Gwangju: notable disagreement by day 7
Every model's forecast from the 16 Sep 12Z run: ECMWF's learned model (AIFS), its physics model (HRES), Google DeepMind's WeatherNext 2 and the ECMWF ensemble mean, then four more learned models as NOAA runs them, each against the ensemble's 10–90% range.
Driven by temperature. 50 means the models disagree as much as the ensemble expects.
Disagreement by day
How it's measuredDay number after the run. Hover a chip for the date and the variable behind it.
2 m temperature, every model
Hover or tap the chart to read each model at that hour. The grey band is the range the ensemble puts 80% of its members in.
Four more learned models, run by NOAA
GraphCast, Aurora, FourCastNet v2 and Pangu, run by NOAA from the same ECMWF analysis, to ten days. They count towards the Disagreement Index. Same ensemble band behind them.
Does the starting point matter?
NOAA also runs each of these models from its own GFS analysis. Same model, different picture of today's atmosphere: the daily high from each start, and the difference (IFS start minus GFS start), °C.
| Model | Day 1 | Day 3 | Day 5 | Day 7 | Day 10 |
|---|---|---|---|---|---|
| GraphCast | 27° / — | 27° / — | 27° / — | 26° / — | 20° / — |
| Aurora | 26° / — | 26° / — | 26° / — | 26° / — | 25° / — |
| FourCastNet v2 | 25° / — | 25° / — | 26° / — | 24° / — | 25° / — |
| Pangu | 27° / — | 27° / — | 28° / — | 26° / — | 22° / — |
Each cell: high from the IFS start / from the GFS start. A dash means NOAA has not published that run yet; the GFS-started runs usually land first.
Day by day
Each model's high and low for the 24 hours ending at the time shown, °C. Δ is how much that model's high moved since its previous run, twelve hours earlier.
| Day | DI | AIFS | HRES | WN2 | ENS mean | ENS range | Spread of highs |
|---|---|---|---|---|---|---|---|
| Day 1to Thu 17 12Z | 37temperature | 29° 18°Δ 0 | 29° 18°Δ 0 | 27° 19°Δ 0 | 28° 18°Δ 0 | 27–29° | 2° |
| Day 2to Fri 18 12Z | 26temperature | 26° 21°Δ +2 | 24° 19°Δ -4 | 24° 20°Δ 0 | 26° 20°Δ -1 | 23–28° | 2° |
| Day 3to Sat 19 12Z | 38temperature | 29° 18°Δ 0 | 29° 18°Δ -1 | 26° 18°Δ -1 | 29° 18°Δ -1 | 28–30° | 3° |
| Day 4to Sun 20 12Z | 39temperature | 29° 19°Δ 0 | 29° 18°Δ 0 | 27° 18°Δ 0 | 29° 18°Δ 0 | 28–30° | 2° |
| Day 5to Mon 21 12Z | 50temperature | 30° 17°Δ 0 | 30° 17°Δ 0 | 27° 17°Δ 0 | 29° 17°Δ -1 | 28–30° | 3° |
| Day 6to Tue 22 12Z | 43temperature | 30° 19°Δ +1 | 30° 18°Δ 0 | 27° 18°Δ 0 | 29° 18°Δ 0 | 28–30° | 3° |
| Day 7to Wed 23 12Z | 55pressure | 25° 20°Δ +2 | 27° 19°Δ +1 | 25° 19°Δ +1 | 27° 19°Δ -1 | 25–30° | 2° |
| Day 8to Thu 24 12Z | 38temperature | 27° 17°Δ 0 | 29° 18°Δ -2 | 26° 18°Δ +1 | 28° 18°Δ +1 | 26–30° | 3° |
| Day 9to Fri 25 12Z | 44temperature | 22° 19°Δ -4 | 30° 18°Δ +4 | 24° 18°Δ -1 | 28° 18°Δ -1 | 25–30° | 8° |
| Day 10to Sat 26 12Z | 46temperature | 25° 18°Δ +2 | 23° 19°Δ -6 | 23° 19°Δ 0 | 26° 19°Δ -2 | 21–30° | 3° |
Open this place on the map
Values are what Forecast Euro's ingest sampled from each model's grid at this city, rounded to whole degrees at source. This page compares models; it is not a forecast to plan by.