For the weather wonks who divine the future of Washington's skies with streaming satellite images and glowing screens of raw numbers, the snow carpeting yesterday's morning commute brought humility at missed forecasts and awe at Mother Nature's age-old ability to surprise.

Meteorologists took a tongue-lashing yesterday from slip-sliding drivers caught in epic traffic jams -- the same ones who have come to expect forecasters to get it right, down to the hour.

The snow "started several hours earlier than I expected," WJLA (Channel 7) meteorologist Doug Hill said. "I got totally bamboozled" -- along with most of his colleagues.

"That doesn't make me feel any better," Hill said. "A lot of people trust me. When I let them down, I don't feel good about that."

But for Hill and the other forecasters -- many of whom began their careers as kids with backyard weather stations and penciled precipitation logs -- yesterday's snowfall was simply another example of what hooked them on weather in the first place: its very unpredictability.

"It's fascinating that we don't have the answers. Every single storm is different," said Barbara Watson, the National Weather Service's warning coordination meteorologist for the Washington area. "We're better at saying it's going to snow than exactly when. . . . That's just where the science is."

Yesterday, timing was everything: Snow once predicted for afternoon and then for late morning began falling before dawn.

A series of Weather Service bulletins issued Tuesday shows a forecast in flux. At 10:04 a.m., the forecast for yesterday read: "A CHANCE OF SNOW IN THE AFTERNOON." By 3:30 p.m., it had changed to: "SNOW LIKELY . . . DEVELOPING LATE IN THE MORNING." A "forecast discussion" released by the Weather Service at 9:40 p.m. said: "PRECIPITATION COULD START BY EARLY MORNING."

Forecasters said disparate weather models were sending complex signals that meteorologists had to interpret based on experience and instinct.

"New model information came in that supported the idea that it could snow anytime," said Kerry Schwindenhammer, an AccuWeather forecaster. The models showed a low-pressure system moving closer to the Washington region than expected, meaning that snow might -- might -- fall earlier than expected.

But interpreting the models is the hard part, said Schwindenhammer, 28, who was a fan of the Weather Channel as a boy. "The models, of course, have their own biases. In some situations, they do well. . . . Then, all of a sudden, they're way off."

Schwindenhammer said his job is to "decide what's believable and what's not believable." In the end, he said, he predicted snow for yesterday, without saying exactly what time it would start.

Hill said he got the amount of snow right but the start time wrong.

"Weather is . . . going to do what it's going to do. Your game is to try to stay on top of it," he said. The goal remains perfection, but "no one's been able to do that yet."

The degree of remorse had more to do with the result of the storm than with its size. "It wasn't like there was a big, ugly storm that sneaked up on us," Hill said. "It just came at the worst time. All of a sudden you've got people sliding around like they're on an ice rink on a half-inch" of snow.

One problem, he said, is that detailed computer models need data that aren't always available: "The bottom line is, out over the ocean there are no reporting stations. The radar only travels so far."

Watson noted that local weather has distant origins. The low-pressure system causing Washington's snow had been centered in Texas. The moist air in its flakes began in the Pacific Ocean. In a science of vast distances, missing something by a mere 40 miles "could mean everything" for Washington's forecast, she said.

"We feel pretty good," Watson said, noting that the Weather Service announced "days in advance" that it would snow yesterday and today. "If people say, 'Hey, you missed it by three hours . . .' that's a pretty good forecast."

Staff writer John F. Kelly contributed to this report.