Satellite view of Hurricane Dorian on Sunday. (NOAA/CIRA/RAMMB)

As Hurricane Dorian sets its sights on the Southeast, forecasters are working feverishly to narrow down its most likely path. Predicting its strength has also been fraught — largely because of a phenomenon known as rapid intensification.

Dorian intensified from a 110 mph Category 2 storm on Friday morning into a monstrous 150 mph Category 4 storm Saturday. Then, on Sunday, its winds surged to a 185 mph Category 5, tied for the second most intense Atlantic hurricane on record (with Gilbert in 1988, and Wilma in 2005), only trailing Allen in 1980 (with 190 mph winds).

While the possibility of rapid intensification was predicted, forecasters were forced to play catch-up as the storm kept growing stronger than expected, adjusting their projections based on each observation.

It’s not just Dorian that forecasters have struggled with.

Scientists with the National Oceanic and Atmospheric Administration have been wrestling with how they can get a better understanding of what happens inside a storm when there’s an increase of sustained winds of at least 35 mph in a 24-hour period.

One recent workday, with no hurricanes imminent, NOAA scientist Frank Marks and his colleague Xuejin Zhang stared at an oversize computer monitor in Zhang’s office inside NOAA’s Atlantic Oceanographic and Meteorological Laboratory on Virginia Key, just outside Miami.

An animation on the screen showed splotches of colors — yellow, orange, red — coalescing over a map of the Gulf of Mexico. The colors converged into a red swirl and then spiraled north into the Florida Panhandle. The animation was a model of Hurricane Michael’s path in October 2018. The forecast had been generated the moment the storm formed in the gulf near Belize, three days before making landfall. It was precisely accurate.

“This is pretty good,” Marks muttered as he revisited the work of the model, known by its abbreviation HWRF.

The path may have been correct, but the details of the storm itself were not. The model showed a central pressure of 956 millibars coming ashore as a Category 2 hurricane with wind speeds of about 100 mph. Instead, the actual Michael raged ashore as a Category 5 — only the fourth recorded storm of that intensity to make landfall on the U.S. mainland — with a minimum central pressure of 919 millibars and maximum sustained winds estimated at more than 161 mph.


Infrared satellite image from the time of Hurricane Michael's landfall. (CIRA/RAMMB)

What the models failed to predict were dramatic surges in power right before landfall, in which wind speeds ratcheted up more than 35 mph in a 24-hour span — rapid intensification. Michael did this three times in three days.

“The big stretch goal,” Marks, the director of hurricane research at the lab, has called efforts to understand rapid intensification. “The big nut.”

The latest tools Marks and his team have been experimenting with include specialized drones that can fly around a hurricane’s eye and sophisticated plane-mounted radar units that can measure wind motion, all to better understand what happens inside the storm during intensification.

Improving rapid intensification forecasts is important because of the danger it poses to coastal populations. “Forecasting rapid intensification is hard, and so many decisions on the emergency management side depend on that forecast,” said James Franklin, former chief of forecast operations at the National Hurricane Center.

“Emergency managers will evacuate different groups of people based on whether a hurricane is a Category 1 or a Category 3,” he said. The variables that create the conditions for rapid intensification have bedeviled scientists so far.

“What makes the hurricane intensify is a series of random processes,” Marks explained. The storm needs an environment where there is a quick exchange of energy from warm tropical ocean waters into the center of the storm’s low-pressure zone.

But a lot of things can disturb such an ideal environment, such as pools of cool water, dry air and wind shear, when wind is moving at different speeds or directions at different levels. In an isolated environment without these inhibiting factors, quick strengthening would always occur.

“The real question is why it doesn’t happen all the time,” Marks noted.

Marks has spent the past 10 years studying this phenomenon, ever since NOAA identified it as a priority. He and his team are constantly refining the two types of computer models used in hurricane forecasting: dynamical, which takes the current state of the atmosphere and calculates how it will act in the immediate future, and statistical, which uses the behavior of past storms to evaluate how a hurricane might act.

Right now, forecasters get much of their information about a hurricane from dropsondes, cardboard tubes carrying data-gathering instruments launched from hurricane hunter aircraft. The tubes descend through the storm on parachutes before they are lost below. The information gathered during descent is transmitted to the NHC in real time, and the forecast models use it to predict where the storm will go and how strong it is.

But it also helps researchers later understand how the models processed the information. This allows them to refine the models and increase accuracy the next time. But oftentimes, what happens during rapid intensification occurs very quickly in parts of the storm not captured by the dropsondes.

Hurricane drones

To augment the dynamical models and get a more intimate understanding of what happens during intensification, Marks and his team are looking at innovative ways to get information from deeper inside the storm where hurricane hunters can’t go because of safety concerns. To do this, the scientists have turned to specialized drones, also launched from aircraft, which can fly around and through the hurricane’s eyewall, the most dangerous part, for up to an hour or more gathering data before the battery runs out or the storm destroys them.

Dropsondes give pinpricks of information, explained Joseph Cione, who is leading the program’s drone research. The drones can cover a much larger field: “A movie versus a snapshot,” Cione said. The information is vital to help scientists better understand the storm’s structure and how it behaves right before intensification.

Marks is optimistic. “The data we’re getting from them is fantastic,” he said.

Experiments started in 2005, when they launched the first drone into Hurricane Ophelia from a truck on a beach just to test the technology’s viability. The first time research-quality sensors were used in a drone was 2014.

In 2017, several drones built by Raytheon were launched from a P-3 aircraft into the eye of Hurricane Maria, and for the first time, information was instantly sent to forecasters at the NHC. The data was not assimilated into the models, but it was made available for forecasts. Cione is working on making the drones a regular part of the program by coming up with cost-effective designs.

As of Friday, no drones had been launched into Dorian.

The drone prototypes being used run about $20,000, with one military-grade prototype costing $70,000. They need to be a lot cheaper for their one-time use in a storm. Cione said he would like to see a design that costs less than $10,000. While that is considerably more than a dropsonde, which cost about $800 each, the per-minute value is a lot higher.

Dropsondes are active for only about five to eight minutes in a storm. Drones can operate for more than an hour gathering information and can be directed to specific locations. This data has the potential to improve the physics of the models NOAA uses to predict hurricane intensity.

New radar systems

Advanced radar systems provide an even wider field of vision than that of the drones. Marks and his team are experimenting with the use of Doppler radar and Doppler lidar, which can measure wind motion and potentially record data about wind speeds near the core of a hurricane that could help predict when intensification will occur. The lidar is owned by the U.S. Defense Department and has been loaned to NOAA during the past five hurricane seasons. It is mounted on NOAA’s P-3 aircraft, which fly right through the hurricanes. Use has been sporadic.

One of the two P-3′s that researchers use has not been available, and there were already too many instruments on the remaining plane, so the lidar could be used only during part of the season. The 2019 season is the first in which the lidar will be in operation all season.

Marks and others don’t expect an immediate breakthrough in the effort to predict rapid intensification. It’s a long-term goal, he said. But every year, there are prediction improvements. And because of the warming oceans (warm water is hurricane fuel), it’s never been more crucial.

Several recent studies have indicated that intensification has been occurring more frequently, probably because of climate change. The most recent, published in the journal Nature Communications this year, calculated 24-hour wind-speed changes between 1982 and 2009 and detected that the proportion of rapid intensification in storms has increased by a magnitude that natural climate variability could not explain.

As far as Hurricane Dorian is concerned, forecasters had warned about the elevated odds of a leap in strength. A promising sign.

Tristram Korten is a journalist based in Miami and author of “Into the Storm,” about Hurricane Joaquin and the sinking of El Faro in 2015. Follow him @TristramKorten.