It was not the news residents of the Florida Panhandle wanted to wake up to: Hurricane Michael had strengthened overnight into a category 4 storm, sporting winds of 140 miles per hour. And it didn’t stop there. The hurricane remarkably continued to intensify until it slammed into land Wednesday afternoon with devastating 155 mph winds at its core—just shy of a category 5 designation.

Michael’s jump from category 2 to 4 in just 24 hours was far from the first time a hurricane has made such a major leap in strength: Hurricane Harvey ramped up quickly before it hit Texas last year, and in 2004 Charley made the same jump as Michael in just three hours before battering southwest Florida.

In the 14 years since Charley meteorologists have made significant strides in understanding why storms undergo such rapid intensification—generally defined as a storm strengthening by 30 knots (34.5 mph) within 24 hours—and in predicting when and where they will do so. Even within the last two years some new forecast methods are showing promise for a problem that had long seemed intractable. “I think we’re seeing a huge leap forward,” says Chris Rozoff, a project scientist with the National Center for Atmospheric Research.

3 Ingredients

For a tropical storm or hurricane to rapidly intensify, it needs three key ingredients: low wind shear, warm ocean water and high humidity. The first refers to winds blowing at various speeds and in different directions in the upper and lower layers of the atmosphere. Strong wind shear knocks a hurricane off-kilter and disrupts its development. Warm ocean water increases the temperature contrast between Earth’s surface and cooler layers higher in the atmosphere, which increases air instability and fuels the convection that drives the storm. Reduced humidity would also rob the convection process of fuel.

All three of these ingredients were present with Michael—an unusual situation for this time of year, this far north in the Gulf of Mexico. Generally, by October cold fronts have started to press southward across the U.S., bringing relatively high wind shear and dry air with them. But Michael has been embedded in a fairly humid air current, says Brian Tang, an atmospheric scientist at the University of Albany, S.U.N.Y. Ocean waters have also typically begun to cool off from their end-of-summer peak, but this year stagnant air patterns have kept the waters in the Gulf unusually warm, he adds.

But even with all those ingredients, “it’s not always a guarantee” rapid intensification will occur, says Brian McNoldy, a hurricane researcher at the University of Miami. It will not happen without the processes in the core of the hurricane “coming together to complete the job,” Rozoff says. Accurately modeling those processes—even just a day in advance—is extremely difficult because they happen on such a small scale. But researchers are starting to understand what to signals to look for, and models are continually improving. “We’re definitely understanding more and more of the details of how that occurs,” Tang says.

A more compact storm with higher winds and thunderstorms concentrated near its center seems to have an easier time undergoing rapid intensification than a bigger one does. Its smaller size means it can more efficiently pull air toward its center, concentrating angular momentum; the classic analogy is how figure skaters pull in their arms to spin faster. Hurricane Charley was a prime example of a small storm that was able to do this. In addition to the favorable environment it was in, Michael was also showing some of these signs on Tuesday, including lightning that indicated strong thunderstorms close to its core.

Improving Predictions

Computer models have become far more detailed over the last decade, and some are used specifically to look for the potential for rapid intensification. With Michael, “they saw the conditions would all be primed for this to happen,” McNoldy says.

Rozoff is working on another method to help predict rapid strengthening: He uses machine learning to have computers take the hurricane model predictions and look for historic examples that are close matches. Whether or not those matches involved rapid intensification can help predict whether it might happen with the current storm. This method captured what would happen with Michael better than some other models did, Rozoff says. He hopes the National Hurricane Center might eventually use it as one of its forecasting tools.

There is still room to improve predictions of rapid intensification, as models become more complex and as observations provide more examples for scientists to dissect in order to understand the underlying physics. But there may be a limit to just how far in advance such strength jumps can be predicted, because of the small scale of the key processes involved; any tiny errors present at beginning of the forecast are increasingly magnified the farther out the forecast looks. Such errors matter less when predicting the path a hurricane will take, by contrast, because that path is determined by larger patterns in the atmosphere.

Hurricane Michael will surely be one of the case studies that scientists dig into to improve their models and forecasts. The potential for such last-minute strengthening before landfall is one of the reasons Rozoff and others have cited to study rapid intensification and improve storm warnings. But although this kind of situation is “stuff we talked about in theory,” he says, no meteorologist actually wants to see it happen in real life.