How Hurricane Forecasting Got So Good

As Hurricane Sandy approached New York and New Jersey last week, schools were closed, subways shut down, and low-lying coastal areas evacuated. The storm preparations were aided by what turned out to be remarkably accurate forecasts from the National Hurricane Center (NHC), which had warned residents living in Sandy's warpath as early as five days before the storm arrived. By accurately predicting the track and intensity of the storm, the forecasting may have saved lives as well as millions of dollars.

Superstorm Sandy was more complicated than a standard hurricane. Under normal circumstances, a hurricane like Sandy would have blown out to sea without giving the Northeast so much as a dirty look. But a high pressure area over Greenland steered Sandy back toward shore, where she crashed into a cold front. Despite all of that, the NHC managed to predict Sandy's track to within 50 miles and two days ahead of time. (For comparison, typical two-day hurricane forecasts are off by 100 miles on average.)

Two-day forecasts also predicted wind gusts up to 80 mph, 12 inches of rainfall, and up to 11 feet of inundation from the storm surge?predictions that almost exactly aligned with last week's storm.

"What you saw for Sandy was, I would say, the best weather forecast of a major weather system ever," says Robert Gall, who directs the Hurricane Forecast Improvement Project for the National Oceanic and Atmospheric Administration (NOAA). "I've never seen anything that accurate that far in advance."

How to Model a Hurricane

Hurricane forecasting begins with lots and lots of data. More data than weather modelers know what to do with, really. Satellites collect information about the hurricane's position, wind movement, and the atmosphere's temperature and moisture levels. Aircraft fly into the storm, collecting barometric pressure and wind intensity data. The planes drop probes that measure a variety of features, including the ocean's salinity and temperature profile as depth increases. Buoys and floats collect data about ocean currents, waves, and important interactions between the sea and the atmosphere. Meanwhile, land-based radar learns about hurricane wind fields, rain intensity, and storm movement.

Much of that data is analyzed and assimilated into a computer-generated numerical prediction model. The NHC bases its forecasts on models from all over the world?including from the Global Forecast System, the European Centre for Medium-Range Weather Forecasts, and NOAA's Geophysical Fluid Dynamics Laboratory. The models digest data differently and may use different equations to simulate how the atmospheric and oceanic conditions will change over time. Forecasters at the NHC have a pretty good idea about which models tend to perform better, and they're able to correct for the biases of each based on its past performance. They take an average of all of the models, which tends to compensate for biases in each individual model, says NHC meteorologist James Franklin.

How Sandy Forecasts Were So Accurate

"A lot of things came together this year," Gall says. "One thing we're doing is running the models at higher resolution. We're able to do that because we've got more computing power." He explains that a typical model might use a grid that consists of 1 million east?west points by 1 million north?south points by 100 vertical points. That's a lot of points (100 trillion, in fact). And at each point the models have to calculate temperature, pressure, and humidity changes at regular intervals?usually every 30 seconds. "So it takes a huge amount of computing to generate a five-to-seven day forecast," Gall says.

The NHC's ability to predict the track of a hurricane has been getting dramatically better in recent decades, thanks in part to better satellite data. In the 1970s five-day hurricane forecasts were off by 400 miles on average. In the 1990s it was down to about 200 miles. For 2010 and 2011 five-day forecasts erred by an average of just 100 miles.

It's relatively easy to forecast a hurricane's path, Franklin says, because "all hurricanes get steered by the features that surround them. When those features tend to be large, then the models do a much better job." For Sandy, the high-pressure block off Greenland, as well as the wintry front blowing in off the land, made life miserable for people in the storm's path but actually helped meteorologists to predict Sandy's track more accurately.

"The other thing that made things easier," Franklin says, "was that Sandy was not a purely tropical storm." Typical hurricanes are smaller than last week's superstorm, and the models do better with larger systems. As it moved inland, Sandy also began to take on more wintry characteristics which are easier to model, Franklin says. The reason is that hurricanes are more driven by the ocean, whereas winter storms depend on what's going on in the atmosphere, and hurricane models are stronger at modeling conditions in the atmosphere than in the oceans.

Room for Improvement

Marshall Shepherd, president of the American Meteorological Society, says it is widely recognized that U.S. weather models are not as advanced as European models. (European weather models predicted Sandy's path several days before U.S. models caught up.) Shepherd says the lag is partially because Europeans have invested in faster computers. NOAA's National Centers for Environmental Prediction already have two supercomputers dedicated to weather prediction, but Robert Gall says they need more computing power to improve hurricane prediction. In Boulder, Colo., he's been running experimental models on a supercomputer that's 10 times more powerful than NOAA's.

Having more computing power would allow meteorologists to incorporate more data into the models?currently much of the data that's collected gets thrown away to simplify things. Faster computers would also increase the models' resolution. As of now, the finest-scale hurricane models in operation have a resolution of about 2 miles, Franklin says. "That's getting to the point where we can start to depict some of the features in and around the core, [but] just barely."

A second area for improvement is in predicting the intensity of a hurricane. Intensity forecasts haven't significantly improved in several decades, because hurricane intensity is derived from its interactions with the oceans, and the ocean physics are less well understood in the models. Intensity also depends on thunderstorm activity in the core?higher resolutions are needed to replicate and predict those small-scale processes.

One promising development is the possibility of using Doppler radar observations in the core, and incorporating those observations into operational models. There again, this requires higher-resolution models and more computing power. Still, Franklin says, we might start to see this incorporated into operational models within the next few years.

For now meteorologists are better than the computers at analyzing the activity within the core and using that information to predict a hurricane's intensity, says Franklin?the models haven't learned to interpret those patterns very well. "There are some things forecasters can see that the models just don't get."

Source: http://www.popularmechanics.com/science/environment/natural-disasters/how-hurricane-forecasting-got-so-good-14544602?src=rss

ncaa tournament 2012 peyton manning 49ers andy pettitte tyler clementi kevin kolb sarah shahi george clooney