Very fair questions, which I've addressed in past threads like this. Given that this is an RU forum, I think New Brunswick is a good location to look at and the average January snowfall in NB over the 116 years we have records for is 7.1", so the 2nd half of January would be a little over 3.5", on average.
Defining what would be "snowier than normal" could be as little as 7.2", but probably a fairer number might be being in the top quartile of January snowfalls, which would be getting above 10.0" of snow for the month or 5.0" of snow for a half month. So, I'd say if the 2nd half of January is above 5.0", then that would count as "significantly snowier than normal." Others may feel differently - for example getting into the top 10% of Januarys would require 16.7" of snow or 8.4" for a half month. The problem is snowfall is a single-sided non-linear variable, where the range of the top 10% is huge (16-31") while the range for the bottom 10% is tiny (0-0.6") - that's why the median snowfall for January in NB is only 5.0" vs. the average of 7.1".
There's also no agreement on what constitutes a "significant" snowfall. My personal definition has generally been about 4-8", but I think just looking at the total snowfall over the last half of the month is the way to go. I also see little value in having others put up "guesses" as I'm simply interested in whether the actual snowfall is well above normal or not for what looks like a potentially colder and snowier than normal pattern. Obviously, though, others are free to put up their guesses.
http://climate.rutgers.edu/stateclim_v1/monthlydata/index.php?stn=286055&elem=snow
Finally, at this range operational models are fairly useless for guesstimating snowfall even for Sunday's possible event, let alone the 2nd half of the month, given the huge uncertainties introduced by chaos theory for numerical weather predictions, given the known uncertainties in model inputs for initial conditions and boundary conditions. That's why most pros look at ensemble forecasts for any particular model, where the model is perturbed in random ways for the initial and boundary condition inputs, to get a sense of the sensitivity of the model around the operational model run.
And right now, the ensemble mean for the Euro, the best model, is showing well above normal snowfall for the 9th through the 24th (with none of that before the 15th), below, including about 7" for NB, which would be well above the 5" I proposed above for a significantly above normal snowfall for the 2nd half of Jan in NB. That's not a forecast and it may be 100% wrong, but it's simply showing that the potential for snow is much greater than usual (ensembles through Dec were typically showing 1-3" for the next 15 days).
