Evaluate GFSv17 performance in 2026 winter storms

[yuanxue2870]

Copied from Clara's email:

1. Prepare plots of the two snow events, comparing v16 and v17 against SNODAS, focussing on binary snow/no-snow. Dan liked the way I split out the categories (hits, misses, etc) in my slides, so lets stick with that approach.

2. Prepare plots of the monthly binary snow/no-snow (to look into the low snow in the west). We'll have to think about how to present monthly averages of snow cover (maybe just total days with snow cover since we're looking at bias here? Of we could also look into mean snow depth?). A timeseries of the snow-covered area in each region could also be useful.