Click on the table cells to select other images. Clicking on an already selected cell enables hover mode. Then the images will switch on mouse movements, but be patient the first time an image loads.
Animations of the first forecasts or analyses are also available.
Digits are offsets in days from the forecast initialization date; older forecasts are on the right.
Ice concentration is fixed in the GFS model; it doesn't change over time. Instead, modeled ice thickness may decrease to values near zero. Ice thickness changes are 24 hour changes. Note that the model doesn't account for ice movements.
Temperature is a mean of 8 values (every 3 hours) per day. The "valid" time is the time of the last value (in hours since model initialization time).
Wind streamlines should be correct now; the first few days (where I didn't save the original data) still suffer from a bug in GrADS where streamlines near the poles are wrong and too dense (patch here).
The ground heat flux is the heat that is transferred from the surface to the ground. The surface can also gain or lose heat through downward and upward radiation, and by transfer to or from the air and by evaporation or condensation. I don't know the model details. 1 MJ/mē can melt about 3.25 mm ice, but I guess that most of the melting heat comes from the ocean.
Direct links to the images: