Blizzard of 2016 forecast challenges: what went wrong

Good evening.

I hope all readers in the states and regions being affected by this massive winter storm are handling themselves well tonight. It has been a very remarkable storm that has performed almost as advertised, with massive snow totals over large areas, plus high winds, devastating coastal flooding and freezing rain. Snow totals have been in excess of 20 inches along most of the I-95 corridor with some amounts over 30 inches. Along the Blue Ridge front, amounts in excess of 40 inches have been reported. This update from the Weather Prediction Center (WPC) shows many snowfall amounts.

Many cities, including New York City, Philadelphia, Baltimore and Washington DC, are essentially shut down tonight. Driving bans are in place, mass transit is skeletal or nonexistent and roads are, for the most part, impassable. Along several Interstates including the Pennsylvania Turnpike, hundreds of motorists including buses were stuck in the heavy snow. As a result, everyone in that region should stay at home or stay where they are for the next little while. Let’s leave the roads to cleanup and emergency crews.

Storm total snowfall accumulation map for the region (NWS Eastern Region).

Storm total snowfall accumulation map for the region as of Saturday evening (NWS Eastern Region).

Assessing the forecast

Literally, the final call yesterday gave me headaches. It was a difficult forecast for over 30 million people, particularly in the New York metropolitan area. The more reliable GFS and ECMWF models, along with the similar but less reliable CMC model, all insisted on the heaviest snow reaching no farther north than central New Jersey, and also stopping in Pennsylvania around the Turnpike corridor. However, the NAM and its derivatives like the SREF plumes – traditionally less reliable – believed the storm would reach farther north, into the southern New England coast, and also forecast phenomenal snow amounts in New York City and northern New Jersey.

The difference was huge – between about 4 inches (very manageable, below warning criteria) and over 20 inches (enough to paralyze the region). I opted to go with a general 8 to 12 inches for the region, splitting the difference, higher south and lower north. However, I also highlighted in red that it was low confidence, knowing the NAM was consistent on a historic storm, and tried to highlight that fact as much as possible. Clearly, I (and most forecasters) greatly underestimated the totals, and the region was playing catch-up all day in extreme measures. I haven’t heard of any fatalities from traffic accidents in that region so clearly they paid close attention. Note that just 60 to 70 miles north of NYC, snow amounts were near the forecast, suggesting that the storm was a miss by just that much, but it had enormous implications for the most heavily populated region in North America.

Most of the forecasts in the mid-Atlantic region, outside of the northern periphery, were largely correct or a bit conservative. However, I did mention yesterday that amounts could be much higher than forecast in some spots, and made note that 40 to 50 inches were not out of the question. The greatest difference was in south-central Pennsylvania, near the Turnpike, and western Maryland. In both cases, amounts of 30 to 40 inches were reported in numerous locations and some higher, which was more than expected. Deformation bands are always hard to predict and snow ratios were low there (suggesting fluffy or dendritic snow), enhancing totals.

What could have been done differently?

Normally a forecast that is regionally off due to location of banding is not a big deal. It just happened that this time over 30 million people were in that area, and the stakes were much, much higher than usual. Last year, a sense of over-reaction occurred after New York essentially shut down for a storm that turned out to narrowly miss them. It became clear that care had to be taken to not go overboard. As it was, I was more aggressive than the majority of models. Yet the NAM rang warning bells. Finally, the 00Z models last night started to turn towards heavier snow farther north, including into New England. That was after the whole event already started and significant snow and ice was underway in western areas, meaning it was too late to be a “final call”.

Maybe I should have gone higher for some? I might have gone a tad higher there but it still would have under-estimated the actual amounts. Also I ran the risk of being vilified for crying wolf if that did not verify. No wonder I got headaches writing the final call map. Clearly, though, placing the confidence levels of the forecast was the right move in order to express uncertainty. I always said the potential was there, but the map could not show it. That is something I advise everyone to do – express uncertainty however they can.

As for the high end amounts, the map stopped at 30 inches which was clearly too low in a swath from northern Virginia to east-central Pennsylvania into New Jersey. I actually recommended showing a peak of 36 inches on the map but couldn’t get team agreement. Nonetheless, the text explicitly mentioned higher amounts were possible, possibly 40 to 50 inches. While that is an extremely aggressive forecast which was not made by many other forecasts, it appears to have verified or nearly verified. However, we should not be competing on forecasts – instead should have simple, single messages as much as possible.

I strongly recommend that everyone reads this article by Becky Elliott of AccuWeather that was posted in the Washington Post this week before slamming any meteorologists or forecasters. Even the best, most professional meteorologists make mistakes frequently and miss forecasts sometimes. Dr. Marshall Shepherd, a meteorologist who makes appearances on The Weather Channel’s show WxGeeks and is also a professor at the University of Georgia, also has an excellent article written today.


Although most of the snow has fallen, the impacts continue as storm surge flooding continues to inundate parts of the mid-Atlantic coast, in some areas as bad as or worse than that from Hurricane Sandy in 2012. It will take at least 24 hours just to get the main roads and highways back to safe condition and re-establish just a skeletal mass transit system, plus to open the airports and railroads. Getting everything cleaned up and plowed could take several days, or even longer in rural areas. Please do everyone a favor and stay indoors tonight and until things are safe to travel. You may be restless, but safety for everyone is paramount. Also be careful shoveling – there have already been people who lost their lives from heart attacks from cleaning up.

The Blizzard of 2016 will be remembered for a long time to come. Please stay safe in the aftermath as we all work to get everything back to normal. Make sure you listen to local officials as they work to restore normalcy to many areas from southern New England to the Carolinas.

Forecaster Craig Ceecee – @Ceecee_Wx

One comment

Leave a Reply