wplg logo

How good are the models at predicting hurricanes? Experts try to make them better

At the National Hurricane Center in Miami, experts are working to make computer forecast models better at predicting storm strength.
At the National Hurricane Center in Miami, experts are working to make computer forecast models better at predicting storm strength.

PEMBROKE PARK, Fla. – Once a tropical system develops, people spend a lot of time looking at the computer forecast models, trying to see where the storm might head.

On TV, we often compare the American model and the Global Forecast System (GFS) to the European and other models, along with the cone as a way to show the uncertainty that comes with forecasting.

That cone represents the official National Hurricane Center’s manmade forecast.

But how good were they during the intense 2020 season? And were the computer models better?

Local 10 spoke to Dr. Mike Brennan from the National Hurricane Center in Miami.

“I think the NHC forecast was as good or better than all of the model guidance pretty much across the board save for some of the consensus aids for track, but we pretty handily beat a lot of the intensity guidance last year, which is pretty impressive in a challenging year that had a lot of rapid intensification,” he said.

The top winds in hurricanes Hanna, Laura, Delta, Zeta, Eta and Iota all strengthened by at least 35 mph in a 24-hour period.

Hurricanes Eta and Iota both intensified by 80 mph in a day, and although Eta just brushed South Florida while causing a big flood, both storms made landfall in Nicaragua, causing widespread destruction across that country and Honduras.

Neither the computers nor humans are good at forecasting these storms that suddenly get a lot stronger.

Still, the forecasts are better than they seem.

“A lot of the error comes from these rapid intensifying storms where we’re able to predict some degree of even significant strengthening, but if you get the timing wrong even by 6 or 12 hours you can have really large errors even if you get the overall picture right,” Brennan said. “We base at least five years’ worth of errors on them so you try to get a sample across multiple seasons, multiple different types of storms. ... That’s how we devise a cone.”

Brennan notes that the further out they forecast, the bigger the challenge — and the more hope there is science will come through.

“I think there’s still plenty of room for improvement out on days three, four, five — even out to days six to seven where we can still get some of those very large errors,” he said.

Those big errors are just one reason the Hurricane Center is thinking long and hard whether they want to make 6- and 7-day forecasts public, among others.

Just this year, Brennan says four models have or will be upgraded. A big one is the American GFS model.

“The track forecast error statistics are pretty similar to the old model, although we did see an improvement in the shorter-range forecast for weaker systems in the Atlantic with this new model,” Brennan said. “But we’re hopeful that this improvement with the GFS and the new HWRF (Hurricane Weather Research and Forecasting) model, HMON (Hurricanes in a Multi-scale Ocean-coupled Non-hydrostatic model) that are going in this year will have some improvements as well.”

CLICK HERE to view and download our 2021 Hurricane Survival Guide.


About the Author:

Bryan Norcross is currently a hurricane specialist at Local 10 News, the station where he began his stretch on television in Miami in 1983.