Improving wave forecasts

It starts with advanced wave models: computer simulations that predict wave behaviour based on physics and weather forecasts. The more accurate these models are, the more reliable the predictions.

If we understand wave behaviour better, we can not only reduce risks but also seize opportunities.

Elias de Korte, Oceanographer specialising in waves and coastal dynamics at Deltares.

Buoys and satellites

“Wave models have reached a reasonable level of accuracy for some of the applications,” explains Elias, a wave researcher at Deltares. “But there are still many uncertainties, for example in wind and wave growth.” To validate these models with reliable data, Deltares uses a network of measurement buoys in the North Sea. These buoys are fixed in position and continuously collect data on wave height, wind, and currents.

This buoy data forms a crucial basis for validating and calibrating our models, but it only provides a limited picture: measurements at specific points, with no spatial coverage. That’s why we also look at complementary sources, such as satellite observations, to improve the overall picture and enhance safety at sea.

Satellite altimetry adds real value: satellites measure wave heights from space across large areas. Because these data are publicly available, researchers can integrate them into their models, improving both short-term forecasts and long-term hindcasts (retrospective analyses of sea behaviour).

What is data assimilation?

Measuring alone isn’t enough for reliable wave forecasts. You need to process the data intelligently. That’s where data assimilation comes in: a technique that combines real-time observations—such as buoy or satellite data—with model predictions. This approach uses both physics and measurements to produce realistic forecasts. At the heart of this system is the Ensemble Kalman Filter (EnKF).

How does the Ensemble Kalman Filter work?

The EnKF improves model predictions by blending them with actual measurements, applying mathematical rules to account for uncertainties in both the model and the satellite data. It works with models such as SWAN (for waves) and Delft3D (for water movement).

Elias explains: “We use wind forecasts to drive our wave model—average wind speeds at 10 metres above the water surface. But real wind fields can differ significantly, with gusts or sudden directional shifts. Wave growth is very sensitive to even small errors in wind forecasts.”

An ensemble of predictions

“To capture these uncertainties, we run the wave model 64 times with slight variations in input data—each run with a slightly different wind speed or direction. This creates an ensemble of predictions, each showing a slightly different scenario for wave development.” Why 64 runs? “It’s the sweet spot between detail and speed. Enough variation to map uncertainties, while remaining efficient on the supercomputer we use.”

“When new satellite measurements arrive, we compare them with our predictions. Then we adjust all forecasts slightly, based on uncertainties in both the model and the satellite data. With this improved set of predictions, we calculate the next time step.”

High-Speed technology

Developing this system was no small feat. “The wave model is computationally heavy,” says Elias. “Running 64 parallel simulations is a challenge—even on the national supercomputer Snellius, managed by SURF.” Modelling uncertainties also requires a careful balance between simplification and complexity: “You need to know where you can simplify, and where physics must be included.”

Impact at sea and on the coast

The applications are wide-ranging and socially relevant. Better forecasts help shipping and the offshore industries to operate more safely and efficiently. They also support storm surge barrier operations, as waves can locally raise water levels—critical for coastal safety. Improved hindcasts provide valuable insights for climate research and can serve as training data for machine learning models, which in turn can be deployed operationally.

But what does this mean for society at large? Elias explains how these insights contribute to sustainable development, policy, and innovation:
“If we understand wave behaviour better, we can not only reduce risks but also seize opportunities—optimising shipping routes, planning offshore wind farms, or protecting fragile ecosystems. This knowledge helps policymakers, businesses, and researchers make informed decisions that balance economic and ecological interests.”

Looking ahead

Our ambitions go beyond the North Sea. Elias: “We aim to deliver more accurate forecasts for Rijkswaterstaat and expand to regions with high flood risk and few measurement buoys.” The next step is to test the system over longer periods and validate it with independent buoy data. We’re also exploring ways to integrate other uncertainties beyond wind.

By smartly combining data and models, we make the North Sea—and potentially other coastal areas—more predictable, contributing to a future-proof marine environment where economy, ecology, and energy production are in balance.

This project builds on earlier Deltares work by colleagues such as Sofia Caires and Martin Verlaan, who focused on buoy measurements. Elias uses existing tools like the SWAN wave model and the OpenDA data assimilation platform—both part of the Deltares portfolio. The project is not only an innovation but also strengthens our knowledge base.

Elias doesn’t just study waves through models and data – he also experiences their natural power. On his surfboard, he feels what he explores in theory: how waves behave. It’s where science meets the sea.