The Blizzard of '93 slathered the Atlantic coast in 10 to 50 inches of snow, scattered record lows from Florida to upstate New York, fanned winds up to 110 mph, poured a quick-freezing mix of snow and hail on Baltimore, blocked interstate highways, sank ships and caused at least 193 deaths.
It was a monster, a killer, a once-in-a-century snowstorm. But it did a lot less damage than it might have because people had plenty of time to get out of its way.
What made that early warning possible, said Robert Derouin, deputy chief of meteorological operations at the National Weather Service's National Meteorological Center, is bigger computers, more raw data and a better understanding of how to put the two together to accurately mimic the daily clash of forces in the atmosphere.
Forecasters at the center, located in a run-down office building in Camp Springs, first spotted the brewing atmospheric imbroglio on Monday, March 8 -- four days in advance.
Day-by-day, the scientists watched the storm develop, just as their computers predicted. Meanwhile, they fed their increasingly gloomy weather maps and data to 274 local weather service offices scattered around the country, as well as private forecasters, television and radio stations.
By Tuesday, local forecasters began spreading the news that snow was expected over the weekend as far south as Georgia. By Wednesday, "we had the first indication that it was going to be a superfantastic East Coast storm," said Paul J. Kocin, a research meteorologist with the meteorological center.
Road crews and the National Guard were alerted. Motorists canceled weekend travel plans. Most folks had ample time to strip supermarket and video store shelves, then hunker down in their dens and club rooms for the long winter's weekend.
Fewer people were killed, chilled, stranded and stressed than if the storm had sneaked up unnoticed. The credit, colleagues say, belongs mostly to the meteorologists and computer jockeys at Camp Springs.
Eugene M. Rasmusson, a senior meteorologist at the University of Maryland at College Park, called it "a stunningly good forecast" that took some courage to make.
"I told my wife Friday evening, this storm's either going to be one tremendous hit or one tremendous bust," he said. "The weather service didn't hedge on it. They went all out. And it worked out beautifully."
In 1955, computers could calculate local weather patterns with reasonable accuracy for only about three days in advance. Today, local forecasts are relatively reliable out about six days.
Those three extra days may not seem like much, but predicting the weather is very difficult because of what scientists call the butterfly effect. The weather is so sensitively balanced, that it can be strongly influenced, in the long run, by tiny changes in conditions.
The beating of a butterflies wings in Brazil, it is sometimes said, can spawn a hurricane in the Caribbean three months later.
Of course, meteorologists hope the weather is not quite that sensitive.
But so far they have little luck with extremely long-term forecasts.
The weather service's detailed, local 24-hour forecasts are now more than 95 percent reliable.
By contrast, monthly forecasts -- which predict broad temperature and precipitation ranges for regions of the country -- are just 14 percent better than random guesswork for temperatures and 5.7 percent better for rain or snow.
Ninety-day forecasts score just 5 percent for temperatures and 3 percent for precipitation.
Predicting that the Blizzard of '93 would develop four days in advance was not an astounding feat in itself, meteorologists said.
The storm was a classic example of what is called a Northeaster -- where cold polar air sweeps south along the Mississippi Valley into the Gulf of Mexico, picks up prodigious moisture and then takes a U-turn up the East Coast, spreading snow in winter and rain in summer.
"This is just sort of a hybrid of a fairly typical weather system, just very extreme," said Mr. Kocin, an expert on Atlantic storms.
What was remarkable was that forecasters saw early on that the storm would do something very unusual -- quickly approach peak intensity over the Gulf and the southern United States. Typically a Northeaster doesn't turn vicious until it reaches New England.
Meteorologists were able to predict the storm's size, its production of rain and snow, its timing and its track with great precision, Dr. Rasmusson said.
"I started forecasting weather in 1952 when we drew our own maps with a pencil, so I've seen the changes," he said. "Even as recently as three years ago, I don't think this kind of accuracy would have been possible."
Every day, the Camp Springs scientists harvest thousands of readings of air temperature, pressure, wind speed and other data from all over North America, the Caribbean, North Atlantic ,, and the North Pacific.
The numbers come streaming in from satellites, weather stations, about 100 balloons launched daily, a few advanced Doppler radar stations that track rainfall, 29 ground-based microwave "sounders" that measure wind speeds, ships, buoys, planes, oil platforms and other sources.
Twice a day, at 7 a.m. and 7 p.m., the Camp Springs scientists plug the data into four computer programs that simulate, or "model," the churning atmosphere.
The programs are run by a $20 million Cray supercomputer, a number-crunching behemoth housed in a building in Suitland. Two programs produce 3- to 5-day forecasts; two produce long-range, 10-day forecasts.
Each program takes the raw numbers and manipulates them using mathematical formulas that reflect the physical laws that govern changes in the atmosphere.
In this way, the computer decides what the weather is most likely to do 10 minutes into the future -- the fraction of a mile a warm front will cover, for example, or how far the jet stream might wiggle from its path.
Then the computer takes the new pattern of predicted temperatures, pressures and wind speeds and cycles them through the same series of mathematical formulas. That fast-forwards the computer's digital "picture" of the weather forward another 10 minutes.
It does this over and over again.
To simulate 10 minutes of weather, the computer must do 45 million calculations. To simulate 10 days, the Cray does about 70 billion calculations.
The number of calculations is mind-boggling. "It's almost like the national debt," Mr. Derouin said.
Humans couldn't handle all this math without extremely fast, powerful machines. The Cray works so fast, in fact, it calculates a 10-day forecast in just 1 1/2 hours.
The picture isn't always clear. "Sometimes we get a tremendous amount of disagreement among models, which makes you pull out your hair because you have to try to decide which one you're going to believe," said Mr. Kocin.
That didn't happen with the recent storm. By 8:30 p.m. Monday, March 8, when the new maps were put up on the 4th floor of the Camp Springs building, National Meteorological Center scientists knew they faced a major Northeaster. All four models predicted it.
Over the next three days, the emerging picture of the storm grew more detailed. But there was still a lot of anxiety in Camp Springs.
"If storm had passed 50 miles to the east, the sleet that hit Baltimore Saturday afternoon, for example, could have been another foot of snow," Mr. Kocin said. "Further west 50 miles, it would have been 2 inches of snow and a lot of rain."
People living in areas that got a lot more or less snow than predicted would not have been very impressed with the forecast.
(People on the Eastern Shore were probably less than impressed with the forecasters prowess, in fact. Ocean City got barely a 1/2 inch of snow, and the hurricane-force winds never showed up.)
Never mind that the storm slammed most of the East Coast, as predicted. "People's expectations are very high for locally-specific forecasts," Mr. Kocin said.
Normally, activity at the National Meteorological Center is less intense than at its two sister agencies, the National Hurricane Center in Miami and the National Severe Storm Forecast Center -- which tracks tornadoes -- in Kansas City, Missouri.
The Camp Springs center could be considered the unglamorous workhorse of the weather service. It gathers, massages and dispenses data for routine local forecasts.
But the center had a chance to shine with the Blizzard of '93, which dumped more water over the East Coast than flows out of the mouth of the Mississippi River in 40 days.
As the storm approached, scientists at Camp Springs began coming to work early and staying late.
"We're kind of a strange group," Mr. Kocin said. "A lot of us have a fascination with extreme weather events. We find it very exciting."
About a dozen meteorologists normally gather in the 4th floor meeting room for "map discussions" after the two daily computer model runs. But at 11:30 a.m. on March 11, the day before the storm began, about 75 people packed the meeting.
"That was a new world record," Mr. Kocin joked.
Mr. Kocin came in to work early Friday morning and began plotting maps each hour tracking the development of the storm. "It actually was developing faster than I expected," he said.
He stayed until 5 a.m. Saturday, when, exhausted, he drove back to his apartment. He tried to return Saturday afternoon, but couldn't get his car out of the underground parking garage. "I was very disappointed," he said, because he was cut off from detailed information about the storm.
Arriving at work Sunday, he began to piece together the history of the blizzard -- which was the first, he said, to blanket the entire East Coast since 1899.
The blizzard didn't bring record snowfalls everywhere it hit. In Baltimore, the blizzard of 1983 spread 22 inches over the city. In February 1979, the city was hit with 20 inches on top of 10 inches from an earlier storm. The recent storm, meanwhile, produced 10-12 inches around Baltimore.
But, Mr. Kocin said, the Blizzard of '93 probably caused more disruption in more lives than any previous storm in East Coast history.
"I don't think there's been another storm that produced at least 10 inches of snow over as wide an area as this one did," Mr. Kocin said. "It was very crippling. You're affecting tens of millions of people."
The mixture of snow, sleet, hail and freezing temperatures also contributed to the storm's power to gum up the gears of civilization.
"This was an unusually hard snow to manage," Mr. Kocin said. "It had extremely heavy, dense snow that was hard to dig out from."
Some people were skeptical that a big storm was coming so late in the season, because since 1987 the East Coast has had extremely warm winters and very little snow.
"Now," Mr. Kocin said mischievously, "they have a slightly different perspective."
No one knows, he said, whether this rapidly waning season marks the beginning of a new cycle of snowy winters.
"It's possible, or it might just be an isolated event," he said. "We've got some snow fanatics here that are hoping it's a sign of better things to come."
Douglas Birch covers science for The Baltimore Sun.