Winter Storm Juno: Did We Get It Wrong?

Nick Wiltgen
Published: January 27,2015




 

The Great New York City Blizzard of 2015.
It didn’t happen. As millions of New Yorkers – and Philadelphians and New Jerseyans – woke up Tuesday morning, it was clear that Winter Storm Juno’s impacts were far from historic for their neighborhoods. Yet, not far away, some New Englanders had seawater and ice floes in their streets, while others struggled to pick out backyard sheds from the snowdrifts encasing them.
For much of Sunday and Monday, The Weather Channel predicted 12 to 18 inches of snow for New York City, while the National Weather Service called for 20 to 30 inches there.

Forecast snowfall map for Winter Storm Juno as it appeared on weather.com at 3:22 p.m. EST on Sunday, Jan. 25, 2015. The National Weather Service issued a blizzard warning for New York City at 3:19 p.m.

Snowfall totals in the five boroughs ranged from as low as 6 inches to as high as 12.1 inches in Glendale, Queens. The official total at Central Park: 9.8 inches.
So what happened? Did we – the meteorologists charged with predicting storms like this – get it wrong?
Dr. Louis Uccellini, Director of the National Weather Service, didn’t go quite that far. But in a teleconference Tuesday afternoon he did say, “We recognize the need to work harder and smarter to produce better forecasts and to better communicate forecast uncertainty and manage expectations.”
Whether it’s out of genuinely scientific motives or simple self-preservation, we meteorologists tend to see things in shades of gray. Where others see right or wrong, we see percentages and numerical skill scores.
Still, there’s a point at which everyone, meteorologists included, can agree that a forecast went wrong. As Supreme Court Justice Potter Stewart famously put it in a 1964 decision about obscenity: “I know it when I see it.”
It’s fairly plain to see that New York City did not get a blizzard. What is perhaps not as plain is that we knew the forecast could go haywire.

What was the forecast?

The National Weather Service and all major media outlets were calling for a major snowstorm in New York City.
When the NWS issued a blizzard watch for the five boroughs just before 4 a.m. Sunday, its forecast called for 12 to 18 inches of snow in New York City. The Weather Channel forecast agreed.
But at 3:19 p.m. Sunday, NWS issued a blizzard warning, calling for a “crippling and potentially historic blizzard” with 20 to 30 inches of snow, “locally higher amounts possible” and north winds of 30 to 40 mph with gusts to 55 to 65 mph, “strongest across eastern Long Island.”
NWS stayed with those numbers through late Monday night, except for a nine-hour period Monday morning when it called for 18 to 24 inches and locally higher amounts.
The Weather Channel stuck with the 12-to-18-inch forecast, except for a brief period Sunday morning when we upped New York City into the 18-to-24-inch range, based on raw forecast numbers of 19 inches at Central Park and 20 inches for Queens.

Our forecast snowfall for Central Park in New York City is seen on this weather.com graphic as it appeared at 9:19 a.m. EST, Monday, Jan. 26, 2015. The 15.1-inch raw number translated into a 12-to-18-inch range in on-air graphics. The actual storm total for Winter Storm Juno was 9.8 inches.

Once the storm began to crank up offshore, it became more evident that the zone of heaviest snow might be setting up a bit farther east than expected. The Weather Channel snow forecasts dropped into the 8-to-12-inch range Monday afternoon. As some of the future snow became past snowfall, we trimmed our forecasts accordingly during the storm to essentially keep the storm total near or just under the one-foot mark.
Throughout the storm, our on-air experts pointed out that a slight eastward shift in Juno’s track could result in a significant change in the outcome for the Big Apple.

Just how wrong were we?

We did forecast a blizzard for New York. So did everyone else that we know of. And we were all wrong about that.
But we weren’t wrong by far. Some 50 miles east of the Queens/Nassau border, many Long Island communities were pummeled by 40-mph wind gusts and near-whiteout conditions for hours – likely meeting blizzard criteria.
There are no snowfall criteria for blizzards, but those same areas were also buried by more than 20 inches of snow. So for eastern Long Island (not to mention Boston, Providence, Cape Cod, and many other areas), everyone got this forecast very, very right.
New York City did not meet blizzard criteria for winds and visibility. And even our forecast, which was among the most conservative among major outlets, overplayed the snow for most of the runup to the storm. By the time we lowered our forecast into the ultimately correct 8-to-12-inch range, much of the region was already preparing for a total shutdown of traffic and commerce.

Beaten by the snow band

Snowfall forecasting involves the intersection of many variables on scales from local to global: atmospheric pressure patterns, wind directions, air temperatures at various levels of the atmosphere, and the amount of moisture in the air.
The tremendous computing power of today’s forecast models allows us to predict these interactions with astonishing skill days in advance – a huge advance in a short time. “When I started my career, these types of storms were at times not even predicted a day in advance,” Dr. Uccellini said Tuesday.
But the complexity of the atmosphere is still greater than our ability to fully comprehend, measure, analyze and predict its behavior.
"Winter storms almost always exhibit sharp edges, or gradients, in snowfall,” says Jonathan Erdman, one of my fellow digital meteorologists at The Weather Channel. "In Juno's case, the gradient in snowfall was predicted to lie right over the most populous metro area in the country. If you were in central and eastern Long Island, Juno probably delivered what you expected. If you're in the five boroughs, however, perhaps the storm didn't match what you had been hearing."
Predicting with confidence exactly where these gradients will set up is still beyond the capability of our science. Forecasters recognize that if the edge of the heaviest snow shifts by 20 or 30 miles that it can mean a huge difference in the resulting accumulation for people near that edge. How is a meteorologist supposed to handle this?
One solution is to smooth out the sharp edges of the storm’s forecast snowfall footprint by essentially blending various contradictory computer model scenarios. In that case, the goal is to get the snowfall amounts as right as possible – or to be blunt, the least wrong – for the largest area, knowing that the forecast is probably not going to be correct for everybody. Obviously, that strategy is complicated when that area includes millions of people.
Another solution is for a forecaster to choose one scenario that seems most likely to happen, and place all of his or her chips on that scenario, hoping to get the forecast right for everybody. In the case of potentially extreme storms like Juno, the consequences of choosing the wrong scenario can be embarrassing and costly – especially when the forecaster bets high and reality comes out low, or vice versa, for a major metropolitan area.

Is there a different way to do this?

Forecast snowfall guidance from the Short Range Ensemble Forecast model, run by the U.S. government's National Centers for Environmental Prediction at 10 a.m. EST Monday, Jan. 26, 2015 for LaGuardia Airport in New York. The colored lines represent 22 model forecasts of snowfall accumulation over an 84-hour period ending 1 a.m. Thursday, Jan. 30. The forecasts ranged from 4.01 to 39.97 inches. The black line is the average of the 22 models, which was about 18 inches. The final storm total at LaGuardia was 11.4 inches.
(NOAA Storm Prediction Center)

Forecast snowfall guidance from the Short Range Ensemble Forecast model, run by the U.S. government's National Centers for Environmental Prediction at 10 a.m. EST Monday, Jan. 26, 2015 for Logan International Airport in Boston. The colored lines represent 22 model forecasts of snowfall accumulation over an 84-hour period ending 1 a.m. Thursday, Jan. 30. The forecasts ranged from 19.51 to 38.87 inches. The black line is the average of the 22 models, which was about 31 inches for the whole storm and 28.5 inches through 7 p.m. Tuesday. As of 7 p.m. Tuesday, Logan reported 23.3 inches with snow still falling.
(NOAA Storm Prediction Center)

One possibility would be to broaden the ranges of snowfall we typically use in our forecasts. For example, instead of predicting 12 to 18 inches for New York, we could have predicted 4 to 18 inches. But that introduces another set of problems, according to The Weather Channel winter weather expert Tom Niziol.
"The trouble with forecasting 4 to 18 inches is that to the snowplow driver – actually, to anyone impacted by the snowfall – there's a huge difference between 4 inches and 18 inches,” Niziol told me. “Large snowfall ranges like that wouldn't provide value-added information for the user."
A relatively recent approach that is beginning to gain some traction is the use of probability-based forecasts. For instance, NOAA’s Weather Prediction Center regularly issues snowfall forecasts for various confidence levels for the next 24, 48 and 72 hours.
If you can understand their percentage-based forecasts, you can actually glean some interesting information from them. You might have seen, for instance, that on Monday morning the agency showed just a 40 percent probability of New York City receiving at least a foot of snow between 7 a.m. Monday and 7 a.m. Tuesday, reserving the highest probabilities for southeastern Massachusetts.
Getting that kind of information into an easy-to-understand format for the general public is something many meteorologists would like to do. Niziol said there was a pilot program aiming to do that at the National Weather Service office in Buffalo, where he spent much of his career.
But right now, most weather information is presented in what meteorologists call “deterministic forecasts” – it will be sunny, it will be 44 degrees, and so on. And many people believe that a more precise forecast is a more accurate one.
But as the New York Non-Blizzard of 2015 shows, it may be better to more thoroughly acknowledge the limits of our forecasting abilities. Finding a way to clearly communicate the full spectrum of possibilities during a major storm would go a long way toward helping society make better use of the imperfect yet amazingly powerful tools we have to see the tempests heading toward us.
Or maybe toward the next county over from us.