Weather model evaluation seems similar to machine learning on just about any dataset. Want to reduce variance and have a more generalized prediction (it'll probably snow, but not sure how much)? Random forest classifiers can help. Want to reduce bias and get more precise with potential overfitting (don't worry about downsloping, it's gonna snow 9.8")? Gradient boosting. Want a mixed bag, but high cost and a potential black box (it's gonna snow somewhere between 1 and 15"), go with ensembles that have a variety of algorithms.
Judging by the median (which I'm assuming is the wise thing to look at lol), looks like the AIFS ensembles are leaning towards snow tv, which is still a win in early December.