I'd love to see this on a globe rather than a Mercator projection.
It's interesting how they still need to do some fudging around the poles. Non-Euclidean geometry is hard, I guess. I wonder if that means the models are all done with Euclidean geometry or is it just a display thing?
Most often, final global models are output on to an orthogonal (plate-caree) lat/lon grid. Data at the poles (singularity) is always a bit fidly