Chet Henricksen lived closer to the office than any of his colleagues, and on one particular occasion that turned out to be to his great disadvantage.
It was the early morning hours of Feb. 19, 1979. The phone rang. The mid-Atlantic region was under extreme attack from nature — a snowstorm had exploded off the coast.
Being the lead forecaster nearest National Weather Service headquarters in Suitland, Md., he was desperately needed to come in and run the show. He quickly encountered a weather-related obstacle: What he saw from the windshield of his Chevy was a “blinding” snow that was falling and drifting so rapidly that it “was almost impossible to drive through.”
Henricksen, longtime head of the Philadelphia region’s weather service office and now retired and living in Downingtown, was summoned because no one had seen this coming, thus no staffing contingency plans were in place.
What happened on Presidents’ Day 1979 was an absolute shock to some of the best minds in meteorology and computer models. By the time it ended, two feet of snow had creamed the Washington area, with just over 14 inches in Philadelphia and close to 13 in New York City.
The forecast whiff set off storms of criticism, finger-pointing, internal soul-searching, and analysis, and ultimately helped lead to significant improvements in forecasting, said Louis W. Uccellini, perhaps the nation’s foremost expert on winter storms, who in his spare time now runs the National Weather Service.
“When it hits D.C., you know it’s going to get people’s attention,” said Dave Dombek, a senior meteorologist at AccuWeather Inc.
It certainly got Uccellini’s attention. Uccellini, fresh from his doctoral studies at the University of Wisconsin, had moved to Columbia, Md., six months before the blizzard to take a job at the Goddard Space Flight Center’s Laboratory for Atmospheres. He has indelible memories of the storm.
On Presidents’ Day eve, a little under five inches of snow had fallen on Washington from a storm that was interacting with Arctic high pressure. The high that day was just 15 in D.C., and the snow resulted from warmer air climbing over the bitter-cold lower levels of the atmosphere. The snow shut off late that night, and the weather service took down its snow advisories.
Uccellini awoke around 5 a.m. to a shocking sight. “I immediately went outside because I had never seen it snow so hard,” he said. “You didn’t even have to put the lights on, it was snowing that hard. It paralyzed everything.” He was seeing as much snow in an hour as the area had received all day the day before.
In Suitland, Henricksen and his team had to play catch-up. The black-and-white satellite images were stunning. Off the coast of Salisbury, Md., they showed an actual “eye,” similar to that of a hurricane, around the center of the furious circulation. By 4 a.m. they increased the forecasted accumulation to call for two feet in Washington, but by then the region was already at a standstill.
The forecasters had missed the offshore explosive development as the storm encountered relatively warm waters. They were relying on computer modeling that at that time was somewhere between infancy and pre-K. “It did not handle the deepening,” Dombek said; the storm “went nuts.” The models’ flaws were well known: “There were certain things that always had problems,” Dombek said.
One major factor was an ongoing Achilles’ heel for modeling, said Jim Eberwine, former marine forecaster for the weather service office in Mount Holly. He was in the Atlantic City area, where 17 inches fell, at the time. Computer models rely on observations to capture the “initial state” of the atmosphere. Using equations for the laws of motion, they look at how systems evolved in the previous six hours to calculate what will happen in the next six.
But the world and its 10-mile deep atmosphere are imperfectly observed, particularly over the oceans, and that was an issue in February 1979. “There was a data void area off the coast that even exists today at the surface," said Eberwine.
Uccellini noted that since 1979, the quality and coverage of observations have become far better, but they still aren’t perfect.
“We can now rely on remote sensing,” said Eberwine. “It helps some, but there is nothing like ground-sea truth.”
Unquestionably, computer modeling has undergone quantum improvements since what the meteorological community affectionately calls PDI, to distinguish it from PDII, the Presidents’ Day storm of 2003, when Philadelphia was shut down under 18 inches of snow.
Warnings for PDII were issued well in advance, said Uccellini, thanks to global forecast models that have shown considerable success at seeing storms several days before they actually form. Model input is available from Canada, the European Union, the United Kingdom, and elsewhere.
Meteorologists can run “ensembles,” model runs that are tweaked to account for weaknesses, such as imperfect observations. That helps them arrive at likely solutions.
One side effect, however, is that many forecasters have complained about a certain paralysis of having too many possible solutions. It’s “chaos,” said Henricksen.
In the 21st century, for the most part, the mid-Atlantic and Northeast have been spared surprise storms, but they have had issues with snow amounts that were over-forecast, last March’s storm being a prime example.
Dombek said that, in 1979, simplicity had advantages, if you knew how to make adjustments. At least, he said, “you weren’t over-modeled to death back then.”