Rogue — forecast extreme wave events

Why this exists

The problem no statistical model can solve.

$2B+ / year in shipping damage, ~100 cruise-ship deaths / yr, offshore-platform damage, submarine-cable outages and wind-turbine failure — all from a few minutes of rogue-wave energy.

Today's tools are WAVEWATCH III / SWAN (spectral, no individual waves), CFD (accurate but far too slow for real time) and Rayleigh statistics (systematically underestimate extremes).

Rogue sits on the 2026 Breakthrough-Prize mathematics of finite-time blow-up in supercritical defocusing NLS (Merle–Raphael–Rodnianski–Szeftel) to flag the precise moment a field is about to concentrate — directly, in real time.

How to use it

No install. No browser of heavy setup. Just press go.

Pick a window — the interactive education playground or the real-data production forecast. Or run the same engine from the CLI and scripts.
Set the state — choose a scenario and resolution, or type a real sea state (Hs, Tp, peak γ) for a physical forecast.
Run & watch — the wasm solver integrates the field with a symplectic scheme; cast-colored crest factors and a risk label appear.
Interpret — < 2.2 σ is everyday, ≥ 2.2 σ is rogue territory, flags bless blow-up.

The parameters

What the knobs mean, and the equations behind them.

Parameter	Meaning	Where
`p`	Nonlinear exponent in `i∂tψ = −(β₂/2)∇²ψ + γ\|ψ\|^p−1ψ`. Supercritical (p > 5) admits finite-time blow-up.	edu · blow-up
`β₂`	Dispersion coefficient (m²/s in production, tuned by sea state).	solver
`γ`	Nonlinearity — JONSWAP spectrum peak in production, constant in edu.	prod + edu
`Hs`	Significant wave height (m). Drives amplitude `A = Hs/4`.	production
`Tp`	Peak period (s). Sets deep-water `ω₀ = 2π/Tp` and wave number `k₀ = ω₀²/g`.	production
`resolution`	Grid size (256→1024). More points, finer crests, slower run.	education
`speed`	Time integrator speed multiplier for the education view.	education
`seed`	Randomness for the random-sea field, so every run is reproducible.	both

Command line

Same engine, no browser — for scripts, servers and automation.

Command	What it does
`cargo install --path . --root ~/.local`	Build & install the `rogue` binary to `~/.local/bin`.
`rogue predict data/sample-buoy.csv`	Physical sea-state forecast from a CSV/TSV of `t, hs, tp` (optionally `gamma`).
`rogue predict sea.csv --json`	Same forecast as machine-readable JSON (for hooks).
`rogue akhmediev` `rogue peregrine` `rogue ocean` `rogue blowup`	Scenario solvers with conservation + blow-up + rogue diagnostics.
`./dev p sea.csv`	Repo alias for the production forecast path.

Runs are seeded and reproducible; the report prints mass / momentum / energy conservation and flagged rogue events.

Connect your instruments

Live feed from a buoy, bridge, coastal station or vessel — the solver runs where the data is.

1 · Pipe in your feed

Any instrument that outputs t, hs, tp works — wave buoys (Datawell, Spotter, TRIAXYS), shipboard X-band radars, coastal ADCP stations, drifting Waveriders, or a helmsman typing observations. CSV over serial, NMEA log, HTTP POST, or a plain file — Rogue reads all these.

2 · Forecast continuously

Run rogue predict /dev/stdin --json in a watch loop or systemd timer. Every new row is a physical, dimensionality-correct forecast. JSON output drops straight into an alarm, a bridge display, or a routing-aid panel — no middleware.

3 · Run on the web, anywhere

No install on the vessel — the same WASM forecast engine lives at /production. A captain, a tug operator, or a coast-guard watch can enter a sea state in a browser and get a forecast in under a second.

4 · Reproducible audit trail

Deterministic and seeded — replay a 12-hour AIS / S-AIS log and Rogue produces identical alerts. Critical for post-voyage review, insurance claims, port-state inspections, and regulatory compliance (SOLAS, MARPOL).

Forecast the wave
that appears from nowhere.

Choose your window