If you had asked the Europeans a week ago, they could have told you a truckload of snow was headed for Baltimore on Thursday. The Canadians came around soon afterward. It took the Americans until just a few days ago to get a whiff of a winter storm.
But as quickly as weather forecasting models developed by each can converge, they can shift, pulling the rug from underneath meteorologists who had warned of a possibility of more or less snow than what might actually fall.
By late Tuesday, the consensus was that six to 10 inches of snow could be expected across the region, with the heaviest precipitation falling through daybreak Thursday. Warm air high in the atmosphere could turn snow into sleet or rain around midday, cutting back on accumulation close to Interstate 95 and the Chesapeake Bay. Another round of snow was expected later Thursday evening.
Such a storm would be the biggest in a season that already has many yearning for spring. Gov. Martin O'Malley declared a state of emergency.
As meteorologists fine-tune their predictions of exactly how much snow might fall, they said the shifts illustrate the challenges of predicting winter storms in a geographically diverse region, amid a growing chorus of weather watchers, and often to the chagrin of a public that can obsess over every snowflake.
"It's getting crazy; especially with social media now, everybody's becoming a meteorologist in their basement," said Mike Masco. One of three meteorologists at WMAR-TV, he has been posting updates to his Facebook page for nearly a week on the potential for Thursday's snow.
When early forecast models suggest major snowfalls that can grab headlines and web clicks, some outlets share them without proper caveats, Masco said.
"It's really killing our credibility, and it's not even us forecasting this," Masco said.
Meteorologists rely on the three major forecasting models to develop expectations for everyday temperature and precipitation, as well as for storms from major winter snowfalls to hurricanes.
The complex computer systems factor in massive amounts of data on past weather and ocean and atmospheric characteristics to make predictions. But each model has its strengths and weaknesses professional forecasters consider, along with their own take on things.
The European model is considered the most robust, best handling dynamic systems such as Hurricane Sandy and its evolution into a "superstorm" as it merged with another disturbance in the atmosphere. Meteorologists also said they have found the Canadian model to be reliable and nuanced.
The American models, the flagship Global Forecast System, or GFS, and North American Mesoscale Model, or NAM, are generally taken with more skepticism. The GFS is known for skewing drier and wetter, for example, while the NAM can go overboard on snowfall predictions, said Eric Luebehusen, a meteorologist with the U.S. Department of Agriculture. Luebehusen has operated a Maryland-focused weather email list for friends, family and other weather watchers since 2000.
Most models have difficulty interpreting unique geographic conditions across Maryland and the Mid-Atlantic, with influences from the Appalachian Mountains, Chesapeake Bay, Atlantic Ocean and Piedmont Plateau. That's where forecasters come in.
Forecasts have erred in both directions this season when it came to wintry precipitation. Meteorologists anticipated freezing rain across the region last Wednesday, tapering off as temperatures warmed, but when temperatures remained hovering in the 20s into midday, the glaze of ice thickened, causing nearly 200,000 power outages across the state.
Models had also suggested snowfall measured in feet on Sunday, a prediction cited by many blogs and weather watchers on social media. But only an inch or two fell.
"It's easier to do than the lottery, but like the lottery and the stock market, there are no guarantees," said Marty Sharrow, an adjunct instructor at the Community College of Baltimore County who teaches a class on Maryland weather.
Even given the complexity of models, forecasts are not considered at their most reliable until a day or two in advance.
"You rarely have the final answer 48 hours out," Luebehusen said. "It's almost unheard of."
That has challenged those who have to make decisions based on forecasts.
If snow or ice hasn't accumulated overnight, officials in school districts across the state make decisions on delays or closures by about 5:20 a.m. Tricky forecasts have put officials in some parts of the state under fire for their choices — as in Harford County, when schools started on time amid rain last Monday morning, only to release students early as snow fell.