Louis Uccellini, head of environmental prediction for the National Oceanic and Atmospheric Administration, explains the unique challenges of tracking Hurricane Sandy.
Update: 6:15 p.m. ET | As Superstorm Sandy barreled its way west from the Caribbean to the Mid-Atlantic states and then swiveled north up to Canada, forecasters scrambled to understand it. The storm was unusual. It arrived late in the season and was fed by unseasonably warm waters. It was pushed inland by a high-pressure system and then merged with a cold front moving east from California. Add to all that a full moon and resulting high tide, and you have a real forecasting challenge.
At the National Center for Environmental Prediction’s National Hurricane Center in College Park, Maryland, scientists gathered global data from weather balloons, satellites, commercial airplanes, buoys at sea and weather stations. That data gets fed into a supercomputer, which uses differential equations to create a model that predicts the track and intensity of the storm. Every six hours, the center comes up with possible weather outcomes, which scientists use to make a forecast.
Such forecasting takes a tremendous amount of computing power. The supercomputer at NCEP processes 74 trillion calculations per second. It would take the computing power of 10,048 Intel processors to do the same. The resulting model creates picture of the atmosphere that forecasts the weather as far as two weeks into the future and provides hundreds of possible weather outcomes. In the video above, Louis Uccellini, head of environmental prediction for the National Oceanic and Atmospheric Administration, explains the unique challenge Sandy presented for forecasters and how those models carry life-saving information for emergency personnel.
Of course, weather doesn’t always cooperate with the models, said Kerry Emanuel, professor of meteorology at the Massachusetts Institute of Technology. Building forecast models requires tracking weather conditions thousands of miles out. In order to predict next week’s weather in Boston, forecasters must know today’s atmospheric conditions in Hawaii, he said.
And forecasters are also at the mercy of chaos, or something better known as “the butterfly effect,” which, simplified, means that slight changes in the atmosphere can make enormous differences over time, Emanuel said.
Every day, meteorologists at NCEP collect data from 200 worldwide models, and then, based on past storms, decide which outcomes are most likely to occur. But sometimes the models vary widely. Five days before post-tropical storm Sandy made landfall in the United States, a U.S. Global Forecast System model predicted that the storm would move out to sea, while another model from the European Centre for Medium-Range Weather Forecasts showed Sandy’s path colliding with the New Jersey shore. When Sandy made landfall, it did so by slamming into the New Jersey coastline, causing flooding, fires, snuffing out power for millions and resulting in more than 100 deaths. In other words, the early U.S. model was wrong.
“The European center had pretty much pegged it, and we were literally out to sea,” Emanuel said. The U.S. model relies on a slower supercomputer, he explained, which means it’s consistently less accurate than the European one. It’s ironic, he said, considering it was the U.S. that invented numerical weather modeling in 1950s.
But better forecasting isn’t just about computing power. Other factors can impair observations, said Jeff Masters, meteorologist at Weather Underground. Damaged equipment sensors, for example, might send back inaccurate information. Weather balloons fly 150 miles apart and only twice a day, and NOAA satellites’ sensors lack the ability to see through clouds. Bottom line: the more observations taken, the better the predictions, Masters said.
But that requires more investment. Funding for NOAA’s weather monitoring, including new weather monitoring satellites, would be cut under a proposed Republican budget.
“The bang for the buck we get for investing in forecasting is huge. Like other things in government, it’s being put on chopping block,” Masters said.
For example, one of the space weather satellites that monitors wind is broken and in need of replacement. That would cost $250 million dollars, Masters said, and there simply hasn’t been the funding to replace it.
The challenge among forecasters is communicating the uncertainty inherent in predicting the weather to journalists and the public.
“One of the great unsung advances in numerical prediction is we got very good at quantifying uncertainty in any forecast,” Emanuel said. “The communications problem could be solved by systematic communication between scientists and journalists to keep out hype but keep in the notion of risk.”
Still, modeling has improved significantly over the last 10 years. Despite the damage, the massive storm was considered a forecasting success, said Mike Smith, meteorologist at Accuweather and author of “Warning: How Science Tamed the Weather.”
“If Sandy had hit without warning, which is distinct possibility even 10 years ago when the high resolution models used today didn’t exist, we would be measuring death tolls in the thousands. People died, and that’s a horrible tragedy, but it was so much less than it could have been,” he said.