prediction accuracy metalayer
An odd and unlikely idea.
I've sometimes wondered if forecast accuracy could be itself used as a measure of climate (in)stability. If you used the same forecast setup over multiple years and recorded the accuracy of the predictions, you could provide an overall global weather predictability graph and see whether it changes much over the years. Not possible to do that here, I guess, but - on that note ...
It'd be interesting to have a "meta" prediction accuracy layer, one that compared current readings with the predictions from a few days ago (maybe using some combination of pressure, temperature, and precipitation). I guess you'd want to update it rarely, to reduce the data needing to be stored. But it could possibly help people learn how reliable weather forecasting is at their location.
Also (to cap off this totally random topic), some of these might be cool :)
(especially the butterfly)