GRIB Files: Offshore Weather Data Explained

Weather Routing · Offshore Tools

Sailors increasingly talk about "downloading GRIB files" without always being clear on what a GRIB file actually is, what data it contains, or how to interpret what a viewer is showing them. The format is compact and efficient — designed for transmitting large amounts of weather model data in minimal bandwidth — which is exactly why it became the standard for offshore weather delivery. Understanding what's inside it makes you a better consumer of the data.

What a GRIB File Is

GRIB stands for GRIdded Binary. It is a file format standardized by the World Meteorological Organization for storing and exchanging weather model data. The key word is "gridded": a GRIB file divides the world (or a region of it) into a regular grid of points, and at each grid point stores numerical values for specific meteorological variables at specific times.

What those variables are depends on what you requested when you downloaded the file. A typical offshore sailing GRIB might contain: wind speed and direction at 10 meters above surface, significant wave height, wave period and direction, mean sea level pressure, and precipitation. More detailed files can include upper-atmosphere wind data (useful for understanding the steering flow for storm systems), sea surface temperature, and cloud cover.

What a GRIB file does not contain: observations. Everything in a GRIB file is model output — a numerical simulation of the atmosphere's future state based on current observations fed into a mathematical model. The quality of the forecast is only as good as the model's ability to simulate the real atmosphere over the time period in question. For 24 hours, that's quite good. For 5 days, it's reasonable with caveats. For 7–10 days, treat it as directional guidance only.

Understanding Model Run Times

Weather models run four times per day, initialized at 00Z, 06Z, 12Z, and 18Z — Z meaning UTC (Zulu time). The 00Z run begins at midnight UTC; the 12Z run begins at noon UTC. Each run ingests current observations from weather stations, radiosondes, satellites, aircraft, and ocean buoys worldwide, then calculates the atmospheric state forward in time.

The newest model run always has the freshest data. A 12Z run initialized at noon UTC is working from observations collected that morning; a 00Z run from the previous midnight is 12 hours stale. This matters because it affects the model's starting state — if a low-pressure system developed or intensified since the last run's initialization time, the previous run doesn't know about it.

The practical implication: when you're downloading GRIB files for departure planning, use the most recent model run available. Note that there's a delay between when a model run starts and when the data becomes publicly available — GFS typically takes 3–4 hours to process and post. So the "12Z run" for GFS becomes available around 15:00–16:00 UTC. Your routing app or GRIB downloader will show you which run you're using; pay attention to it.

When two consecutive model runs tell significantly different stories about conditions in your passage window, that divergence is important information. A single model run showing a gale is worrying. If the next run removes it, uncertainty is high. If the next run makes it worse, you should believe it.

Resolution and What It Means

GFS runs at 0.25° resolution, which works out to approximately 28 kilometers between grid points at mid-latitudes. ECMWF runs at roughly 9 kilometers. This gap matters in specific situations.

In open ocean, far from land, the difference between 28km and 9km resolution is relatively small. A low-pressure system 500 miles offshore is large enough that both models represent it at adequate resolution for routing purposes. The difference becomes significant near complex coastal terrain — mountainous coastlines where thermal effects and orographic lifting create local wind conditions that a 28km grid simply cannot resolve. California's Channel Islands, the Pacific Northwest, Norwegian fjords, the Greek islands — anywhere the topography is complex and scales smaller than 28km matters, GFS will smooth over detail that ECMWF captures better.

Beyond 5 days, resolution is increasingly irrelevant — the atmospheric evolution becomes inherently uncertain at scales far larger than either model's grid. A 10-day GFS forecast is not more useful than a 10-day ECMWF forecast because of resolution differences; both are working at the limits of atmospheric predictability.

How to Download GRIB Files Offshore

Satellite: Iridium Go! or Similar

The most reliable method for offshore passages is satellite-based GRIB download via an Iridium Go! or similar satellite modem. Iridium's network provides truly global coverage — including the polar regions where most satellite phone services fail. The PredictWind Offshore app is designed around this workflow: it connects to the satellite modem via your device's wifi, sends a request for a specifically bounded GRIB file, and downloads the compressed result. A typical 5-day offshore GRIB for a 1,000-mile passage corridor might be 50–150 kilobytes — a few minutes of satellite airtime.

The cost is real: Iridium Go! units run $700–1,000, monthly satellite plans add $50–150, and PredictWind Offshore adds another $30–40/month. For a transatlantic passage or a Pacific crossing, these are reasonable costs relative to the stakes. For coastal cruising, the system is overkill.

SSB Radio and Winlink

Single-sideband (SSB) radio remains a viable offshore weather option. The traditional SSB weather workflow involves receiving NOAA weatherfax broadcasts — analog fax images of synoptic charts and wave analysis — directly on your radio receiver. The images are low-resolution by modern standards but fully functional for understanding the large-scale pattern. NOAA's broadcast schedules and frequencies are available in the NWS Marine Prediction Center's documentation.

Winlink extends SSB capability considerably. Winlink is an amateur radio email system that allows licensed ham operators to send and receive email via SSB radio. Using SailMail (a maritime version of the same infrastructure), you can request and receive GRIB files via email — the file is compressed and sent in binary through the radio link. Speed is low (typically 300–600 baud), so file size discipline matters: request only the variables and area you need. A well-constructed Winlink GRIB request for a transatlantic passage might take 5–10 minutes to download.

SSB radio requires an investment in equipment (marine SSB transceiver, antenna tuner, good ground plane) and, for Winlink, an amateur radio license. But it is a system with no recurring data costs once installed, which makes it attractive for extended offshore passages and liveaboards.

Marina Wifi Before Departure

The simplest approach: download everything you need before you leave the dock. Pull the most recent model run for your passage corridor. Download 5-day GFS wind and wave files, ECMWF if your app provides it, and save them locally on your device. Most GRIB viewers can display saved files without a connection.

The limitation is obvious: you're working from data that's aging from the moment you leave. For passages of 24–36 hours, this is manageable. For longer passages, plan to update via satellite or SSB once you're out of coastal wifi range.

Reading GRIB Files in a Viewer

The three most common tools for reading GRIB files are LuckGrib (Mac/iOS, paid), OpenCPN's weather plugin (free, cross-platform), and PredictWind's app (iOS/Android, subscription).

In any of these, you're controlling a time slider that moves through the forecast period. At each time step, the map displays the values for your selected variables at every grid point. Wind is typically shown as barbs (the traditional meteorological notation) or as colored arrows. A long barb represents 10 knots, a short barb 5 knots, and a pennant (filled triangle) 50 knots — add them up for total wind speed. Wave height is shown as a color field and/or contour lines.

The most important thing to look for in a GRIB viewer is model disagreement over your passage. If you overlay GFS and ECMWF and they tell the same story, confidence is higher. If GFS shows a low staying north of your route while ECMWF clips your route with its southern edge, you have a genuine forecast uncertainty that should affect your planning.

Where Model Accuracy Degrades

Models lose accuracy in predictable ways. Beyond 5 days, the inherent unpredictability of the atmosphere dominates — even ECMWF, the best global model available, is producing probabilistic guidance rather than reliable point forecasts at that range. Use 5-day data to understand the pattern; don't bet a passage on 7-day model detail.

Mountainous coastal areas create local wind effects that model grids cannot resolve. Cape Hatteras, Point Conception, the passages between Greek islands, much of Norway — in complex coastal terrain, the model is averaging over geography and will miss local acceleration zones, shadow areas, and thermal effects that a local sailor knows from experience.

Tropical regions are particularly difficult. Convective thunderstorm development — the organized squalls that characterize the tropics — happens at scales below model resolution and is driven by instability that models characterize statistically, not precisely. In the trade winds, this usually doesn't matter much: the large-scale pattern is stable and predictable. In transitional tropical weather — developing disturbances, the edge of a tropical wave — model uncertainty can be very high even at short ranges.

Know the model's limits. It is a tool with known failure modes, not an oracle.

← Weather Routing

Frequently Asked Questions

What is a GRIB file and what data does it contain? GRIB stands for GRIdded Binary — a World Meteorological Organization file format that stores numerical weather model output at regular grid points across a region. A typical offshore sailing GRIB contains wind speed and direction at 10 meters above surface, significant wave height, wave period and direction, mean sea level pressure, and precipitation. Crucially, a GRIB file contains only model output — a numerical simulation of the atmosphere's future state — not observations. The quality of the forecast is only as good as the model's ability to simulate the real atmosphere over the time period in question: good at 24 hours, reasonable at 5 days with caveats, and directional guidance only at 7 to 10 days.

What do model run times mean and which run should sailors use? Weather models run four times per day, initialized at 00Z, 06Z, 12Z, and 18Z (UTC). The newest model run always has the freshest data — a 12Z run initialized at noon UTC is working from observations collected that morning, while a 00Z run from the previous midnight is 12 hours stale. GFS typically takes 3 to 4 hours to process and post, so the 12Z run becomes available around 15:00 to 16:00 UTC. When downloading GRIB files for departure planning, use the most recent model run available. When two consecutive runs tell significantly different stories, that divergence is important information about forecast uncertainty.

What is the practical difference between GFS and ECMWF resolution for offshore sailors? GFS runs at 0.25 degree resolution — approximately 28 kilometers between grid points at mid-latitudes. ECMWF runs at roughly 9 kilometers. In open ocean far from land, the difference is relatively small for routing purposes. The difference becomes significant near complex coastal terrain — mountainous coastlines where thermal effects and orographic lifting create local wind conditions that a 28km grid cannot resolve. California's Channel Islands, the Pacific Northwest, Norwegian fjords, the Greek islands — anywhere topography is complex and scales matter below 28km, GFS smooths over detail that ECMWF captures. Beyond 5 days, resolution is increasingly irrelevant as atmospheric predictability limits dominate both models.

What are the three methods for downloading GRIB files offshore and what are the tradeoffs? Satellite via Iridium Go! or similar is the most reliable globally — the PredictWind Offshore app connects via the satellite modem and downloads a compressed GRIB in minutes, but hardware costs $700 to $1,000 with monthly satellite plans at $50 to $150 and app subscriptions at $30 to $40/month. SSB radio with Winlink or SailMail allows licensed ham operators to receive GRIB files via email at 300 to 600 baud — no recurring data costs but requires equipment investment and an amateur radio license. Marina WiFi before departure is the simplest option but only works for short passages since data ages from the moment you leave.

What is the most important thing to look for when reading GRIB files in a viewer? The most important analysis is model disagreement over your passage. If you overlay GFS and ECMWF and they tell the same story, confidence is higher. If GFS shows a low staying north of your route while ECMWF clips the route with its southern edge, you have genuine forecast uncertainty that should affect planning. Wind is typically shown as meteorological barbs: a long barb represents 10 knots, a short barb 5 knots, and a pennant (filled triangle) 50 knots — add them up for total wind speed. The three most common GRIB viewers are LuckGrib (Mac/iOS, paid), OpenCPN's weather plugin (free), and PredictWind's app (subscription).

Where do weather models lose accuracy most predictably for offshore sailors? Models lose accuracy in predictable ways. Beyond 5 days, inherent atmospheric unpredictability dominates — even ECMWF is producing probabilistic guidance rather than reliable point forecasts. Mountainous coastal areas create local wind effects that model grids cannot resolve: Cape Hatteras, Point Conception, passages between Greek islands, and much of Norway have local acceleration zones and shadow areas that models miss. Tropical convective thunderstorm development happens at scales below model resolution — in the steady trade winds this rarely matters, but in transitional tropical weather around developing disturbances, model uncertainty can be very high even at short ranges.