(This post has been updated; please scroll down for the latest.)

Despite some slight weakening over the last few hours, Hurricane Irene is being called the “storm of a lifetime” — on its current path, it will affect over 65 million people in the Mid-Atlantic and Northeastern U.S. over the next 72 hours. Packing winds of 100 mph (as of this writing) and ocean waves in excess of 25 feet near its center, Irene looks every bit as ferocious as it sounds — and forecasters are warning of destructive flooding and wind damage.

And as those of us on the East coast prepare for the storm, at a cost of millions of dollars — there are now mandatory evacuations all along the coastline, flights are being cancelled, utility companies are deploying extra crews, etc. — there’s one question some are wondering:

Just how accurate is that forecast?

Our colleagues at IEEE Spectrum have penned a very timely story describing hurricane forecasting — driven by a variety of computational approaches and tools — and how it’s improved significantly in just the last five to 10 years:

Hurricane prediction is by no means a perfect science. Supercomputers crunch vast amounts of data to model a hurricane’s path, taking information from sensors around the world, including balloons, aircraft, ships, and satellites, but it’s a patchwork collection at best. And there’s a practical limit to how much data the models can consider—a simulation that takes more time to calculate than the storm takes to advance isn’t very practical.

 

That’s why hurricane forecasts aren’t perfect. But they keep getting better. Five years ago the two-to-three-day forecast of Hurricane Rita’s path showed it hitting Houston; hundreds of thousands of people fled in a chaotic evacuation that turned out not to be necessary—Rita missed Houston entirely.

 

But today, the 48-hour forecast of a hurricane’s track is as good as the 24-hour forecast was 10 years ago—and that’s pretty good. Researchers are also starting to get a handle on predicting the intensity of hurricanes—that’s something they had very little skill at ten years ago. Operational models, the ones used for day-to-day weather forecasting have gotten better; research models are taking in data they’ve heretofore only dreamed of getting. And all this prediction capability is being thrown at Hurricane Irene.

One key advance:

New numerical forecast systems that started being tested this summer on supercomputers in Boulder, Colo., run fast enough to use in real time, explained Robert Gall, head of the developmental test bed for the University Corporation for Atmospheric Research (UCAR). So, with Irene, as many as half a dozen research models are being combined with the current operational models to zero in on the most likely track, says Frank D. Marks Jr., Director of the Hurricane Research Division at the National Oceanic and Atmospheric Administration’s (NOAA) Atlantic Oceanographic [and] Meteorological Laboratory.

 

And the models have access to more information than ever before. While NOAA has for years sent planes into hurricanes to take measurements, until recently they only used those planes to provide data about current conditions; but now the data collected by the planes is going immediately into the models that provide active hurricane forecasts. When I talked to Marks, he was about to leave on one of those flights, getting ready to collect data on temperature, wind speed, air pressure, and other parameters. More significantly, the plane he was due to travel on was going to be carrying a Doppler radar, which provides valuable insights into storm development, Marks said. This Doppler radar data, Marks says, “is showing the most promise with helping us make models better. We can use it as a benchmark to judge the satellite data we use, because the satellites can’t see through clouds or rain.”

 

Putting all these tools together means that the National Hurricane Center’s forecasts, at least those only two to three days out, are likely to prove true for most hurricanes these days, but particularly for Irene.

 

“Irene’s been a good storm for predictability,” says Marks. Global weather models spotted Irene’s formation nearly two weeks ago, he said, and predicted that it would cause trouble around Haiti and Puerto Rico. Irene, Marks said, is a “classic behemoth,” easier to forecast than scrappy faster forming storms like Cindy, Harvey, and Gert, he said, that wandered all over the place.

Check out the entire article in the IEEE Spectrum, including links to more info about the computing research — modeling, algorithm development, data collection and mining, machine learning, etc. — at the center of this work. Another great resource: NOAA’s Hurricane Research Division Blog which is providing regular updates of the “hurricane hunter” flights.

And of course, for those in the storm’s path, here’s the latest information on Irene from the National Hurricane Center, as well as a guide to hurricane preparedness.

***

Updated Monday, Aug. 29 at 8:46am EDT: As the remnants of Hurricane Irene make their way into the North Atlantic today, it’s worth revisiting this post. Here’s a msnbc.com article that notes, though the storm proved weaker than anticipated, “[On] Friday morning — 24 hours before landfall — [National Hurricane Center forecasters] had the storm’s next day location to within 10 miles or so.”

(Contributed by Erwin Gianchandani, CCC Director)