The North Sea does not break by region

A swell can pass an offshore buoy, reach the coast, and produce different surf within a few kilometres. One beach lines up. Another stays soft. A harbour arm blocks one direction and accelerates current in another. A shallow bank breaks the energy before it reaches shore.

The offshore forecast can be right while the local call is wrong.

Offshore numbers stop offshore

Wave height, period, direction, and wind are the boundary conditions. Between the model grid and the beach, waves cross uneven depth. They shoal, refract, spread, lose energy to the bottom, and break over banks.

Deltares models these processes with spatial bathymetry and full wave spectra. Belgian coastal researchers use local measurements and a 250-metre wave model because the shallow seabed changes conditions from place to place.

A single regional bank profile cannot reproduce Scheveningen harbour, the bars at Egmond, the engineered coast at Camperduin, the Flemish Banks, and the broad western Belgian beaches.

The coast moves

The bank in front of a beach is not permanent.

Rijkswaterstaat has surveyed Dutch coastal profiles since 1965. Those records show a coast shaped by migrating bars, nourishment, storms, inlets, and harbour works. Research at the Dutch coast has measured intertidal bars moving tens of metres within weeks. The Hondsbossche coast near Camperduin was rebuilt with millions of cubic metres of sand and continues to be maintained.

Belgium is just as varied. Mariakerke is nourished and groyned. Groenendijk near Nieuwpoort has several natural intertidal bars. De Panne has a broad, gently sloping ridge-and-runnel beach. Oostende and Blankenberge are shaped by harbour structures.

Their forecasts need dated local geometry. Old bathymetry should lower confidence instead of passing as current fact.

Tide is a moving part

“High tide” is not a complete forecast.

Belgium's spring tidal range reaches about five metres. Nearshore currents can exceed one metre per second. The same water height can arrive on the flood, leave on the ebb, or sit near slack water. Wind and pressure can push the actual level away from the astronomical prediction.

Current matters most where depth and structures concentrate it. TU Delft work around Scheveningen shows harbour jetties accelerating flood current and forming eddies. Belgian studies found local changes in wave height and period when time-varying depth and current were included.

The right response is local: use continuous water level and setup everywhere; add detailed current effects where measurements and geometry justify them.

Local knowledge has a job

Goedegolven and Surfweer describe what institutional models do not label: which bank tends to work, when a harbour corner stays clean, and how a swell behaves after the turn of the tide.

That knowledge is valuable. It is also conditional. Banks move. Reports come from different boards, abilities, and parts of the beach. A good session is evidence for that time and place, not a permanent coefficient.

We use local reports to form a testable rule. Then we compare the rule with forecast snapshots, measured water level, nearby buoys, and later outcomes. A rule earns more weight when it survives new events.

The forecast needs four layers

1. Describe the sea. Keep wind sea and separate swell components, with the model run and uncertainty attached.
2. Transform it locally. Use the bank profile, harbour structures, tide, setup, and current regime for the named break.
3. Check the miss. Compare the model with suitable wave, wind, tide, and current stations. Correct short-horizon bias without treating an offshore buoy as shorebreak height.
4. Score the surf. Map the resulting break state to one NORDR tier using held-out local outcomes. Keep quality, hazard, and confidence separate.

NORDR already combines spot profiles, tide fit, model forecasts, and buoy context where coverage exists. The next layer is dated local geometry, measured setup, and held-out calibration across more breaks.

Confidence should expose the gap

A fresh weather model can still meet an old bank profile. A nearby buoy can be fresh but too deep or too sheltered for the break. A tide prediction can omit wind-driven setup. Two models can disagree on a northwesterly swell entering the North Sea.

Those gaps belong in the forecast.

Confidence should fall when local geometry is old, a station is missing, the model and observation disagree, or the spot has too few verified outcomes. A sparse forecast can still be useful. It should not present itself as certain.

What changes at the beach

Wijk and IJmuiden can share offshore swell while the harbour moles change shelter and breaking height. Domburg needs its own wave transformation around a curved, tide-affected coast. Oostende, Mariakerke, Nieuwpoort, and De Panne need different Belgian profiles despite sharing the same regional sea state. De Koog and Paal 17 should not be trained as one Texel break.

This is the practical standard: forecast the named break, preserve the source and model version, and test each tuning change on weather events it has not seen before.

When the local evidence is thin, say so.

Method

We reviewed the previous year of public Surfweer forecasts and their attached comments, then compared the recurring local patterns with every current Goedegolven regional guide. The practitioner findings were tested against work from Rijkswaterstaat, KNMI, Deltares, TU Delft, MDK, Flanders Hydraulics, VLIZ, and peer-reviewed coastal research.

The review covered 18 Surfweer forecasts, 555 public comments, and 17 Goedegolven regional guides. The counts describe the material screened; they do not turn comments into instrument data or establish a forecast hit rate.

Primary sources

• Rijkswaterstaat / Deltares JARKUS coastal profiles
• Deltares NWECS wave modelling
• Deltares D-Waves nearshore processes
• Flanders Hydraulics: The Wave Climate in the Belgian Coastal Zone
• Belgian field measurements and beach morphology
• MDK Hydrometeo and Meetnet Vlaamse Banken
• Deltares North Sea swell-propagation study
• TU Delft Scheveningen boundary-rip study