My examination of the available 00Z guidance reveals that the 00Z high-resolution guidance has come to much-better agreement with the 12Z coarser-resolution guidance. If you missed the earlier tweets: https://twitter.com/VORTEXJeff/status/1357788292205449219?s=19

I'm not surprised (I forecast 1-3" of snow for Champaign-Urbana in lab this morning, even without the 12Z Euro, despite the high-res 12Z guidance going much higher) because most of the coarser-resolution models are quasi-two dimensional, and this is important.

Theory states that purely two-dimensional turbulence is infinitely predictable. (Imagine living on a completely flat world with no vertical dimension.) That's obviously not the truth. However...

Synoptic-scale weather systems are 1000s of km across and only ~10 km deep because they're limited by the depth of the troposphere. A midlatitude cyclone may extend 1000+ km horizontally, but only 10 km vertically

Midlatitude cyclones are like atmospheric pancakes. And over large-enough scales and long-enough time scales, even synoptic-scale cyclones become turbulence

Smaller-scale phenomena, like thunderstorms, are much smaller in the horizontal direction, but still exhibit the same depth in the vertical, making them fundamentally three dimensional.

With finer grid spacing, like in the high-res models, three-dimensional atmospheric circulations can be resolved (length and width are both 10 km). Unfortunately, three-dimensional turbulence is chaotic; it's not predictable beyond a point, unlike 2-D turbulence.

If you've seen the movie "Jurassic Park," you might be familiar with chaos theory. Chaos theory was developed in the 1960s by examining mathematical solutions to equations of atmospheric flows.

This means that the large-scale predictability from high-resolution models is limited compared to coarser-resolution models, which is why models like the HRRR and 3-km NAM are not run as far into into the future as models like the GFS and ECMWF.

TLDR: When examining small-scale phenomena, give deference to the high-res models, but when you're seeing differences on the synoptic scale, go with the coarser global models.

Did I teach this in forecasting class this week? Absolutely! Resolution != Accuracy.

Final call for snowfall in Champaign-Urbana Saturday evening is 1-3".