→Rough and Dirty ideas
Since snowtam and metar snow report consider mostly the runway, and there can be snow on the ground even if the runway is dry and clear (sunny spring days for example) there could be a method to get a rough idea whether there is snow on the ground at a given location and certain time of year.
Maybe an idea worth poking at would be some kind of geometrical lat/lon grid matrix for each month (or week?) of the year with either boolean value (snow / no snow) or perhaps something like
the 8 cloud coverage stages in METARs. 0 would be no snow, 8 would be total snow coverage, and the values in between could be for example implemented as somewhat randomized probability of snow, taking current METAR into account. Like, value of 1 would give you snow if temperature is below freezing and there is rain / snow in METAR. Value of 7 would give snow as long as temperature is not too warm, or something like that.
This might give us somewhat reasonable winter/snow season globally at most locations with reasonable effort..?