NewsMarch 7, 2013

D.C. Snow No-Show A Lesson in Forecasting Uncertainty

Search results placeholder
Andrew Freedman

By Andrew Freedman

The storm known to many as “Snowquester” appeared for a time like it would be one of modern meteorology’s shining moments, in which early warnings saved the government and millions of people time, money, and exposure to dangerous weather conditions. Instead, it turned into every weather forecaster’s worst nightmare — a bust of a forecast for a heavily populated and politically influential region, in which the predicted 5 to 10 inches of snow for Washington, D.C. and Baltimore actually yielded just 0.2 inches of slop.

The extent of snowfall accumulation seen in Northwest Washington on March 6.
Credit: Lauren Morello, Climate Central.

In anticipation of the storm, the government shut down Wednesday, keeping federal workers — well, the ones who had not yet been furloughed by across-the-board budget cuts known as the sequester, at least — off the roads. Schools were closed, appointments were cancelled or rescheduled — yet all of those disruptions, and more, turned out to be for nothing, more like a snow drill.

The failure to accurately predict the storm in the I-95 corridor offers lessons in communicating risk and uncertainty, which can be applied to both weather and climate forecasting.

In meteorology, there is an emerging trend away from so-called “deterministic” forecasting, as represented by the weather maps showing zones of snowfall amounts without any expressions of confidence or percentages, and toward “probabilistic forecasting,” in which consumers of weather information are given clear indications of just how confident forecasters are in a given prediction. For example, a probabilistic snowfall forecast might assign odds to predicted snowfall amounts, thereby allowing forecasters to fully express the range of possible outcomes, rather than being rigidly committed to a particular number. Probabilistic forecasting is, after all, already in use every day in limited ways, the best example of which is when forecasters communicate a “50 percent chance” of rain.

Along the same lines, climate scientists have been struggling for years to devise ways to convey both what is known with high confidence — such as the basic fact that the world is warming largely as a result of manmade emissions of greenhouse gases — and what is less clear cut, such as exactly how precipitation patterns will change in the Corn Belt region of the U.S.

Near Unanimous Forecast Guidance

On Tuesday night, before the snow began, forecasters were faced with what appeared at first to be a clear-cut forecast, but in reality was a plethora of slightly mistaken projections from high-tech computer models.

Nearly every computer model used to guide weather forecasts indicated that significant precipitation would fall, and fall heavily enough to be in the form of snow, thereby overcoming surface temperatures that would be above freezing, at least at the outset of the storm. That caused weather forecasters to think that the high March sun angle, which can make it extremely difficult to get accumulating snow during the daylight hours, could be overcome, and to not worry too much about the potential for the low pressure system to intensify before it moved to the east of Washington, leaving the city with a relatively warm flow of air off the Atlantic Ocean.

The Washington Post’s Capital Weather Gang, which nicknamed the storm “Snowquester” and is one of the city's leading weather information providers, forecasted a 5 to 10 inch snowfall along the I-95 corridor, to which it assigned “medium confidence.” Television weathercasters such as Bob Ryan of WJLA-TV forecast similar amounts, after seeing the same information streaming in from the computer model simulations.

The “Snowquester” storm seen from space, swirling out in the Nortwest Atlantic, off the New England coast.
Click on the image to enlarge. Credit: NASA.

“When all the guidance points one way and it is toward a major event, I don’t think you have much choice but to go with the guidance, though we may not have given enough weight to the March Sun,” said Wes Junker, a veteran forecaster and a winter weather expert for the Weather Gang.

On Thursday, Ryan, the Weather Gang, and other forecasters were in damage-control mode, explaining to the public why their forecasts didn’t pan out.

“I thought the snow forming above us (all the rain that fell was snow until 1,000 to 2,000 feet up) would be intense enough to 'beat' the mild air near the surface. Even with a mild east wind,  the morning rain around D.C. should have changed to snow. Wrong,” Ryan said in a blog post on his station’s website. 

In an interview, Ryan said that in hindsight the biggest red flag should have been the lack of cold air to the north of the Washington area, which would have been drawn into the storm to ensure snowfall instead of a slushy mix of snow and rain. He noted that for areas further west and southwest of D.C., the forecast was on target, as 6 inches or more fell across northern and western Virginia and western Maryland.

Jason Samenow, chief meteorologist of the Capital Weather Gang, wrote a lengthy mea culpa in which he not only laid out the reasons why the forecast failed, but also took a look at what this event might suggest for future forecasts.

“Given the problems with the modeling as well as the more intuitive reasons to be skeptical of snow prospects, the most critical error we and every single forecaster in this region made was the failure to communicate the uncertainty in the forecast,” Samenow said.

“We actually did this reasonably well 2-3 days prior to the storm,” Samenow said. “For example, on Sunday, we presented a 'lower probability' scenario (given a 30 percent chance) in which 'any snowfall accumulations around the city would be limited to an inch or so on grassy surfaces' and we displayed a complementing graphic. On Monday, we re-iterated there was bust potential . . . But we watered down some of this uncertainty information the day before the storm — a major communication error.”

That illustrates the challenge that weather forecasters have when it comes to clearly communicating uncertain information to the public, an issue that the climate science communities are constantly wrestling with.

“One of the reasons, as we get closer to the onset of the storm, that we drop some uncertainty information is that some readers want to know the bottom line, without qualification. They view scenarios and percentages as 'cop-outs,' ” Samenow said. “Ultimately, there has to be a sweet spot, where we can effectively communicate uncertainty concisely and effectively while also presenting a most likely forecast.”

Marshall Shepherd, a professor of atmospheric sciences at the University of Georgia and president of the American Meteorological Society said that computer models, while advanced in many ways, still are wrestling with an atmosphere that is a “non-linear, chaotic system,” which means there will always be uncertainties involved in their projections.

The trick for forecasters of weather and climate alike is to devise ways to convey such uncertainties without making it seem like they have no idea what the weather, or climate, will be like. After all, anyone can just make guesses and be right once in awhile. Forecasters are the ones who are supposed to have the expertise that others lack.

“With the increased capabilities in forecasting, maybe there are cases (where) the weather community portrays certain scenarios too concretely — even while noting that the confidence is not high at the same time. Most weather consumers look at a map with ranges and generally expect those ranges without necessarily reading the fine print,” said Ian Livingston, a forecaster for the Weather Gang and a researcher at a D.C. think tank. 

Hopefully, forecasters can learn the lessons from this event and implement them quickly. In Boston on Thursday night through Friday morning, snow, or rain and snow, is in the offing from that city's close encounter with the “Snowquester” storm, with anywhere between 1 and 10 inches predicted to fall, depending on the finicky surface temperatures.

Related Content
Coastal Flooding, High Winds Battering New Englandtorm Spares D.C., Pounds Coast With Wind, Water
'Snowquester' Threatens to Cause Major Coastal Flooding
The Latest on the Developing March Snowstorm
Myriad Impacts Loom as Intense Snowstorm Moves In
High-Impact Snowstorm Headed for Mid-Atlantic
Lack of Warning on Drought Reflects Forecasting Flaws
Scientists Seek Insights Into Outlier Drought Predictions