These computer models from Oct. 26 of then-Hurricane Sandy show different predictions for the storm's path.
These computer models from Oct. 26 of then-Hurricane Sandy show different predictions for the storm's path. NOAA
Better satellites, smarter computer models and faster computers helped government forecasters correctly predict the devastation from Hurricane Sandy, scientists say.
It's unlikely the forecast would have been nearly as accurate just a couple of decades ago, they say.
"The National Hurricane Center did a fantastic job, particularly with the track forecast and the intensity forecast as it was moving toward the Northeast," says Sharan Majumdar, an associate professor of meteorology and physical oceanography at the University of Miami.
Days before the storm hit, the National Hurricane Center was warning people about the power and reach of the storm and the potential for an extreme storm surge in New York City.
"We're going to have a very, very large area that's affected by strong winds, storm surge, heavy rainfall, and inland flooding, and, in fact, even snowfall," forecaster James Franklin said in one of the center's numerous broadcasts to the public.
The accuracy of those forecasts was especially impressive because Sandy was a very odd storm. It turned left when most hurricanes turn right. It maintained its strength even as it struck land. And ultimately, it joined forces with a winter storm.
One reason the forecasts were so good is that the models used to predict weather around the globe are more accurate now and run on much faster computers than they used to, Majumdar says.
"What that means is that we can run the global models at a much higher resolution," he says. "So in the same way that our TV sets are going more high-def, computer models of the globe are also becoming more high-def."
And higher definition leads to higher accuracy. A week before Sandy arrived, one of the major computer models was already showing that Sandy would make a sharp left turn toward New Jersey.
Collecting Better Data
All the computer models now get more and better data than they used to, Majumdar says. One source of these data is dropsondes, capsules dropped from an airplane into a hurricane. As they fall, they transmit a constant stream of weather data.
"The dropsonde data tend to improve hurricane forecasts by up to 30 percent," he says. "So there's been some real significant improvements to the track forecasts over the last 15 years or so through the deployment of these dropsondes."
Government scientists already use more dropsondes than they did just a few years ago. But during Sandy, they put extra dropsondes into the storm to help resolve early differences in their forecast models.
Satellites have also played a big role in improving forecasts, says Tom Renkevens from the National Oceanic and Atmospheric Administration. During Sandy, he says, satellites that sit over the equator were churning out high-quality pictures.
"We were taking images every 7 1/2 minutes continuously through the storm," he says. In the 1990s satellites took 30 minutes to produce an image, he says.
Improvements to special satellites that orbit the Earth's poles also helped forecast Sandy, Renkevens says. He says these satellites report things like sea surface temperatures and humidity in the atmosphere at different altitudes.
"It's the information from these satellites that really is what drives a lot of the forecast models and the hurricane models," Renkevens says.
All these technological improvements allowed the Hurricane Center to predict an extraordinary event in Sandy's brief life, scientists say.
As the hurricane's rotating winds moved north, they began to approach the rotating winds of a cold winter storm, says Adam Sobel of Columbia University. "As the two got close to each other they both got caught up in the other's circulation," he says.
This complicated interaction helped steer Sandy sharply to the left and transform it into a monstrous winter storm, Sobel says.
It's remarkable that the Hurricane Center was able to see this coming several days before it happened, he says, adding that this probably would have been impossible just a couple of decades ago.
"The models could have simulated it in some sense," he says. "But we wouldn't have seen a forecast being accurate that far ahead of time."