AccuWeather Inc.'s seasonal outlook last month called for a mild December - "quite mild," meteorologist Bob Smerbeck reiterated Friday. Yet its most recent shorter-term forecast, derived through different methods, called for below-normal December temperatures.
For further evidence of the questionable state of the art, consult last year's snow-heavy winter outlooks - issued before one of the least-snowy winters in the period of record, at 8.3 inches.
What's the problem? The atmosphere is what scientists call a "nonlinear chaotic system," in which nothing happens in a vacuum.
Computer models have done wonders for short- and medium-range forecasting, not to mention speculation, which is why talk of a major holiday nor'easter began last week.
But they have not yet cracked the seasonal case. For short-term forecasts, meteorologists can count on guidance from computers that solve immensely complex equations about how the atmosphere will change over time.
Seasonal forecasters have to rely heavily on analyses, probing the atmosphere for various signals, such as pressure patterns over the ocean and the Arctic that ultimately affect winter in the United States.
"There's always going to be things we don't understand," said Mike Halpert, acting director of the climate center, where he has been trying to understand it all for 30 years.
Lacking this year are "high confidence factors," AccuWeather's Smerbeck said.
Halpert said forecasters were getting no help from the tropical Pacific, which is an atmospheric engine room.
When surface water temperatures are above or below normal, certain weather impacts in the United States are reliable for months, but the surface-water temperatures are now right about where they should be.
Halpert said the winter's intensity would depend heavily on the career of the North Atlantic Oscillation, or NAO, a pressure pattern that is a key factor in directing cold air into - or withholding it from - the Eastern United States.
The state of the NAO, however, is not predictable beyond 10 days, he said.
As for snow forecasting, the ice does not get much thinner.
The long-term average snowfall for the 126 years of record-keeping in Philadelphia would be about 22 inches, but in this case normal is abnormal: Snowfall has been within 15 percent of that average in fewer than a third of the winters.
In the winter outlook that he posted last week, which called for 14 to 18 inches of snow, NBC10 meteorologist Glenn Schwartz noted the last seven years had been exceptionally feast or famine.
More than 10 feet of snow fell over the winters of 2009-10 and 2010-11, but just over 1 foot combined in the two winters since.
Schwartz said he suspects the eccentricity has something to do with the way newly liberated sea-surface waters in the Arctic - long locked under ice - have interacted with the overlying air and affected air-pressure patterns.
Halpert said that such affects were possible but that analysis was hampered by a lack of long-term data; satellite observations have been available only since about 1970.
Schwartz said the possible link between Siberian snow cover in October and the U.S. winter is promising, pointing to research by Judah Cohen, at Atmospheric and Environmental Research. "It has been a better predictor than anything else I've seen," Schwartz said, noting that Cohen's analysis this year argues for less snow in the Philadelphia area.
"It looks like there's something there," Halpert said, adding that some of his colleagues were taking a hard look at a possible connection. Eventually, it might become part of a government outlook, but even if it did, it would be only one piece of an immense puzzle, he said.
None of the above is likely to be helpful for our winter planning. But meteorologists are going to keep at it, if only for the atmosphere's amusement.
Said Halpert, "We're trying to uncover the clues."