Model rankings

Compare forecast models head-to-head.

Global verification against ERA5 reanalysis. Updated daily after each GFS cycle. All scores are Northern Hemisphere extratropics unless noted.

Variable:
Rank ▴ Model Type T+24h RMSE ▴ T+72h T+120h T+240h Skill

Rankings are based on the selected variable's T+24h RMSE (Root Mean Square Error) against ERA5 reanalysis. RMSE is computed over the Northern Hemisphere extratropics (20°N-90°N) for the most recent 30-day rolling window. Lower RMSE indicates better accuracy. Models marked "AI" use machine learning; "NWP" indicates traditional numerical weather prediction.

Visual comparison

Side-by-side skill assessment at different forecast lead times.

T+24h RMSE comparison

2m temperature (K) — lower is better

T+120h RMSE comparison

2m temperature (K) — lower is better

T+240h RMSE comparison

2m temperature (K) — lower is better

Skill retention at day 10

T+240h RMSE / T+24h RMSE — lower ratio = more graceful degradation