Models are "smart" enough these days to recognize snowpack, the problem is that they will lay it down where they think it should go and as you say the effects cascade downstream. So for instance if we think we get a correction vector colder, we probably don't want to assume the storm that follows is a good bed where the GFS says it is.