A good discussion of the uncertainties in prediction is at:
Snowstorm’s Forecasters Were Mostly Right Even if the model gets it right in terms of amount of snow, even a slight variation in storm track makes a big difference in how hard a given spot gets hit.
In this storm, the predicted snowfall gradients — charts showing how much would accumulate where — were very steep. “So a little bit of track error means a big snowfall error,” he said.
Mr. Miner pointed out that in this storm, the distance between an area that was swamped with snow and one that received little more than a glorified dusting was often small — as little as 30 miles. Parts of eastern Nassau County, on Long Island, for example, got as much as 18 inches, while parts of New York City received only four.
Mr. Miner added that, as with any storm, there were uncertainties about its epicenter. “This was a good case of where we could probably have done a better job of communicating the uncertainty,” he said.
Another discussion with more details about the various models, with a more technical discussion is at:
Blizzard Questions, Including Why a European Weather Model Excels at U.S. Forecasts