Validation · classic wave-lab benchmarks · published reports

We test the model where the answer is known — before trusting it with your wave.

Four classic wave-laboratory benchmarks, run in our production toolchain and checked against published measurements: wave transformation over a submerged bar, breaking on a slope, wave focusing across a model basin — and piston wave generation. The classics are closed; the full reports, with data and limitations, are published as citable DOI reports.

A render proves the software runs. Validation proves the numbers mean something.

Any CFD setup produces convincing pictures of water. The engineering question is whether the same setup reproduces waves that were physically measured. So we run the classic wave-laboratory benchmarks — the ones wave models have been tested against for decades — and publish every comparison in full, as citable reports. This page is the short version: what was tested, and how it came out.

Beji–Battjes: a wave crossing a submerged bar.

The flume classic (Beji & Battjes, 1993): a regular wave crosses a submerged bar and gets reshaped — exactly what a surf-pool bottom does to a wave. The computation is checked against laboratory measurements at all 10 wave gauges along the flume: mean surface error 2.7 mm on waves 2–4 cm high. Full report, with the data and the limitations: DOI 10.5281/zenodo.21776521.

Computed free-surface time series overlaid on measured data at 10 wave gauges, Beji-Battjes case A
Ten gauges, calc vs measured Computed surface elevation (line) on top of the laboratory measurements (points), from the flat approach to the lee side of the bar. Per-gauge error 1.5–4.1 mm.
Harmonic amplitudes along the flume: measured vs computed first, second and third harmonics
Harmonic transfer First, second and third harmonic amplitudes along the flume — the energy hand-off from the primary wave into higher harmonics over the bar, measured (dashed) vs computed (solid).

Ting & Kirby: a spilling breaker on a 1:35 slope.

The reference surf-zone experiment (Ting & Kirby, 1994): a regular wave runs up a 1:35 slope and breaks. For a surf pool this is the regime that matters — the breaking wave. The computation lands where it matters for design: where breaking starts, how the mean water level behaves, and the return current under the wave, at the measured strength and shape. The full write-up of both flume cases is in the same report: DOI 10.5281/zenodo.21776521.

Wave crest and trough envelope and mean water level: computed vs Ting and Kirby measurements
Surface envelope & setup Crest/trough envelope and mean-water-level setup through the breaking zone — computed vs the measured points.
Undertow velocity profiles at four cross-sections: computed vs measured
Undertow profiles The offshore return flow at four sections across the surf zone. The turbulence closure (red) reproduces the measured shape and magnitude better than the closure-free run (blue).

Berkhoff shoal: refraction and diffraction across a whole model basin.

The third classic (Berkhoff, Booy & Radder, 1982): a 20×22 m model basin with an elliptic shoal focuses the incoming wave into a narrow beam — a check of what the entire bathymetry does to the wave field at once. It is also how we work on real projects: the basin as a whole is computed by a wave model, the wave-forming zone by CFD. Agreement with the experiment across the four measured sections: r = 0.97. Report: DOI 10.5281/zenodo.21804625.

Normalized wave height along four measured sections of the Berkhoff shoal basin: two wave models vs digitized experimental points
Four measured sections, calc vs experiment Normalized wave height along the four published measurement sections — two independent wave models (lines) vs the laboratory points. Green–Naghdi model: r = 0.97 against the lab points.
Computed wave-height amplification map over the elliptic shoal: focal beam up to 2.3 times the incident height with shadow bands
Focal beam over the shoal Computed wave-height amplification across the basin: the shoal focuses the incident wave into a beam of up to ~2.3× H₀ with interference minima on both flanks. Dashed lines — the measured sections.

Ursell / Madsen: the piston wavemaker.

The fourth benchmark closes the start of the chain — the wavemaker itself. A piston paddle in a numerical flume against classical wavemaker theory and Madsen's (1970) laboratory experiment: wave height at the paddle — 100.5% of theory; and the first waves after start-up run 1.22× the steady height, exactly as in the lab (1.22×). Report: DOI 10.5281/zenodo.21858419.

The classics, closed: four regimes at a glance.

Four benchmarks — four physical regimes of a surf pool: what the bottom does to the wave, the breaking itself, the wave field of a whole basin, and the wavemaker. The details live in the published reports below.

Beji–Battjes 1993submerged bar · benchmark 1 Ting & Kirby 1994spilling breaker · benchmark 2 Berkhoff 1982elliptic shoal · benchmark 3 Ursell / Madsen 1970piston wavemaker · benchmark 4
Physical regime Wave transformation over the bottom, before breaking Breaking on a slope — the surf zone itself Refraction, diffraction and focusing across a whole basin Wave generation: from paddle stroke to wave
Compared against Surface elevation at 10 wave gauges (Delft Hydraulics) Wave envelope, mean water level, return-current profiles Wave height along the 4 published measurement sections Classical wavemaker theory + Madsen’s (1970) start-up experiment
Result Mean surface error 2.7 mm on 2–4 cm waves Breaking point, mean level and return current land on the measured picture r = 0.97 against the experiment 100.5% of theory at the paddle; start-up overshoot 1.22× = lab 1.22×

How these validations are run.

01

Open solvers

OpenFOAM (interFoam, volume-of-fluid) for the wave-forming zone; Basilisk (Green–Naghdi, layered non-hydrostatic) for basin-scale propagation. Nothing proprietary between you and the numbers — anyone can re-run the cases.

02

Published data, traced

Measured series and points come from published sources with provenance recorded — flume geometry and parameters cross-checked against the original papers, digitized points checked against published tables.

03

No tuning to fit

Fixed averaging windows chosen for stationarity, per-wave robust statistics that reject spray artifacts, one global phase alignment — and no parameter adjusted to chase the measurement.

04

Adversarial review

The transformation case passed an independent audit that tried to break the data provenance, the setup and the analysis. Corrections it found are folded in; the verdict is on record.

Error bars you can plan concrete around.

When we calculate a surf-pool wave, every claim rides on the same toolchain shown here. The validated harmonic transfer is what makes bottom-profile predictions trustworthy; the quantified surf-zone limits are what keeps breaking-wave claims honest — we know which outputs to state with confidence and which to bracket with margins or a physical prototype.

All three validation reports are published as citable technical reports (Zenodo) — each with the setup, the data, the full comparison and the limitations: flume CFD, transformation and breaking — DOI 10.5281/zenodo.21776521 (PDF, 1 MB); basin-scale wave models, Berkhoff — DOI 10.5281/zenodo.21804625 (PDF, 0.4 MB); piston wave generation — DOI 10.5281/zenodo.21858419 (PDF, 0.5 MB).

Common questions about CFD validation.

Why validate on lab flumes instead of a real surf pool?

Because flume benchmarks are published: geometry, wave parameters and measured series are open, densely instrumented, and anyone can re-run the comparison. Measurements from operating pools are proprietary and sparse. Qualifying a solver on published benchmarks is standard coastal-engineering practice.

Why publish where the model disagrees with the measurements?

Because the error bars are the product. A model with known, quantified limits supports engineering decisions; a model that only ever shows agreement can't be interrogated. We publish where it holds, where it diverges, by how much, and why.

Which solver do you validate?

OpenFOAM (interFoam, volume-of-fluid) for the wave-forming zone, with standard wave generation and active absorption; Basilisk wave models (Green–Naghdi, layered non-hydrostatic) for basin-scale propagation. All open and checkable; what we validate is the whole practice — setup, closures, resolution, time schemes, postprocessing discipline.

Where are the full data and the limitations?

In three published, citable technical reports (Zenodo, CC BY): flume CFD — transformation and breaking — DOI 10.5281/zenodo.21776521; basin-scale wave models — DOI 10.5281/zenodo.21804625; piston wave generation — DOI 10.5281/zenodo.21858419. Each contains the setup, the data, the full comparison and the limitations. This page is the short version.

Related engineering topics.

Want the numbers behind this page?

The full reports behind this page are published — DOI links above. Send your project and we'll tell you which parts of it fall inside the validated regimes, and which would need their own checks.