As the monster storm Irma approached, some people in South Florida fled to the Gulf Coast, hoping to be safe there. Instead, after clobbering the Keys, Irma followed them, wobbling to the east and taking a course that came right up the state's western edge with a landfall at Marco Island.
Did the National Hurricane Center get the forecast wrong?
"An analysis of the track forecast and intensity error of the storm won't be done for a few weeks, at the earliest," Hurricane Center spokesman Dennis Feltgen said Monday.
But a trio of weather and climate experts say the forecasters did pretty well, given the size of the sprawling storm and the 120-mile width of the Florida peninsula.
"South Florida isn't that wide," pointed out Phil Klotzbach, a research scientist with Colorado State University. "The typical forecast margin of error is going to be within 100 miles."
In order to give people plenty of time to evacuate, Klotzbach said, forecasts often have to be made several days before anyone is certain where the storm will go. That's bound to lead to some evacuations that turn out to be unnecessary, he said.
For everyone who believes he or she didn't really need to evacuate, Klotzbach quoted former National Hurricane Center director Neil Frank: "You have to evacuate three times for every time you really need to evacuate."
Sharanya Majumdar, a professor of atmospheric science at the University of Miami, pointed out that Irma's final track was within the hurricane center's "forecast cone," more commonly known as "the cone of uncertainty."
Even before the storm crossed Cuba, some models showed the storm taking the path it eventually took and losing steam as the hurricane left the warm water that fuels its power. That would mean the powerful Category 5 storm everyone feared would be a tiger would be a far tamer pussycat, a 2 or even a 1, by the time it reached Tampa Bay.
People who paid attention to the line instead of the cone were the ones who were fooled, Majumdar said.
"There's too much focus on the line of the storm, and not enough on the uncertainty of the storm," he said. "There needs to be more emphasis on the size of a storm like Irma," he added, noting that it covered the whole peninsula.
Also, he said, Floridians should learn that there's danger from more than just the winds in a storm's eyewall. He pointed to the flooding in Jacksonville.
If he had been in charge, he would have told people on the Gulf Coast to evacuate at the same time as the people on the east coast, he said. But he found no fault in the forecasting prowess of the hurricane center.
"I think it was handled very well," he said.
The Hurricane Center's ability to see into the future has been questioned before, most strongly after Hurricane Charley became the first of four hurricanes to clobber Florida in 2004.
Originally, the hurricane was predicted to hit Tampa Bay head-on, the first such hurricane to slam into the region since 1921. Instead, at the last minute, it made an abrupt turn and slammed into Punta Gorda, then cut a swath of destruction across Central Florida.
Charley's abrupt change in direction wasn't the only surprise. The storm also intensified very rapidly from a weak Category 1 hurricane to a strong Category 4 within a few hours.
In reviewing what happened, the National Weather Service in a 2006 report found that the Punta Gorda landfall had been within the cone of uncertainty.
"Many people focused on the specific forecast track which indicated the projected path of the center of Hurricane Charley making landfall near Tampa Bay instead of the cone of uncertainty," the report noted.
The report said that a social scientist from Florida International University interviewed 100 people where the storm made landfall and found that more than 90 percent thought Charley was supposed to hit Tampa.
The report suggested the agency spend more time trying to educate people about what the cone means. But the changing demographics of a growing state make that difficult.
The report did note that the change in Charley's intensity was something that would have been hard to foresee.
"It is unusual for hurricanes to intensify rapidly just before making landfall — they usually weaken — and it is rare for hurricanes to make landfall as a category 4 or 5," it said.
As Florida's official state climatologist, Florida State professor David Zierden isn't an expert on day-to-day forecasting, but he does consider himself "a weather nerd" who knows a few things about how it works.
What Irma did "was well within the forecast cone put out by the National Hurricane Center," Zierden said. "As storms like this are making landfall, these kinds of developments can happen."
When a storm is headed for land at an oblique angle, "even a slight deviation can make a big difference," and can be hard to foresee three or four days in advance, he said.
The bottom line is that the amount of notice required to move millions of people conflicts with the forecasters' ability to gauge what a bunch of swirling wind might do next.
"The science is good and it's getting better," Klotzbach said. "But it's not perfect."
Contact Craig Pittman at email@example.com. Follow @craigtimes.