycle.Jurisdiction/contract dependent; full paywalled text not clause-audited.
USACE Coastal Engineering Manual (EM 1110-2-1100) — engineering manualPhysical wave climate, transformation and design-method context.Check part/chapter date and revisions; not a product endorsement.
WMO-No. 702 (2018, updated 2020) — WMO guidanceWave analysis/forecasting and verification practice.Not a universal structural design code or fixed validation pass/fail threshold.
IHO S-44 Ed. 6.2.0 (Oct 2024) — survey standardHydrographic survey quality relevant to bathymetric inputs.Survey-quality compliance alone does not verify nearshore wave modelling.
IOOS QARTOD surface-wave manual v2.1 (2019) — QC guidanceObservation flags, sensor-context tests and data-qualification procedures.Not a numerical hindcast accuracy standard; applicability is sensor-dependent.
CF Conventions 1.13 — metadata conventionTime coordinates, bounds, cell methods, units and ancillary flags for traceable datasets.Metadata conformance is not physical model validation.
Copernicus Marine GLOBAL_ANALYSISFORECAST_WAV_001_027 PUM (issue 1.5, Nov 2024) — product documentationConcrete documented example of 3-hourly instantaneous wave fields.Illustrates semantics of one product/version; not a universal rule for other wave datasets.
NOAA NDBC acquisition FAQ — observational documentationShows why the observation report interval and acquisition duration may differ (20- or 40-minute wave records).Actual sensor station, period and measurement method must be checked individually.

Engineering release gate: minimum information required for defensible downstream use

Release decision: If an engineer cannot establish whether “3-hourly Hm0” means a point diagnostic, a time average or a sea-state assumption, the product should not yet be treated as a verified 3-hour design sea state. Flag the uncertainty, obtain the missing metadata or repeat the temporal representativeness study. The same principle applies to hourly products.

31.26 · Spatially varying winds: resolution, averaging, reference height and the real forcing seen by waves

The input question is not simply “Do I have a 2D wind field?” A horizontally varying u10(x,y,t), v10(x,y,t) is normally needed where fetch, cyclone geometry, sea-breeze convergence, coastal shelter or storm motion materially affect wave generation. A spatially uniform wind can be a legitimate controlled experiment or short-fetch approximation. Conversely, a fine-looking map of coarse, time-averaged or biased winds can be less trustworthy than a verified simpler forcing. Wave-model grid refinement cannot recover unresolved atmospheric structure.

Ten-metre wind-speed magnitude
U10=u102+v102
Approximate vector wind stress
τ→≈ρaCD|U10|U10
Nonlinearity of vector averaging
⟨|U|U⟩≠|⟨U⟩|⟨U⟩

Here u10 and v10 (m s−1) are eastward and northward horizontal wind components, U10 is speed (m s−1), ρa is air density (kg m−3), CD is a dimensionless drag coefficient, τ is surface-stress vector (N m−2) and ⟨·⟩ denotes a stated temporal/spatial average. This is an illustrative bulk-stress relationship, not the full wind-input source term Sin(f,θ); drag and wave growth can change with wave age, stability, sea spray and extreme winds. Do not infer wave-height error directly from a stress-proxy difference.

Wind-field propertyWhat must be checkedFailure mode and engineering implication
Reference elevation and averagingWhether the product is actual or neutral-equivalent 10 m wind, averaged over 1/10/60 min or a point-time analysis; measurement height, MSL vs ground/surface and stability corrections.Converting gusts to sustained winds, mixing 1-min cyclone best-track and 10-min surface products, or applying an unsupported universal factor can bias wave generation and structural wind loads.
Spatial support and effective resolutionNative atmospheric grid, representativeness over a cell, actual resolved feature scale, coastal land–sea mask, shorelines, interpolation and storm displacement relative to grid nodes.Small high-wind cores and gust fronts are smoothed, shifted or omitted, even if visualising the field on a finer wave grid. Mere bilinear interpolation adds no meteorological information.
Temporal supportValidity time and UTC, instantaneous analysis versus accumulation/mean/max, 10-/30-/60-/180-min archive step, interpolation operator and maximum between archives.Rapid intensification and convective passages are missed; an apparently continuous wind animation can be generated from two inadequate snapshots.
Vector/direction metadataEarth-relative or grid-relative east/north components, rotation on projected meteorological grids, meteorological “from” versus propagation “toward” convention.Incorrect rotation or direction convention creates false fetch, shifts the wave approach direction and changes coastal exposure.
Extremes and air–sea couplingStorm intensity, radius of maximum wind, travel speed, duration, pressure, drag formulation, wave-current interaction and model-specific wind-input physics.A plausible wind map can give wrong wave-growth timing or wave height; parameter uncertainties are strongly regime-dependent.
Independent evidenceUse quality-controlled buoys/platform anemometers, satellite scatterometer or SAR winds where fit-for-purpose, coastal stations with exposure correction, best-track/radar event timing and wave measurements.Point observations and satellite cells do not share the same spatial/temporal support as a model grid cell. Satellite rain contamination, coastal proximity and instrument limitations require explicit screening.

Reference boundary: ISO 19901-1:2026 Clauses 3.21, 3.43 and 7 define terms, including 10 m / 10-min mean wind reference and squall context; they do not require every coastal spectral model to use WRF or prescribe a universal meteorological grid size. WMO-No. 702 provides wave-analysis/forecasting guidance. The SWAN technical manual illustrates how one spectral implementation uses 10 m wind and a wind-input source term; its parameterisation is not a rule applicable to all solvers. ISO preview · spectral wind-input formulation · WRF wind components.

Original wind-input and wave-model evidence chainAtmospheric analyses or observations enter optional event-aware downscaling, are tested against independent observations, drive spectral wave calculations, and yield metocean outputs assessed for structural design. False precision and unvalidated transformations can occur at each interface.Atmospheric inputsU10(x,y,t), pressure,track, surface physicsQuality + time supportEvent-aware optionsMeso-NWP / parametricstorm reconstructionValidate before useWave computationAction balance +bathymetry, boundaryWind + wave checksEngineering releaseH, T, θ, spectra,extremes, uncertaintyFit for intended design?Independent QC and validation at EACH handoverFiner output ≠ finer forcing; better wind RMSE ≠ adequate cyclone / squall peaks; wave-model fit ≠ design certification.
Figure 31.12 · Original quality-assurance chain. The optional atmospheric reconstruction box can be bypassed: use it only where justified by a quantified improvement and available evidence.

31.27 · Tropical cyclones: the high-wind core, asymmetry and wind-to-wave consequences

Many gridded reanalyses and global forecasts represent cyclone-scale circulation well but can underestimate or displace the sharp eyewall wind maximum, particularly for compact storms or rapid intensification. A local refinement of the wave mesh cannot correct an atmospheric track, intensity or high-wind-core error. Conversely, a very fine wind field is not sufficient if the wave model mesh, source terms, currents, bathymetry or numerical transport remain inadequate. Chen, Ginis & Hara (2018) found a strong wave-grid-resolution sensitivity in idealised cyclones and explicitly noted that wind smoothing is not the only mechanism.

Storm properties to audit together: centre position and uncertainty (km); translation velocity and heading (m s−1, degrees); intensity and sustained wind reference (m s−1, stated averaging period); minimum central pressure (hPa); radius of maximum wind (km); azimuthal asymmetry and environmental flow; eyewall replacement; and event cadence (minutes/hours). A peak Hm0 comparison alone cannot establish correct storm-wave evolution: errors can compensate between winds, drag and numerical source terms.

Idealized tangential wind profile
Uθ(r)=UmaxrRexp(1−rR)
Radius from the storm centre
r=(x−xc)2+(y−yc)2

This original, deliberately simple analytic vortex has maximum tangential wind Umax at radius R; it is used only for the resolution demonstration below. It is not a Holland pressure–wind model, operational cyclone hindcast or physically realistic inner-core simulation. The toy field neglects asymmetric boundary-layer winds, land interaction, inflow, turbulence and storm translation effects.

Lab L · The cyclone eyewall against a coarse 2D wind grid

Compare the fine analytical vortex to cell-averaged horizontal wind vectors on a coarse grid. Grid-cell averages can cancel/attenuate adjacent turning winds. The chart highlights the radius of maximum wind and shows the resulting maximum resolved cell-average speed. No wave-model result is computed.

On a narrow screen, scroll the diagram horizontally to see both panels and all annotated axes.

Idealised 2D cyclone wind-resolution experiment

Calculating cyclone-resolution indicators.

The analytical peak is 50 m s−1. Cell values are spatial vector means computed over each cell, not maxima and not physically time-averaged atmospheric model outputs. The experiment demonstrates information loss; it does not quantify the error of any named atmospheric dataset.

For engineering reconstruction, alternatives include higher-resolution analysed winds, event-specific satellite or airborne estimates, data-assimilating mesoscale atmospheric simulations, and parametric tropical-cyclone wind models when they are fit for purpose. Parametric approaches may repair an underspecified compact vortex but must be checked for track, pressure–wind relationship, asymmetry and land interactions. They cannot be assumed superior to dynamically downscaled winds. Wind averaging conversions should use documented measurement-specific guidance, not a fixed gust multiplier; see WMO/TD-No. 1555.

Important evidence distinction: An ECMWF ERA5 field at 31-km native spatial resolution is not a 31-km-resolved eyewall. Technical documentation warns that very high analysed 10 m winds can be unrealistic at that scale; other studies demonstrate the opposite problem—underprediction of strong winds. Neither finding justifies applying an automatic blanket multiplication to all ERA5 winds. ECMWF documented limitation · observational assessment of offshore strong winds.

31.28 · Convective squalls in the Strait of Malacca and Southeast Asia: space–time aliasing

A Sumatra squall is an organised convective line originating over Sumatra or the Strait of Malacca and typically travelling eastward toward Peninsular Malaysia and Singapore. The Meteorological Service Singapore describes sudden gusty surface winds of 40–80 km h−1 and heavy rain often lasting about 1–2 hours over the island; MET Malaysia discusses west-coast storm lines. These are weather phenomena, not a guarantee that a similarly long period of sustained offshore wind-wave forcing occurs at every site.

Distinguish (i) a few-second gust affecting local loads; (ii) the spatially coherent, sustained component that can grow short-period, fetch-limited wind sea; (iii) the pre-existing swell; (iv) wave-propagation time and lag behind the moving wind front; and (v) tide/current/depth effects at the engineering site. An abrupt anemometer peak is not an instant increase of Hm0 by the same proportion. Some small or sheltered basins may not develop severe local wind sea despite violent gusts; vessel operations and wind-sensitive structures may still be exposed.

The moving-disturbance sampling criterion is more general than a fixed “hourly versus 3-hourly” rule. If a front has characteristic horizontal width ℓ (m) and speed V (m s−1), its local passage scale is τpass ≈ ℓ/V (s). This is not a Nyquist-compliant reconstruction rule for a Gaussian front: meaningful peak capture requires a forcing archive substantially shorter than the event passage scale, and its suitability must be demonstrated through event and wave validation. The wave-model internal time step cannot restore wind peaks missing between forcing timestamps.

Lab M · A moving squall line sampled by a coastal wind archive

A synthetic 2D space–time wind-pulse travels eastward. The upper diagram is a distance–time view; the lower diagram is the time history at the indicated coastal site. Change front width, movement speed and forcing archive interval to observe aliasing. The pulse describes a sustained-equivalent toy wind field, not individual measured gusts or a convection-resolving NWP simulation.

On a narrow screen, scroll the diagram horizontally to see both panels and all annotated axes.

Moving squall: time-space aliasing illustration

Calculating sampling effects.

The fine 1-min synthetic reference is an exact evaluation of the prescribed pulse, not a measurement. Reported peak loss refers only to the archived wind samples at the fixed site; it is not the predicted wave-height loss. Severe squall verification should use radar/satellite event timing plus calibrated marine/coastal winds and wave observations where available.

Observation verification must reflect both the hazards and the intended wave output. A meteorological model can have the right rain-line position but an inaccurate marine wind vector; a wave model can correctly reproduce swell while underestimating locally generated steep wind sea. Case-based studies of Singapore's Sumatra squalls demonstrate that convection-system forecasts warrant independent objective evaluation, not visual acceptance of a high-resolution weather plot. See the published SINGV forecast evaluation.

31.29 · Should atmospheric conditions be dynamically downscaled? WRF is an option, not an acceptance criterion

When coarse atmospheric products miss spatial gradients or short-duration events important to the quantity of interest, a nested meteorological hindcast such as WRF or another mesoscale numerical weather prediction system is a candidate method. It should not be selected solely because “high resolution is better.” Dynamical downscaling can misplace convection, over/underproduce gust fronts, change cyclone intensity, generate spurious land–sea-breeze patterns, inherit erroneous global lateral boundaries or be sensitive to cloud microphysics, planetary boundary-layer and convective parameterisations.

Trigger / questionOptions worth testingEvidence needed to select them
Strong but broad monsoonal winds; large fetchQuality-controlled regional/global gridded winds may be sufficient; test coastal gradients and large-scale direction.Wind/wave skill across seasonal sectors, representative extreme periods, native support, and adequate coastal masks.
Compact/rapid cyclone with displaced eyewallHigher-resolution assimilating NWP, event reconstruction, parametric vortex with verified track/asymmetry, multi-source blending; nested wave mesh if necessary.Track, peak and radial wind profiles, independent point/satellite/radar observations, storm swell arrival and tail-conditioned wave error.
Moving Sumatra squall / gust frontConvection-permitting modelling, radar-supported analysis and case-specific marine wind reconstruction; separate short-duration gust hazard from wave hindcast.Gust-front passage, wind-vector duration and track, coastal/marine validation, and downstream wave response at appropriate time support.
Coastal topography, island shadowing, sea breezeCoastal meteorological mesh, observed direction corrections, alternative local wind analyses; investigate land mask and atmospheric stability.Site exposures, buoy/mast observations, shoreline orientation, low-level profiles and fetch-direction skill.
Insufficient high-resolution observationsUncertainty envelopes and competing wind reconstructions rather than a uniquely “corrected” wind field.Identify irreducible uncertainty; do not present unvalidated downscaled weather as ground truth.

Suggested comparison experiment: Hold the wave solver and bathymetry fixed while changing only the atmospheric forcing dataset; then compare both the winds (speed, direction, event peaks, location, duration and time support) and the resulting waves (Hm0, Tp, directional spectra, storm timing and high-quantile errors) against independent observations. Next, hold winds fixed and alter wave physics, grid or source terms to separate causes. Where possible use an event-and-site holdout rather than tuning all scenarios against the same buoys. This is an original recommended experimental design, not a specified ISO pass/fail algorithm.

Scientific evaluation protocol for possible atmospheric downscalingTwo atmospheric forcing candidates flow into matched wind validation and the same wave model, followed by matched wave observations and engineering interpretation. Rejection or acceptance requires both atmospheric and marine evidence.Baseline atmospherenative 2D gridded U10,V10known product time supportCandidate alternativeWRF / other NWP / vortexsame wind conventionsStage 1: wind skillIndependent wind + radar;event timing, direction,peak, width, durationRecord unresolved peaksStage 2: wave skillSame wave setup;compare buoy spectra,storm peaks / phasesNo retuning per windStage 3: decisionQuantify design QoIsensitivity / extremesand uncertaintyAdopt or reject, document
Figure 31.13 · Original sequential validation design. A finer downscaled atmosphere is not automatically accepted. This follows the distinction between evaluating wind forcing and validating wave-model output.

Lab N · Why temporally averaged wind vectors can alter a nonlinear stress proxy

Two wind regimes alternate within a notional hour. Their direction, intensity and duty cycle determine (i) the magnitude of average wind, (ii) the magnitude of the mean stress proxy ⟨|U|U⟩ and (iii) the proxy computed from the average vector |⟨U⟩|⟨U⟩. The unequal values arise from nonlinear averaging, not from a wave-height formula.

On a narrow screen, scroll the diagram horizontally to see both panels and all annotated axes.

Nonlinear wind-stress proxy and vector-averaging demonstration

Calculating wind averaging comparison.

Use ρaCD constant for this diagnostic only; plotted values are in m² s−2 before multiplying by ρaCD. The source-term physics in a spectral wave model is more complex than this proxy, and the strong regime must not be substituted for a sustained wind unless its support is justified.

For WRF, archived U10 and V10 are documented as grid-relative wind components in standard output. Rotate them to true earth east/north when needed before applying meteorological-direction or coastal-wave conventions; verify projected-grid units and native output intervals. The WRF guide also describes interval maximum wind diagnostics, which are not equivalent to interval mean forcing fields. WRF user documentation.

31.30 · Wind temporal variability is not an instantaneous sea-state response: closing the design evidence loop

Wind sea responds through a coupled spectral source/propagation balance, not simply by mapping the maximum wind speed to a maximum wave height. Wave growth depends on the spectral action, duration, fetch and directional structure of wind, currents and sinks. A squall can strongly influence local short-period wind seas without producing a coincident long-period offshore swell response. A moving cyclone creates wave fields with memory and fetch asymmetry; maximum waves may not lie at the maximum instantaneous wind pixel. Model-output time spacing and the sea-state reference duration remain distinct (Section 31.20).

Educational linear relaxation model (not a wave solver): The next experiment uses a dimensionless response proxy q, with Q(U)=[max(U−5,0)/20]² and dq/dt = [Q(U)−q]/τ. Wind U is in m s−1, t and τ are in minutes. The differential equation is a one-box response analogy; it omits frequency–direction spectra, fetch, refraction, wind growth physics and dissipation. It intentionally does not calculate Hm0, wave period, structural loads or an actual sea state.

Lab O · Why high-frequency wind detail and wave-response timescale both matter

Compare high-rate synthetic wind with linearly interpolated archived winds, then feed each into the same response analogy. Changing wind-pulse width and response time shows how forcing aliasing and system memory are separate effects.

On a narrow screen, scroll the diagram horizontally to see both panels and all annotated axes.

Wind temporal sampling and response-memory analogy

Calculating response proxies.

One-dimensional synthetic forcing and an independently defined ODE, included to demonstrate timescales—not a validated sea-state calculation. A real offshore-to-nearshore hindcast would require a full directional spectrum and comparison to observed wave evolution.

Engineering-grade acceptance: the questions a modelling team must answer

Decision consequenceEvidence / diagnosticWhen to withhold design use
Storm-dependent extreme Hm0 and periodsIndependent cyclone/squall track and wind-vector skill; storm spectra, peak/timing errors and time-sampling sensitivity; physically appropriate extreme-value methods.High-wind cores or extreme sea states are systematically missed, or observation time supports are incompatible and unexplained.
Breakwater armour and overtoppingIncident directional spectra, wave height/period joint conditions, water levels, local transformation and sensitivity of a chosen approved downstream engineering formula.Nearshore wave direction, breaking or water level has not been validated for relevant extremes; a bulk height comparison alone is insufficient.
Port/harbour operations and downtimeDirectional fetch/squall timing, short-period wind-sea spectra, long-period energy and duration/persistence analysis matched to operational thresholds.Only 3-hourly instantaneous wave output exists but one-hour weather windows or short transients are being certified without further evidence.
Offshore structures and vessel/mooring responseAdopted design basis and code, environmental contours/joint conditions, kinematics/response spectra where applicable, fatigue exposure duration, uncertainty.Wind averaging and sea-state duration conventions are undisclosed; wave products are extrapolated outside validated regimes.

Practical acceptance decision: The correct atmospheric product is the one that demonstrates sufficient validated performance for the specific wave and design quantities that matter, across representative conditions including relevant extremes. Document an explicit decision to retain coarse forcing where justified, or a substantiated need for higher-resolution reconstruction where not. No WMO, ISO, USACE, DNV or WRF source cited here imposes a universal requirement to dynamically downscale every coastal wave study.

Additional primary guidance and peer-reviewed source trail

Source / typeDirectly supportsDoes not establish
ISO 19901-1:2026 — international standard, official publicly available definitions / indexDefinitions of sea-state, mean and gust winds, squalls; design metocean context. Relevant visible headings include storm types, directionality and thunderstorm/squall considerations.Any numeric universal input spacing, accepted cyclone model or guaranteed downscaling improvement; full paywalled clauses have not been audited.
WMO-No. 8 (2024), Vol. I, Ch. 5 — observational guidanceSurface-wind observing instruments, exposure and methodological context for quality-controlled wind measurements.The October 2026 provisional edition is not yet formally approved as of this release; checking the index is not a detailed clause audit and no model accuracy is certified.
WMO-No. 702 (2018, updated 2020) — professional guidanceWave-forecast and analysis methods, sources and verification context.Automatic approval of a wind field or site-independent RMSE limit.
Chen, X., Ginis, I. & Hara, T. (2018), JMSE 6, 116 — peer-reviewed researchDependence of TC wave simulations on wave-grid resolution, radius and translation speed; coarsened winds are not the only issue.A transferable percent error for all cyclones, locations and solvers.
Osinski, R. D. & Radtke, H. (2020), Ocean Science 16, 355–371, DOI: 10.5194/os-16-355-2020 (2020) — peer-reviewed researchSensitivity of wind-driven waves to atmospheric forcing's spatial and temporal resolution, particularly short events.That running WRF always improves the result, or that subhourly wind is always required.
Meteorological Service Singapore and MET Malaysia — national meteorological agenciesSumatra squall weather characteristics and regional context.Direct wave-height estimates at coastal engineering locations.
Objective/subjective evaluation of Sumatra squall forecasts — peer-reviewed researchIndependent weather-model validation for regional convective systems.Validation of any particular coastal wave model under squalls.
WRF official output guide and SWAN computational-grid guide — implementation documentationNative gridded wind components and the distinction between computational step and forcing-field accuracy.A general certification of either numerical system.

31.31 · Reconstructing tropical-cyclone winds: when the Holland family is appropriate

Reconstruction is a method, not an accuracy guarantee. A parametric tropical-cyclone (TC) vortex is one defensible candidate when coarse atmospheric products poorly resolve the pressure gradient and intense wind core, when historic tracks need reconstruction, or when synthetic-storm ensembles must drive wave and surge calculations. The choice should be based on the engineering quantity of interest: peak wind and surge, outer-core wind fetch and wave growth, wave direction and period, or storm-related design extremes. It must be compared with observed and dynamically downscaled winds wherever these exist.

In earlier Lab L, the analytic toy vortex is deliberately not Holland; it illustrates the effect of 2D wind-cell averaging. The new Labs P and Q below instead use the published Holland (1980) radial pressure and gradient-wind equations. They remain educational, stationary, symmetric and Northern-Hemisphere-only, not a replacement for an operational moving-storm 10 m wind-field reconstruction.

Candidate forcingPotential benefitCritical failure modeEvidence before wave-model use
Observed or evaluated gridded atmospheric windsRepresent background weather and interacting meteorological systems; may capture storm asymmetryIntensity/position errors, spatial smoothing and missed rapid intensificationNative-resolution wind vectors, valid-time definition, core/outer-wind comparisons and event timing
Holland 1980 radial vortexTransparent pressure–wind relationship that can recover a compact core using a few known storm parametersIdeal axial symmetry, uncertain radius/shape; gradient winds are not measured surface windsTrack, central/environmental pressures, RMW, wind radii, 10 m conversion and independent data
Revised Holland 2008/2010 or asymmetric variantsCan constrain peak or outer-wind structure with additional empirical predictors and measured isotachesRegional empirical transferability, structural asymmetry and parameter uncertaintyExact published variant, fitted parameters, observational support and independent validation
Mesoscale NWP/downscaling or blended fieldsCan represent asymmetric circulation, land-sea contrast and other concurrent weatherConvective/physics biases, inappropriate blending or double counting translation/background windsReproducible setup, blending transition, wind/pressure consistency and independent wind/wave skill

Important wave-specific nuance: The strongest gradient wind at the eyewall does not alone determine the largest waves. Storm extent, R34/R50/R64 wind radii, persistence, track, translation, directional fetch and alignment with wave propagation can dominate at a coastal or offshore point. A reconstruction calibrated only to maximum wind can overestimate a compact core and underestimate an expansive outer wind field, or vice versa. This is why surface-wind isotaches and directional wind measurements matter (Holland et al., 2010; ADCIRC GAHM documentation; relevant TC wave-size research).

31.32 · Holland (1980) mathematical formulation: from pressure to gradient wind

For an idealized, axisymmetric cyclone, let r be radial distance from the cyclone centre (m), pc the central sea-level pressure (Pa), pn the environmental pressure (Pa), Δp = pn − pc > 0 (Pa), R a radial scale commonly termed the radius of maximum winds in the cyclostrophic model (m), and B a dimensionless profile-shape parameter. Define x = (R/r)B. Holland's pressure expression is:

Holland radial pressure profile
p(r)=pc+Δpexp[−(Rr)B]

The pressure gradient follows from the chain rule, as shown in the equation panel. At the gradient-wind level, assume steady, circular, inviscid balance for the cyclonic speed magnitude Vg:

Pressure-gradient derivative
dpdr=BΔprxexp⁡(−x)
Dimensionless radius
x=(Rr)B
Gradient-wind balance
Vg2r+|f|Vg=1ρadpdr
Cyclonic gradient-wind solution
Vg(r)=BΔpρaxexp⁡(−x)+(|f|r2)2−|f|r2

Here ρa is air density (kg m−3), f = 2Ω sin φ is the Coriolis parameter (s−1), Ω ≈ 7.292115 × 10−5 s−1, and φ is latitude (positive north, negative south). Speed and pressure use SI units; 1 hPa = 100 Pa. In the cyclostrophic approximation (neglecting Coriolis near a sufficiently compact/intense core), Vc = √{(B Δp/ρa) x exp(−x)} and its exact maximum is at r = R. Consequently Vc,max = √{B Δp/(ρa e)}. With finite Coriolis, the exact maximum of the gradient-wind curve need not occur at R; a measured 10 m Vmax must not be substituted directly into this gradient relation.

Parameter inversion and uncertainty. Under the cyclostrophic assumptions, B = ρa e Vc,max2/Δp. This is not a generic estimator for B from a 1-minute or 10-minute observed surface-wind maximum. A measured surface maximum must first be interpreted at the correct height, averaging duration and flow level; translation asymmetry and the pressure–wind relationship must be considered. The 2008 Holland pressure–wind formulation and 2010 revised radial formulation address some of these issues using empirical and additional radial observations. Neither establishes an invariant B, surface-wind reduction, or wind-period conversion for all storms.

Worked calculation (idealized, not a hindcast): For pn = 1010 hPa, pc = 950 hPa, Δp = 6000 Pa, B = 1.5, R = 35 km, φ = 20° N, ρa = 1.15 kg m−3: p(R) ≈ 972.07 hPa, cyclostrophic Vc(R) ≈ 53.66 m s−1, and gradient-level Vg(R) ≈ 52.79 m s−1. These are NOT 10 m sustained winds or predictions of significant wave height.

31.33 · Building a moving, two-dimensional cyclone field for a coastal spectral-wave model

1 · Recover trackUTC position, central pressure, maximum sustained wind, motion and uncertainties
2 · Specify shapeEnvironmental pressure, RMW, B, observed wind radii and radius/azimuth variation
3 · Form 2D vortexPressure-gradient field, hemisphere sign and rotational orientation
4 · Derive surface forcingCalibrated boundary-layer winds, inflow, translation, land effects and background blend
5 · Validate & nest10 m U/V and pressure QC, wave/surge checks, sensitivities and design release

At every model time t, interpolate physically consistent TC parameters along the storm track (xc(t), yc(t), Δp(t), R(t), B(t)). On a local east–north plane, compute r = √[(x − xc)2 + (y − yc)2]. For the Northern Hemisphere, an ideal tangential unit vector is eθ = (−Δy, Δx)/r; reverse its sign in the Southern Hemisphere. The gradient-level vector is Vg(r) eθ. It must be transformed to an appropriate near-surface, earth-relative wind vector before use in surface-wave generation, accounting for boundary-layer wind reduction, inward inflow and storm motion when justified and consistently defined.

Do not add storm translation speed uncritically. Different parametric models incorporate translation asymmetry with distinct empirical corrections; some empirical peak-wind parameters already contain asymmetry. Applying multiple translation additions or background-wind blends can double count winds. A 10 m wind reference height, sustained-wind averaging interval, land/sea roughness and time interpolation must be documented. Wind profiles do not reconstruct an eye, rainbands, extratropical transitions, environmental shear or eyewall replacement by themselves.

Track archives commonly contain approximately 6-hourly best-track fixes; a model requiring 10-minute or hourly forcing must interpolate or otherwise reconstruct the intervening evolution. Interpolation creates intermediate values, not observed 10-minute rapid-intensification detail. Use observed 34/50/64-knot wind radii (and quadrants) to constrain outer wind structure where available. For a compact storm, nearby coast, rapid intensification, landfall or complex land masks, a blended or asymmetric approach may need stronger observational support than the basic symmetric Holland formulation can provide.

Engineering handover condition: The two interactive experiments below calculate gradient-level winds from a stationary axisymmetric cyclone. They intentionally do not prescribe a universal surface-wind reduction or produce time-varying 10 m U10/V10. A production reconstruction requires verified boundary-layer assumptions, storm translation/asymmetry treatment, georeferenced land-sea masks, consistent time bounds and independent observations before it is used as a wave-model input.

31.34 · Interactive reconstruction experiments: radial physics and 2D grid losses

Lab P · Holland pressure and gradient-wind profiles

Adjust pressure deficit, radius and B independently. The left chart compares the analytical cyclostrophic and finite-Coriolis gradient-wind curves; the right shows the associated sea-level pressure field. Check whether the wind maximum and the gradient-wind maximum radius are exactly the same.

Holland analytic pressure and wind radial profiles

Pressure assumptions: pn = 1010 hPa; pc = pn − Δp. Constant air density ρa = 1.15 kg m−3 for this controlled teaching exercise. Wind is gradient-level, not observed or parameterized 10 m surface wind. Axes are shown in kilometres, metres per second and hectopascals.

Calculating analytical profile…

Lab Q · Two-dimensional Holland wind vectors versus coarse-grid cell averaging

A fine-sampled analytic gradient-wind field is compared with a synthetic coarse-grid vector cell average over the same spatial area. This is an averaging experiment, not a measured atmospheric grid or an actual coupled wave-model hindcast. Cell averages may suppress the vortex core or obscure direction changes even when the mathematical cyclone is unchanged.

Holland two-dimensional wind resolution and spatial averaging

Both panels show gradient-level wind vectors in m s−1, in an idealized 320 × 320 km Northern Hemisphere east/north domain. Colours indicate wind magnitude; the dashed circle is the prescribed cyclostrophic radius R. The fine panel uses 4-km midpoint samples; the coarse panel averages wind vectors using numerical quadrature, rather than sampling wind-speed maxima. The plotted maps are for physics education, not validated forcing products.

Calculating spatial sensitivity…

Lab R · Moving cyclone over a fixed point: can coarse output miss the wind event?

Translate the same idealized Holland gradient-wind vortex along a straight track past a fixed location, keeping intensity and structure constant. Compare a fine, one-minute synthetic reference with instantaneous 1-, 3- or 6-hourly output samples. This isolates temporal sampling and track-geometry effects; it is not a dynamic atmospheric hindcast and does not account for ocean-wave response memory.

Holland moving-vortex time-sampling illustration
Calculating moving storm response…

The closest approach occurs at a prescribed model time of 12.7 h; the reference is evaluated every minute, and comparison samples fall at exact whole-hour, three-hour or six-hour timestamps. Maximum wind at a site need not occur at closest approach if the site passes within the eyewall radius. No wind-speed averaging, surface-wind conversion, wave growth, track curvature or storm intensification is simulated.

31.35 · Evidence, parameter fitting, validation and downstream design decisions

The engineer should evaluate three independently traceable products: (i) cyclone-track and intensity inputs; (ii) reconstructed meteorological fields; and (iii) resulting directional wave conditions (and surge where coupled). A parametric wind-field reconstruction can fit surface wind observations and still produce poor offshore waves because fetch, outer wind radii, track timing, directional alignment and spectral source terms remain incorrect. Conversely, good wave-height agreement alone can conceal errors in period, direction, spectral partitioning or validation sample timing.

LayerMinimum recommended evidenceKey engineering issue
Track/intensityBest-track version and agency; centre coordinates/UTC; maximum wind averaging; pc; pn; RMW; R34/R50/R64 quadrant radii; uncertainty and interpolation methodLocation and timing error can dominate local peak conditions
Wind/pressure reconstructionPressure/radial profile comparison; 10 m scalar/vector winds at independent stations, buoys, scatterometer/SAR where suitable; temporal and spatial matching; boundary-layer and land-decay documentationCorrect core peak but erroneous outer-field fetch or vector direction
Numerical wave responseWave-buoy time series and measured spectra; Hm0, Tp, Tm−1,0 where defined, direction, storm-peak timing; wave-grid/source-term convergence and forcing sensitivityUnresolved spectral peaks, wave height bias, incorrect joint design conditions
Design releaseEvidence for extreme-event representativeness; multiple plausible cyclone size/track/B scenarios; engineering quantity sensitivity; transparent uncertainties; relevant adopted design standardDo not equate a reconstructed wind-field maximum with a design wave or structural load

Independent evaluation: preserve an independent assessment period or event set; fit/calibrate shape and surface reduction only with their designated data; inspect wind radii and storm-sector errors rather than just overall RMSE. For wave validation, match observation integration periods to instantaneous versus averaged model outputs (Section 31.20); inspect timing offsets and extremes separately. Apply plausibility/physical consistency tests, temporal and spatial refinement tests, and uncertainty scenarios over R, B, Δp, storm centre, translation, outer-wind structure and the wind-to-surface conversion. An apparently improved wind maximum does not prove an improved design distribution.

Standards and authority: ISO 19901-1:2026 addresses metocean inputs used in offshore design; WMO-No. 702 supports wave-analysis and evaluation practice, and DNV-RP-C205 is an offshore environmental loads recommended practice. These documents are not represented as mandating Holland reconstruction. The equations and modifications described above come from research literature and documented model implementations, not from a claimed ISO/DNV approval.

Primary research and implementation references (distinct methodological roles)

ReferenceRole and limitation
Holland, G. J. (1980), An Analytic Model of the Wind and Pressure Profiles in Hurricanes, MWR 108, 1212–1218Classic radial pressure-gradient and wind theory; ideal symmetric structure; not a 10 m wind or wave solver
Holland, G. (2008), A Revised Hurricane Pressure–Wind Model, MWR 136, 3432–3445Empirically parameterised pressure–maximum wind relation; its surface-wind parameter is not interchangeable with the 1980 gradient-level B
Holland, G., Belanger, J. & Fritz, A. (2010), A Revised Model for Radial Profiles of Hurricane Winds, MWR 138, 4393–4401Radial-profile extension using outer wind observations and potentially surface winds; not a complete two-dimensional asymmetric/boundary-layer model on its own
ADCIRC Generalized Asymmetric Holland Model, technical documentationDocumented example of quadrant isotaches, asymmetry, surface conversion and storm motion in an operational modelling framework; not an industry-wide compulsory algorithm
UCAR tropical-cyclone best-track / ATCF data documentationStorm-track variables and typical six-hourly cadence; best tracks and wind radii carry observational and analysis uncertainty
Landsea & Franklin (2013), Atlantic Hurricane Database Uncertainty and Presentation of a New Database Format, MWR 141Data uncertainty and smoothing in historical tracks; explicitly cautions against inventing unobserved sub-six-hour storm variability
Interpretation: Holland reconstruction is most useful as a documented physical or statistical option in a comparison and validation workflow. It is not a reason to dismiss gridded meteorology, omit surface-wind transformations, or transfer wave-model predictions directly into structural design without an appropriate design-basis assessment.
Connection to earlier chapters: 28 — Wave instruments and response → 29 — Data QC → 30 — Uncertainty → 31 — Model-to-design evidence chain. These lessons form one workflow, not independent compliance certificates.
\n\n