Quick Facts
| GRIB Format | Binary gridded binary — mathematical model output, not an image |
| GFS Resolution | 0.25° grid (~27 km) — free, global, updated every 6 hours |
| ECMWF Resolution | 0.1° for marine products — premium, more accurate days 5–10 |
| Update Cycle | 00Z, 06Z, 12Z, 18Z — always fetch the freshest run |
| Key Variables | Wind U/V, significant wave height, swell period/direction, MSLP |
| Satellite File Size | Trim to your routing area — a Pacific crossing GRIB can exceed 5 MB |
Most offshore sailors have downloaded a GRIB file. Fewer understand what they are actually looking at. The forecast display in PredictWind or LuckGrib shows coloured wind arrows and wave heights, which feels intuitive — but underneath that visualization is something quite different from a weather map drawn by a meteorologist. Understanding what the model actually produces, where it is reliable, and where it lies is the foundation of competent offshore weather routing.
What GRIB Actually Is
GRIB stands for GRIdded Binary — a data format developed by the World Meteorological Organization to store and distribute the output of numerical weather prediction models. It is not a picture. It is not a forecaster's interpretation. It is a dense binary file encoding mathematical values for dozens of atmospheric variables at regular grid points across the globe, projected forward in time in discrete steps — typically 3 or 6 hours.
When a weather model runs, it solves an enormous system of differential equations describing atmospheric physics: fluid dynamics, thermodynamics, radiative transfer, moisture physics, ocean-surface interaction. The output is a four-dimensional dataset — latitude, longitude, altitude, and time — containing predicted values for temperature, pressure, humidity, wind velocity components, and dozens of derived variables. GRIB is simply the standardized format for packaging and transmitting that output.
The implication for sailors is important: what you see in your routing software is a visualization of mathematical model output, filtered through whatever variables you asked for, displayed at the resolution the model provides. The forecast has no knowledge of your specific anchorage, the local sea breeze that builds off the cape every afternoon, or the confused sea state that forms when a 3-meter northwest swell meets a 2-meter local wind chop from the southwest. It knows what the equations predict for that grid square. That is all.
The Model Hierarchy: GFS vs. ECMWF
Two models dominate offshore sailing weather: the American GFS (Global Forecast System, run by NOAA) and the European ECMWF (European Centre for Medium-Range Weather Forecasts). Both are global models. Both update every six hours. Both produce output sailors can access. They are not equal.
GFS is free. Anyone can download GFS GRIB data from NOAA's servers, and the major sailing weather apps all include it. At 0.25-degree resolution (approximately 27 kilometers per grid square), GFS is more than adequate for open-ocean routing. It is updated four times daily and provides reliable forecasts to roughly 72–96 hours. Beyond day 4, GFS forecast accuracy degrades more rapidly than ECMWF, particularly for synoptic-scale feature position (where the high is, where the fronts are) and intensity.
ECMWF — colloquially the "Euro model" — is widely regarded as the most accurate global model for medium-range forecasting, particularly in the day 5–10 range. ECMWF consistently outperforms GFS in independent verification studies, particularly for storm track prediction and frontal timing. The catch: ECMWF data is not free. Accessing ECMWF GRIB data requires a paid subscription, either directly or through applications like PredictWind, LuckGrib (Mac/iOS), or Expedition that license the data. For serious offshore passages where day 7 weather matters — a transatlantic, a Southern Ocean run, a Pacific crossing — the cost of ECMWF access is one of the better investments in offshore safety available.
The practical approach: use GFS for short passages and initial planning; add ECMWF for any passage where the forecast beyond 96 hours will influence a go/no-go decision or a mid-passage routing adjustment.
Ensemble vs. Deterministic Runs
Most sailors download a deterministic GRIB — a single "best-guess" model run that produces one forecast for each point in time and space. This is useful and familiar. It is also potentially misleading about forecast certainty.
Ensemble forecasting addresses this by running the same model many times with slightly different initial conditions. The GFS Global Ensemble Forecast System (GEFS) runs 31 members; ECMWF's ensemble (ENS) runs 51 members. The spread of those ensemble members tells you something the deterministic run cannot: how confident the model is in its own forecast.
When ensemble members cluster tightly — all showing a high centred near the same position, all showing the cold front passing within a few hours of each other — the forecast is reliable. When they fan out dramatically — some members showing the low tracking north, others tracking south, some deepening it aggressively, others keeping it weak — the deterministic "best guess" may be meaningless. Any of those scenarios might verify. The tight-clustered reliable forecast and the widely spread uncertain forecast look identical in a single deterministic GRIB download.
PredictWind's departure planner visualizes ensemble spread directly. LuckGrib provides ensemble GRIB access. For any passage where weather uncertainty could mean the difference between a comfortable passage and a serious situation, ensemble data is worth the additional download and the cognitive overhead of interpreting probability rather than certainty.
Variables That Matter for Sailors
A full GRIB file contains hundreds of variables. Sailors need a focused subset. Wind is stored as U and V components — eastward and northward velocity in metres per second — which your software converts to the speed and direction display you recognize. For ocean passages, always verify that wind speeds are displaying at 10 metres above the surface (10m wind), the standard for marine forecasting, not at a higher atmospheric level.
Wave data requires separate GRIB products. Significant wave height (SWH) is the average height of the highest one-third of waves — a statistical measure, not the height of the biggest wave you will encounter. Peak wave period tells you whether the energy is wind-sea (short period, choppy) or swell (long period, organized). Swell direction indicates where the energy is coming from, which matters when it conflicts with your heading and creates beam-on seas. The combination of swell height, period, and direction relative to your boat and apparent wind creates the sea state you actually experience — the GRIB alone cannot fully predict comfort.
Mean sea level pressure (MSLP) and its gradient are foundational for understanding the synoptic situation — the large-scale pressure pattern that drives the winds. Isobar spacing in the GRIB pressure field gives you the geostrophic wind estimate independent of the wind forecast itself; tightly packed isobars mean strong pressure gradient and strong winds regardless of what the wind field shows.
The 6-Hour Update Cycle and Freshness
Weather models initialize four times daily at 00Z, 06Z, 12Z, and 18Z (UTC). Each initialization incorporates the latest observational data from radiosondes, satellite retrievals, aircraft reports, ocean buoys, and surface stations. The model run completes approximately 3–5 hours after initialization time, which means the 00Z run is available around 03:00–05:00 UTC.
Forecast quality degrades with age. A GRIB downloaded 18 hours ago was initialized from atmospheric observations that are now 18 hours old — the model has had two additional updates since then, both with fresher data. On a passage, downloading the latest model run every 12 hours (or at every 00Z and 12Z update) is basic practice. Day 1 of an old model run may be less accurate than day 3 of a current run for rapidly evolving situations.
GRIB File Size and Satellite Comms
An offshore passage GRIB covering a large ocean area, multiple variables, 10-day forecast horizon, and 0.25-degree resolution can easily reach 5–10 MB. Over Iridium Go at 2.4 kbps, that is a 5-10 hour download — impractical. Saildocs email requests allow you to specify a geographic bounding box, a variable list, a resolution, and a forecast length; a well-configured Saildocs request for an Atlantic crossing corridor might return a 200–400 KB file covering everything you need.
The discipline of GRIB file management is a proxy for offshore weather discipline generally: know exactly what you need, request only that, and fetch it on schedule. Apps like PredictWind Offshore and Expedition have satellite-optimized GRIB download features built in. LuckGrib's offline GRIB caching is particularly well-suited to Mac-based chartplotter setups. Whatever tool you use, the file-size problem forces you to think carefully about what forecast information actually matters for your passage — which is useful thinking to do in any case.
Continue Reading
Learn how to turn GRIB data into departure decisions: Finding and Using Weather Windows → | Routing Software Compared →