Live Forecasts

Jump to your personalised 14-day planner

Save any location (including your own pins), set swell/wind/tide/weather preferences, and see the best sessions now and next across all your spots.

Go to Forecasts
Back to Learning Center

The Big Two: GFS vs. ECMWF

6 min read

The Battle of the Supercomputers

Two giant computers fight for supremacy every day. One is in the USA (GFS). The other is in Europe (ECMWF). Both attempt to predict the atmosphere's behavior using physics equations and billions of data points.

They often disagree—sometimes dramatically. When GFS shows 12 knots and ECMWF shows 18 knots for the same day, which one should you believe? The answer matters for your session.

The Physics

Both models solve the Navier-Stokes equations for atmospheric fluid dynamics. The difference lies in grid resolution, data assimilation quality, and parametrization schemes (how small-scale processes are approximated). ECMWF uses a finer grid and more sophisticated assimilation, giving it an edge in medium-range accuracy.

Skill score (accuracy) degrades exponentially with forecast lead time—ECMWF maintains higher skill longer

GFS (The American)

The Global Forecast System is run by NOAA (National Oceanic and Atmospheric Administration). It is free and open to the public. Most free weather apps and websites use this data because it costs them nothing.

  • Update Cycle: It updates 4 times a day (00, 06, 12, 18 UTC). This rapid refresh is useful for short-term nowcasting.
  • Resolution: It divides the world into 27km squares (0.25° grid). This is "coarse." It misses small local features like headlands, river valleys, and island gaps.
  • Forecast Range: Runs out to 16 days, though accuracy drops significantly after day 7.
  • Strengths: Free, frequent updates, good for rapidly evolving weather systems.
  • Weaknesses: Lower resolution, less sophisticated data assimilation, struggles with complex terrain.

ECMWF (The European)

The European Centre for Medium-Range Weather Forecasts is based in Reading, UK. It is a commercial product funded by 35 European countries. Access costs money, which is why premium weather services use it.

  • Update Cycle: It updates 2 times a day (00 and 12 UTC). Less frequent, but each run is deeper and more refined.
  • Resolution: It divides the world into 9km squares (0.1° grid). This is "fine." It sees the world in high definition, capturing local wind acceleration zones and terrain effects.
  • Forecast Range: Runs out to 10 days with high accuracy, 15 days extended.
  • Strengths: Superior accuracy (especially days 3-7), better data assimilation, excellent for medium-range planning.
  • Weaknesses: Expensive, less frequent updates, not always available on free apps.

Resolution Comparison

GFS (NOAA)

27km

Low Resolution—misses small terrain features

ECMWF (Europe)

9km

High Resolution—captures local wind patterns

Why ECMWF Wins

Statistically, ECMWF is more accurate than GFS for medium-range forecasts (3-7 days out). Independent verification studies consistently show 15-20% better skill scores for ECMWF in the 3-5 day window.

The European model achieves this through:

  • Better data: ECMWF spends more money buying private satellite and aircraft observations
  • Superior assimilation: 4D-Var algorithm with tighter temporal windows and better bias correction
  • Finer grid: 9km resolution captures local wind acceleration zones (Venturi effects, land-sea interactions)
  • Advanced physics: More sophisticated cloud parametrization and boundary layer schemes

If GFS says 12 knots and ECMWF says 18 knots, trust the European. History favors it.

When GFS Is Better

GFS isn't useless. It has advantages in specific scenarios:

  • Nowcasting (0-6 hours): GFS updates 4x daily, so it's fresher for immediate forecasts
  • Rapidly evolving systems: If a storm is intensifying quickly, GFS captures it sooner
  • Free access: If you can't afford premium data, GFS is better than nothing
  • Ensemble availability: GFS ensemble (GEFS) is free and widely accessible for uncertainty analysis

For next-day forecasts (12-24 hours), the gap between GFS and ECMWF is smaller. Both are reasonably accurate. The real divergence happens in the 3-7 day range.

Accuracy Timeline

0-2 Days

Both Good

GFS and ECMWF are similarly accurate

3-7 Days

ECMWF Wins

15-20% better accuracy for medium-range

7+ Days

Both Degrade

Forecast skill drops exponentially—use trends, not specifics

How to Use Both

Smart riders compare both models. Here's a practical workflow:

  • Check GFS first: It's free and gives you a rough idea of upcoming weather
  • Verify with ECMWF: If planning a session 3+ days out, check ECMWF for confirmation
  • Look for agreement: If both models show the same pattern, confidence is high
  • Watch for divergence: If they disagree by >5 knots or opposite wind directions, the forecast is uncertain—check closer to the day

Practical Tips

Planning trips: Use ECMWF for 3-7 day forecasts—it's more reliable for booking decisions

Same-day sessions: GFS is fine—both models are equally accurate at 0-24 hours

Check update times: ECMWF runs at 00:00 and 12:00 UTC—forecasts are freshest 2-4 hours after

When they disagree: Trust ECMWF for wind speed, GFS for rapid changes—or wait for the next model run

Other Models to Know

GFS and ECMWF aren't the only options. Regional models often outperform both for local forecasts:

  • ICON (Germany): 3-hour updates, excellent for Europe, free access
  • NAM (USA): High-resolution regional model for North America
  • UKMO (UK): UK Met Office model, comparable accuracy to ECMWF
  • Canadian (CMC): 4x daily updates, good for North America

If you ride in areas covered by these regional models, use them—they often beat global models for short-term, localized forecasts.

Summary

If you pay for a premium weather app, you are usually paying for the ECMWF data. It is the gold standard for precision, especially 3-7 days out. GFS is free and updates more frequently, making it useful for same-day decisions and rapidly changing weather. For wing foiling trip planning, trust ECMWF. For quick same-day checks, GFS is fine. When they agree, confidence is high. When they disagree, the forecast is uncertain—prepare for variability.

AI-generated content for research only. Verify with real experts, certified instructors, and official sources.

Forecasts made for you

Save any spot (including your own pins), set swell/wind/tide/weather preferences, and scan 14 days to see where to ride now and next.

View Forecasts