Why forecasts missed the big snow


Storms: Technological advances have improved accuracy, but tracking the climate changes that lead to such weather remains tricky.

February 18, 2000

In a season of volatile weather patterns, the National Centers for Environmental Prediction are where America's weather and climate services begin. Observations from around the globe -- taken by balloon, aircraft, satellite and more -- are funneled continuously to the agency, which is the central operating unit of the National Weather Service. A new, 3,000-square-foot IBM System Parallel supercomputer crunches the numbers and generates weather maps for 121 local forecast offices around the nation.

The system provides a "high-resolution global model," said Louis Uccellini, director of the center. But the weathermen had uncharacteristic trouble foreseeing the storm that hit the area the night of Jan. 24 and the next morning.

On an unusually warm, sunny February day at his office in Camp Springs, Uccellini talked to The Sun's Jonathan Pitts to put that weather event in perspective and to cast a cold eye on the area's meteorological near-future.

The snowstorm of 2000 took a lot of people by surprise. Are snowstorms inherently hard to predict?

An East Coast storm is a relatively rare event, especially for Baltimore-Washington. In 50 years, from 1950 to 2000, only 28 storms have affected the major metropolitan areas. So it's about one every two years. Rare events are more difficult to deal with than things that happen regularly.

Many things have to come together just right. The key factor is getting warm, moist air into the upper layers while still sustaining colder air in the lower layers. Then, whatever starts as snow in the clouds will fall as snow and make the big headline.

If you get one little sliver of warm air in there, you get sleet [instead]. If you get a deeper level of warm air in the lower layer, you get rain. These are very difficult processes to understand, very difficult to model [on computer]. You need very fine resolution to do it. It becomes a difficult thing to forecast.

Yet the National Weather Service usually succeeds at that.

We've gotten better at forecasting the big storms since 1990. For the March storm in 1993, we were out there five days in advance telling people it was going to happen. In 1996, for the storm that buried this whole area, we were telling people three days in advance, two days in advance, a day in advance. The day before, we told people they were going to measure it in feet, not inches. We nailed it. I guess we were just due for a miss this time.

In [this blizzard], the low-pressure system was actually predicted a day in advance. We just didn't have the two- and three-day watches out that people have come to expect.

What does "low pressure" mean?

Low density. The pressure at the Earth's surface is a measure of air density up to the top of the atmosphere. Where mass is removed, pressure will fall. Air from surrounding areas will start to converge into it. That's how warmer air from one source -- from the south and east -- comes together with colder air coming from the north and west.

The day before the big storm -- Sunday night into Monday morning -- the [computer] model started predicting a low, but only along the coast. And each successive [data] run kept pulling that low closer to the coast. For a reason we still don't completely understand -- we're doing a detailed study on this -- the model, even with the developing low, did not [locate] the precipitation shield far enough west. It kept the heavy precip along the coast, with a very, very sharp western boundary.

By Monday, we had a forecast of snow out for the Eastern Shore, but we didn't expect the precip to get to the major metropolitan areas. By Monday afternoon, the western boundary of the precip shield was over Washington and Baltimore. And at 7 Monday night, not only was it over Washington, but now the maximum precip shield was over both cities. It changed that suddenly.

Was it hard getting the word out?

When we saw this trend, we very quickly amended all the forecasts, got on the phone with the emergency-management community, with the TV stations, with the people responsible for the roads, and told them we were going to be in a major snow area tomorrow morning.

We had all the forecasters updated before the 10 o'clock news, and the 11 o'clock news had the warnings out there. We said it would be 10 to 12 inches or more before the first flake hit. Normally we forecast in the 4- to 8-inch area and update as the storm evolves. We really hammered that point.

What have we learned from this storm?

There were some observation issues over the Southeast and over the Atlantic. We had observations that the initial run of the model just threw out. The data came in and was just not used. The wind data that came in from Peachtree, Ga., and Tallahassee, Fla., were so out of whack from the previous runs -- the differences were in the range of 40, 50, 60 knots-- that the models simply rejected the data.

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.