Polar Explorer

Web Mercator is infinite at 85.05°. That is a property of the projection, not a limit of the data — ECMWF has values right up to the pole — and it means no Mercator raster can contain a pole at any size. The global Explorer stops at 85° because that is as far as the projection goes.

So the caps get a polar stereographic render instead: true at 70°, oriented on Greenwich, and guaranteed to 45° on every meridian — the corners reach further still, to about 29°. A stereographic raster cannot be draped on a Mercator map surface, which is why this page is a small purpose-built map rather than another layer on the Explorer.

It matters most here because this site draws global models, and the two places Mercator fails hardest are the two where a hemisphere behaves as one system: the polar vortex, the sea-ice margin, and the cross-pole flow between Siberia and Canada are all things a map cut at 85° simply cannot draw.

All four models are drawn here, and the caps are where that comparison is worth most. The Arctic and Antarctic are where a global model has the fewest observations to be corrected by, and the polar vortex at day ten is exactly the kind of large-scale, medium-range structure a machine-learning forecast — or a fifty-one-member mean — is supposed to be good at. So HRES, AIFS, the ensemble and WeatherNext 2 get the same projection, the same colours and the same instrument — switch between them with the Model buttons and nothing else on screen changes. Each carries whatever its feed publishes: 22 fields for HRES, 18 for the ensemble, 10 for AIFS and 4 for WeatherNext, which is experimental rather than operational.