Which hurricane forecast model is the best?

A satellite image of Hurricane Earl (left) and Tropical Storm Fiona in the Atlantic Ocean on Aug. 31, 2010.
Getty Images
A satellite image of Hurricane Earl (left) and Tropical Storm Fiona in the Atlantic Ocean on Aug. 31, 2010.

As hurricane season comes upon us on June 1, we will be treated to a variety of hurricane forecasts, with different guesses about when and where landfall will hit, how intense the storm will be, and how much flooding we might expect.

All those forecasts will change each day.

And some of those forecasts will look very different than the rest.

So which forecast is the best? Which one do we trust?

This question matters because it's estimated to cost $1 million in lost economic output for each mile of evacuated coast land. The key for local officials is to minimize the evacuation zone—and a model that forecasts landfall needs to be as close as possible. Being off by 50, 100 or 150 miles is a major problem, and it hurts the ability for the public to trust the models in the future.

Europe takes the crown

According to most people in the industry—or just plain statistics—the European model is the best, and has been for years. Almost any report will describe it as the best. Technically, you want to look for the acronym it goes by—ECMWF—which stands for the European Center for Medium-range Weather Forecasting.

A big factor in its dominance recently is due to a 2006 improvement in its model. The Europe model's advantage comes from several sources: Powerful supercomputers that can analyze larger amounts of data, taxes paid by the member nations of the European Union to help keep funding up, and charging money to other forecasters who want its data.

"They don't have any top secret things," University of Miami meteorologist Brian McNoldy told National Geographic. "Because of their (computer) hardware, they can implement more sophisticated code."

Politics hurts the US

One of the reasons the U.S. suffers is a lack of cooperation between government officials and academics. Another big factor is the way research is funded (and defunded) in the U.S.: by the whims of Congress. This creates instability, and a generally permanent always-catching-up status for the U.S. models. The technology isn't there to keep up with what Europe has to offer.

Known as GFS, for Global Forecast System, the U.S. model might sometimes beat ECMWF, but for the most part it doesn't. "The Euro is the king of the global models," said Greg Nordstrom, a meteorologist at Mississippi State University.

This year is a big one for the U.S. models, as the NOAA announced that its two supercomputers will see a "nearly tenfold increase from the current capacity" by October, which will allow for more specific resolution on its forecasts.

It's always better to aggregate

No matter which model is the single best, human forecasters can find more accurate results by averaging the specific models to see what the overall trends suggest. This is true for prediction models in all sorts of fields, not just hurricanes.

"The more skilled models you have running, the more you know about the possibilities for a hurricane's track," said McNoldy.

Other top forecast models—beyond the ECMWF and GFS—come with strange acronyms as well: GFDL (American), UKMET (English), HWRF (American) and NOGAPS (U.S. Navy). There exist many more modelsdozens, in fact—each with its own strengths and weaknesses, used for different types of tracking and predictions. Remember it's not just about tracking location and timing, but also storm intensity, air and water effects and other environmental factors.

The NOGAPS model has struggled in recent years, and as a result of poor accuracy the National Hurricane Center in 2011 dropped it from its list of "consensus models." NOGAPS wasn't specifically designed to forecast hurricanes and is better used as tracking-only model.

Models have improved over time

As you can see from the chart above, the lines keep going lower each year, suggesting better accuracy over time. The red line is the 24-hour forecast, which will always be better than the blue line (120-hour forecast).

As technology improves, data increases, and the computers get faster, more thorough analysis can occur, for a better assessment of where and when exactly hurricanes will hit land.

But they still aren't that great

Even if the models are off by 50 miles, that's still a big difference for who will be affected once a hurricane hits shore. This is why governments are still investing in trying to improve the models.

There has been some talk that we might soon reach a "prediction horizon," meaning the forecasts just can't get any better, but for now, that doesn't seem to be the case. Each year, the trend is toward better forecasts, and that trend should continue for a while.