During the 6 p.m. news reports on Jan. 24, local meteorologists in the Washington, D.C., region forecasted "snow flurries" to begin around noon the next day with "up to an inch of accumulation on the ground." That was the last forecast many D.C. residents heard before going to bed that evening. But on Tuesday morning, Jan. 25, they awoke to find near blizzard conditions outside with 4 to 7 inches of accumulation already on the ground; and the winter storm wouldn’t subside until about 11 p.m. that night. In all, some areas east of Washington, D.C., received as much as 18 inches!

Living in the United States, we have grown accustomed to weather forecasts that are highly accurate as much as three days ahead of time. So why didn’t meteorologists foresee the winter bomb Mother Nature was preparing to drop on most of the eastern United States? Actually, they did says Dave Jones, meteorologist for NBC’s WRC-TV and principal investigator for WeatherNet4.

"The [computer] model runs on Wednesday, Thursday, Friday, and Saturday (Jan. 19-22) were indicating the potential for a major East Coast storm," Jones states. "But for some reason on Sunday and the early run on Monday (Jan. 23-24), the models did not continue to show a strong storm hitting the East Coast."

Jones isn’t sure why the models backed off after so many consecutive days of predicting the storm would hit us. He suspects there may have been some critical data that were needed that somehow either weren’t collected or weren’t entered into the model.

Meanwhile, all day Monday, TV meteorologists displayed satellite images of a huge cloudbank enveloping most of Georgia and North and South Carolina. Radar images revealed a vast blob of precipitation (mostly snow) being dumped on those states at a rapid rate of accumulation on the ground. If the storm moved north and east, there were implications for some of the busiest, most densely populated regions in the United States–Richmond, Washington, Baltimore, Philadelphia, New York, and Boston. But the question remained: What precisely would be the storm’s path?

Forecasters throughout the United States were all examining the potential for this storm and knew that it could be a major event, according to Louis Uccellini, director of the National Centers for Environmental Prediction for the National Weather Service (a part of the National Oceanic and Atmospheric Administration, NOAA). Their challenge was to accurately predict its path. If the storm traversed 50 or 100 miles to the east--as the models were indicating--it would have very little impact on our roads and cities. But if the storm took a more westerly route, then forecasters knew we could be describing the accumulation in feet, not in mere inches.

 The Winter Bomb

The data used in this study are available in one or more of NASA's Earth Science Data Centers.

View: low quality animation (1.5 MB)
high quality animation (4.9 MB)

The Winter Bomb

Over the weekend of Jan. 22-23, meteorologists were tracking a complicated upper atmospheric pattern made up of two separate but converging "frontal systems." Fronts are the leading edge of large air masses moving within the troposphere. Along and behind a front, the rate of change in meteorological variables—such as temperature, wind speed and direction, air pressure, and dew point (humidity)—is usually greater than ahead of the frontal zone.

Uccellini explains that there was a strong high-pressure system over the northern and eastern portions of the United States. Here, "high pressure" refers to a greater number of molecules of gases in the atmosphere. The more molecules—mainly nitrogen and oxygen—you pack into a given column of air, the higher the pressure. Think of blowing up a balloon; the more air molecules you breathe into it, the greater the pressure pushing outward and the more the balloon expands.

So the high-pressure was pushing southward from Canada, bringing frigid Arctic air, at the surface down the eastern side of the Appalachian Mountains, while a low-pressure "trough" was pushing eastward from the west. Here, "trough" refers to the region of minimum pressure, or fewest molecules. The opposite of a high-pressure zone, a trough behaves like a vacuum cleaner—sucking up molecules to fill the vacancy of so few molecules. Thus, troughs tend to create an upward motion of air masses.

According to Uccellini, these high- and low-pressure systems converged over Georgia and South Carolina in such a way that a storm began to develop. The trough was sucking up moisture and lifting it high into the troposphere. The cold Canadian air brought by the high-pressure front was in place at the surface supplying ample cold air for snow. On Monday, towards the east along the coastline it began to rain, but further inland it began to snow.

As the low pressure system continued to develop, it began moving in a large-scale circular motion, similar in size and energy to that typically observed in tropical storms that form hurricanes. Wind speeds increased on Monday until they peaked with gusts of around 50 miles per hour along the North Carolina coastline. These air masses drew up vast amounts of moisture from the ocean and transported it inland into the much colder air where it froze and fell as fat flakes, dropping almost a foot of snow on Charlotte, North Carolina over a 12-hour period. (Ice crystals at or near the 32°F, or 0°C, freezing point are very "sticky" to the touch and tend to aggregate into much larger snowflakes; whereas in much colder conditions, snow flakes tend to be smaller because colder air holds less moisture.)

Moving slowly, the storm trundled northward and laid a thick white carpet on almost the entire eastern seaboard, two feet deep in places such as the suburbs of Raleigh, NC, and Annapolis, MD. Forecasters call such a meteorological event a "winter bomb." What did they learn from this one?

 What’s wrong with the forecast models?
 Second Guessing Mother Nature

What’s wrong with the forecast models?

Uccellini admits that there was some uncertainty in the minds of meteorologists as to what path the storm would take. He says the National Weather Service could have emphasized that uncertainty more in its forecasts. But forecasters have become spoiled by the increasing accuracy of computer models and nowadays they rely upon them quite heavily. Forecasters are reluctant to publicly present weather scenarios that contradict the models. It is far safer for one’s professional reputation, for instance, to follow the mainstream rather than to be the lone voice "crying wolf."

"Forecasters know the models aren’t perfect," Uccellini says. "There should be some way to present alternate scenarios without creating a panic."

Jones adds, "My feeling is that most numerical models have gotten so good in the two- to three-day timeframe, most meteorologists thought the storm would move out to sea [having little impact on land]. But 12 or 24 hours before it hit, there may have been some indication from satellite imagery that the storm wasn’t behaving as the models were forecasting."
 

Jones advises forecasters to use the entire suite of meteorological tools at their disposal. He says satellite imagery is critical to the science of meteorology today and in the future. He also advocates the use of surface-based measurements, upper-air balloons, aircraft measurements and even radar. Perhaps paying closer attention to these latter data sources would have enabled forecasters to predict more accurately the blizzard’s course.

The fact remains, however, that predicting the weather days in advance is inherently prone to error. Uccellini notes that even with all their numerical models and all their measurements and satellite observations, meteorologists still cannot obtain data on every point in the atmosphere and ocean.

"Ocean and atmosphere are two fluids," he explains, "and as they interact they go through many non-linear dynamics that produce our weather. (Some of these ocean-atmosphere dynamics are discussed in-depth in the Ocean & Climate fact sheet.) The best you can do is estimate and interpolate what’s going on there. In so doing, you’re going to introduce errors within the models. Over time, those errors grow until eventually they render the model useless."

For this reason, says Uccellini, the limit for weather forecasts is set at 14 days. The errors within the observations and models are such that after 14 days they have "snowballed" until the model is useless for predicting any specific event.

Jones cites the Chaos Theory, stating that scientists still don’t understand the ways that small-scale events can end up having large-scale effects. "It goes back to the question of whether the flap of a butterfly’s wings in Brazil could lead to a chain of events that causes a tornado in Texas," he says. "The models would never forecast those types of chain events...perhaps one day, but not now."

 Today’s Forecasts Are Better Than Ever
 The Winter Bomb

View: low quality animation (.85 MB)
high quality animation (3.3 MB)

Today’s Forecasts Are Better Than Ever

Uccellini points out that there has been a steady improvement in weather forecasts over the last two decades. He says there was a 20-30 percent improvement in the last 7 years alone.

"There have been two fundamental changes since the late 1980s," he states. "(1) We have extended our predictions to the 7-day range with an accuracy that is really remarkable. Today, our accuracy in the 5-day range is as accurate as it was in the 3-day range 20 years ago."
 

 
 
Top: Weather forecast models have steadily improved over the last 20 years. This graph shows how well model forecasts scored when compared to a grid of actual meteorological measurements from 1977 to 1990. "Anomaly correlation score" refers to the statistical relationship between a geographical grid of measurements and a corresponding grid of numbers generated by models. Notice that by 1990, 5-day forecast models were as accurate as 3-day forecast models were in 1977. (Graph by Robert Simmon, courtesy of Louis Uccellini, NCEP)

Bottom:As model forecasts improve over time, they help improve "manual" forecasts made by humans that incorporate the models into forecasts that also use the larger suite of meteorological tools at their disposal (e.g., satellites, radar, aircraft, and ground-based instruments). The bottom line shows the improvement in the accuracy of model forecasts, from 1977 to 1990, and the top line shows the relative improvement in the accuracy of manual forecasts. (Graph by Robert Simmon, courtesy of Louis Uccellini, NCEP)

He continues, "(2) We are now able to predict major weather events farther in advance than light to moderate events."

Remember the winter bomb that hit the U.S. East Coast in March 1993? The National Weather Service predicted that event 5 days in advance. Later, in January 1996, we had another major snowstorm that produced more snow than any other storm in the 20th century; again, this was forecast 5 days in advance. Uccellini asserts that no one could have made those predictions 20 years ago.

"But Mother Nature has a way of exacting revenge," he laughs, "like last Monday, when we only got a warning out 8 hours in advance. While that didn’t meet our current standards, it really is a testament to how far we’ve come."

Okay, so meteorologists missed the snowstorm of Y2K. They nailed with astonishing accuracy the next winter storm that arrived almost a week later. Forecasters predicted time of day the precipitation would begin, and precisely maps the boundaries of where there would be snowfall, where there would be sleet, and where there would be rain. They even accurately described snow accumulation totals.

Jones not only agrees with Uccellini, he believes the best is yet to come...and soon. He says as meteorologists fine-tune their models and computers get more powerful, in 10-15 years forecasters will be able to accurately predict the weather 10 days, perhaps even 14 days, ahead of time.

 Wacky Winter Weather
 What’s wrong with the forecast models?

Wacky Winter Weather

Across much of the United States, the 1999-2000 winter has produced some unusual weather patterns. For instance, the Pacific Northwest has already received record amounts of precipitation while, until the winter storms in late January, the mid-Atlantic region had received very low amounts.

"Our winter weather pattern is being influenced by La Niña," Uccellini observes. "This is characterized by a strong flow of air from Canada over the Great Lakes region and toward the mid-Atlantic. That is a known factor influencing our winter weather patterns; it was expected and incorporated into the winter weather forecasts."
 

In general, says Jones, La Niña causes wintertime temperatures to be higher than average so we are more likely to get more rain or ice storms, such as the one that occurred along the East Coast on Superbowl Sunday. This winter has certainly been milder than average and, up until the winter bomb, precipitation amounts have been lower than average. As a matter of fact, until the Jan. 24-25 snowstorm, Washington, D.C., had experienced only two snowfalls of 6 inches or more during a La Niña winter.

By now, most people are only too familiar with the El Niño/La Niña climate phenomena, thanks to all the negative press these climatic siblings receive. But Uccellini says there is another strong influence on weather patterns in the northeastern U.S. that many people haven’t yet heard of–the North Atlantic Oscillation (or NAO). It is a distant cousin to the El Niño/La Niña Pacific oscillation. (Stay tuned, a case study on the North Atlantic Oscillation will appear in the Earth Observatory later this month.)

"When we (the National Weather Service) issued our winter forecast in November 1999," Uccellini recalls, "we said that the NAO would be the wild card for the northeastern U.S."

In January, the NAO caused a build-up of very cold air over Greenland and southeastern Canada. This has had two effects on weather in the United States: (1) it became a source of frigid air filtering down into New England, causing temperatures there to plummet, and (2) it provided the colder temperatures needed to turn rain into sleet and snow.

"We do look to the Pacific Ocean as a basis for our winter forecasts," Uccellini says, "but we also look at what’s happening in the Atlantic. We still have a lot of research to do because we don’t have a complete understanding of how these large-scale ocean-atmosphere phenomena work."

Jones adds that it is too soon to tell what scientists have learned from this wacky winter weather. He says it will probably take years, and more data, before they learn lessons that meteorologists can apply toward improving their models.

Reference:

Uccellini, Louis W., Paul J. Kocin, and Joseph M. Sienkiewicz, 1994: "Advances in Forecasting Extratropical Cyclogenesis at the National Meteorological Center," from Proceedings of of the Geophysical Institute's International Symposium at the University of Bergen, Norway. Edited by Sigbjorn Gronas and Melvyn A. Shapiro; published by the American Meteorological Society, pp. 259-274.

 Today’s Forecasts Are Better Than Ever

This image shows an exaggerated topography of the equatorial Pacific region where height corresponds to the difference in sea surface height from normal. Color corresponds to the difference in sea surface temperature from normal. (Image by Greg Shirah, NASA GSFC Scientific Visualization Studio)

Return to: Second Guessing Mother Nature

Return to: Second Guessing Mother Nature

Return to: What’s wrong with the forecast models?

Return to: What’s wrong with the forecast models?