Skip to content
WX Navigator

Documentation

Forecast models and environmental data

Routed wind forecasts, additive current and wave fields, depth, and model consensus.

Forecast models

Every available wind model allowed by the plan is routed. The Model picker chooses which answer you are looking at, not which one gets computed. Currents, waves, and depth are different: they modify or constrain each wind-model route rather than producing candidates of their own.

The consensus

Left on Consensus, the panel shows the track the models broadly agree on and names which model that recommendation came from. Open the picker to step through the individual models and see what each would have had you do.

This is the useful part. A single forecast gives you a course to steer. Several give you a course to steer and a measure of how much to trust it. When they part company, you know before the start that the race has a real decision in it, and you can plan a hedge.

Departure scans are solved under the consensus model, and the panel says so.

Wind and pressure

NOAA GFS — global, 0.25°, four cycles a day (00/06/12/18Z), hourly steps to 120 hours.

NOAA HRRR — US coasts and the Great Lakes, 3 km, hourly. It resolves sea breeze, lake effects and terrain channelling that a 0.25° global model averages away. At All-out it is decoded near its native 3 km grid with 15-minute steps rather than resampled to 0.25°.

NOAA NAM (CONUS nest) — the same US 3 km footprint as HRRR, from a different model core, with hourly steps to 60 hours from the 00/06/12/18Z cycles (published about two hours after each). Past 60 hours its candidate continues under GFS, exactly as HRRR's does past 48. NAM is scheduled for retirement in favour of NOAA's RRFS; when that happens its candidate simply stops appearing — nothing else depends on it.

ECMWF IFS — global, 0.25°, from the open data release. Often the stronger model in the 48–120 hour range, and the standard second opinion against GFS.

ECMWF AIFS — ECMWF's AI forecast system, run alongside IFS. It behaves differently from a physics model, so it functions as a third opinion rather than a second copy of IFS.

Availability depends on the plan, model coverage, requested horizon, and whether a complete cycle has published. A regional model that does not cover the course simply does not appear.

Currents

NOAA RTOFS supplies ocean surface currents for set and drift. It has no Great Lakes coverage.

NOAA operational forecast systems supply hourly tidal currents in supported US coastal waters and the Great Lakes. The product selects the regional system that covers the course. Both current sources are additive: they are sampled alongside a wind model, not routed as separate weather answers.

Waves

NOAA GFS-Wave supplies sea state where the paid artifact set includes waves. The route reports the waves it encounters. Wave avoidance and a maximum-wave constraint can steer the search toward calmer water, but waves do not directly reduce polar boat speed; the polar is assumed to price the boat's performance in the chosen conditions.

Depth

Depth comes from one tiled archive built from three sources:

  • GEBCO 2024, 15 arc-seconds, globally. Its shallowest band starts at 200 m, so it keeps long ocean routes off seamounts and says nothing about anywhere you could run aground.
  • NOAA NCEI Great Lakes grids, about 3 arc-seconds. GEBCO models the lakes as land elevation, so all lake depth comes from here.
  • NOAA ENC Direct Approach, Harbour and Berthing charts, for inshore contours.

The draft on your certificate is used against it. This is a planning constraint and not a navigational one — see the terms.

Reference height

Wind is reported at 10 m TWS, shown in the readout at the top of the screen. That is the height polars are quoted against, so the numbers the router uses and the numbers on your certificate mean the same thing.

Ensemble spread

Some deployments expose an experimental GEFS spread band under Route stability. It samples the 31-member ensemble's wind disagreement along an already computed deterministic route. It does not compute an ensemble route, choose a hedge, or route on an ensemble mean.

The route-clustering work used in research and the admin hindcast harness is not yet a customer feature. If it moves into the product, clustered route families—not an averaged weather field—are the intended representation.