MODELS
Seven models, every run: what each one is, when it last ran, and how accurate it has been against observed degree days.
The roster
Accuracy leaderboard
Lower 48, last 30 days: how far each model has been from what was observed, by days ahead. Samples differ by model; compare identical samples on Verification.
Mean absolute error by days ahead
Against observed degree days (NOAA nClimGrid), 00Z and 12Z runs. Each model uses its available sample; the lowest MAE at each lead is marked.