NEWS | Jan. 17, 2020

Predicting Hurricane Dorian using NRL’s tropical cyclone model COAMPS-TC

By NRL Corporate Communications

WASHINGTON— At the time, senior scientist Jim Doyle was a little worried. Projections by his team’s tropical cyclone model were deviating from those of the other major weather models, which were predicting Hurricane Dorian would make landfall in Florida. His own team’s model showed the storm would curve northward along the coast.

“You know, it’s just human nature,” said Doyle, whose team develops and runs the U.S. Naval Research Laboratory’s Coupled Ocean-Atmosphere Mesoscale Prediction System-Tropical Cyclone model (COAMPS-TC®). “We were watching closely. I'll say that much. And we were a bit concerned, because we thought, well, maybe it’s curving to the north a bit early.”

It turned out that COAMPS-TC was one of the only models to indicate early on that the storm would curve to the north and not make landfall in Florida, according to Doyle and the National Hurricane Center. It was little consolation: the model performed well, but Dorian went on to bombard the northernmost islands of the Bahamas, where it wrought historic levels of destruction.

Making landfall Sept. 1, the Category 5 storm turned homes into rubble and flattened neighborhoods with winds reaching nearly 200 mph, leaving more than 75,000 in need of shelter and food assistance, 1,300 more missing and at least 67 dead. According to various Bahamian government reports, there is an estimated $3.4 billion recovery cost as of late October.

Although they often must think of tropical cyclones in abstract, computational terms when studying and modeling them, Doyle and his team are well aware of what transpires on the ground when people, property and infrastructure stand in the path of one.

“We watch the news, too, and it's devastating,” Doyle admitted. “Nobody wants to see humans suffer and experience loss and death and destruction because of these storms. I think all meteorologists feel the same. It underscores the importance of what we do.”

NRL’s Marine Meteorology Division in Monterey, California, began developing the COAMPS-TC around 2010. COAMPS-TC became operational in 2013. It is a derivative of NRL’s COAMPS computational weather model, which has supported Navy missions around the world for years.

Doyle’s team of 10 researchers works on testing and further refining the COAMPS-TC annually. Those refinements can take years to implement operationally; sometimes, they take just one hurricane season. The most-recent COAMPS-TC version to go into operations transitioned in May 2019.

“Each season, we come up with new ideas on how to improve it,” Doyle said. “Then we'll have to test it on many, many storms across multiple seasons to be able to get a statistical signal on whether it improves the system or not.”

How does COAMPS-TC work?

COAMPS-TC uses multiple nested grid meshes to follow tropical storms. Increasingly smaller grid cells on each mesh descend from the outermost mesh to the innermost. The innermost mesh uses a grid of cells, each measuring 4 square kilometers, to create high resolution representations of the hurricane, its inner core, and various aspects of its circulation.

“At each of these cells, 12 predictive equations are solved and information from the neighboring cells is shared frequently as the model runs,” Doyle said. “The outer mesh doesn't cover the whole globe. It covers only part of the globe, and usually it's an entire basin — for example the Western Pacific, along with other basins such as the Atlantic, Eastern and Central Pacific, and Indian Ocean.

“We have several different areas around the world for COAMPS-TC, where it has a basin-scale outer grid mesh. And then within that outer mesh, the two nested grids that follow the storm operate.”

And where does all the data for all those cells come from? Though the researchers are in the habit of referring to COAMPS-TC in the singular, today there are two versions of COAMPS-TC, each a limited area system coupled with a different global model that provides the large-scale conditions for its initial state. The data come from those two global models.

One version of COAMPS-TC begins with data from the Navy Global Environmental Modeling system. The other version of COAMPS-TC makes use of data from the National Oceanic and Atmospheric Administration’s Global Forecast System. Essentially, these global systems drive the limited area systems. The arrangement is called one-way coupling.
NAVGEM and GFS are both sophisticated numeric weather prediction models that employ data assimilation systems taking in observational data from every conceivable source. These include weather balloons, surface stations, aircraft, ships, and satellites that make up most of the observing system. The number of observations per day used in these systems are in the millions.

“The number of satellites is finite, and it's smaller than the number of observations,” Doyle said. “But, of course, you can have one satellite sensor with hundreds of channels – these are hyperspectral satellites.

“The global models provide the initial state that COAMPS-TC is using, and we specify the hurricane vortex itself using the information directly from the forecasters about the intensity, position, and size. We insert the vortex based on that information into the COAMPS-TC initial model fields.”

Generally, the operational versions of COAMPS-TC run on a supercomputer at The Fleet Numerical Meteorology and Oceanography Center in Monterey, while NRL Monterey runs test versions at the Department of Defense Supercomputer Resource Center at Stennis Space Center in Mississippi.

Further refinements, initiatives

Like all scientific disciplines, the meteorology community still grapples with its share of mysteries. Their models advance as their knowledge of the Earth, the oceans and the atmosphere advances. Among the phenomena that still puzzle them is that of rapid intensification, defined as the change in the intensity of a storm by at least 30 knots over a 24-hour period.

Scientists are still trying to understand exactly why some storms rapidly intensify and how to predict when it will happen. Not every storm undergoes rapid intensification, but the phenomenon is far from rare. Hurricane Dorian underwent not one, but two instances of rapid intensification with the first taking it from category 2 to category 3; the second from category 3 straight to category 5.

“Rapid intensification is really hard for the forecast models to capture,” Doyle said. “There's a lot of subtle processes that we don't understand yet that control this intensification. So, Dorian getting up to category 5 was a bit of a surprise.” Models, including COAMPS-TC, hinted that it could go to a strong category 4 in its forecasts, but the timing of that intensification was not particularly accurate.”

In an Office of Naval Research-sponsored research initiative, Doyle’s team and academic researchers are collaborating with scientists on a field project to learn more about rapid intensification. The project utilizes NOAA’s aircraft, including two Lockheed WP-3D Orion and a Gulfstream IV-SP jet, to drop instrumented packages called dropsondes into storms to collect profiles of wind, temperature and moisture.

“There are some pretty good theories on why rapid intensification happens, but I think understanding the details of when it happens in a particular storm's life cycle is really challenging,” Doyle said. “We want to be able to predict when a storm just offshore will undergo rapid intensification like Hurricane Harvey did. Because that is critical for Navy and civilian decision makers.”

Meanwhile, Doyle’s team is already working on a new data assimilation initiative to initialize the COAMPS-TC model based on satellite radiances, the raw measurements that satellites take using solar-reflected and Earth-emitted radiation. Doing so should enable the model to produce more realistic representations of storms, according to Doyle.
“The best approach is to use the raw radiance information directly in the model data assimilation system,” he said. “At this point, we’re trying to demonstrate a new data assimilation capability using high-spatial and temporal resolution radiance measurements.”

With these and other efforts, the division has set a goal to improve forecasts for more advanced tropical cyclone warnings, and to track forecasts better, the storm size, and its intensity. Accomplishing these goals will help governments, the Department of Defense, emergency managers, and the public to prepare and, if necessary, get people and assets out of harm's way in time.

“One of the decisions the Department of Defense needs to make, for example, in Norfolk [Virginia] is whether to send ships out to sea if a storm is coming close,” Doyle pointed out. “This is one of the applications of COAMPS-TC that's really important. Prior to Dorian, the Navy made the decision to sortie the Hampton Roads-based ships and aircraft.
“We have a very dedicated team that’s trying to improve the forecast, because its impact is so front and center, not only for the DOD but of course for everyone whose lives and property will be affected.”

From Dorian to Lorenzo

But for the most part, the process is automated, essentially a network of supercomputer servers interacting with one another. The servers used by the forecasters — in Dorian’s case that’s the National Hurricane Center — send out a signal every six hours alerting weather systems models that they have identified an investigation area (meteorologists call them “invest areas”) or a tropical cyclone and provides the location, intensity and size of the storm. That information is used to initialize the COAMPS-TC hurricane vortex.
“They initiate a message that has the information in a certain format through the Automated Tropical Cyclone Forecast system, and the models, like the NOAA models and our Navy models, look for that message,” Doyle said. “The software downloads the message, sorts through the different storms, and identifies which storms to run.”

Meanwhile, NOAA conducts special observations from their aircraft that help meteorologists gain a more thorough understanding of the circulation and other processes inside the tropical cyclone. The U.S. Air Force flies C-130s into the storm to get measurements of the peak winds, which tells forecasters about the storm’s current strength, size and location including the center of the circulation (if one can be found).

Doyle’s team of meteorologists conduct weekly meetings to discuss the development of the system, the storm’s track and intensity, according to COAMPS-TC’s projections and those of a half dozen other models that are also predicting the storm, including NOAA, the United Kingdom Meteorological Office and the European Centre for Medium-Range Weather Forecasts.

As an entity, Hurricane Dorian finally dissipated Sept. 10. By then, Doyle’s team was already tracking the development of another storm, Hurricane Lorenzo, which would soon intensify into a category 5 storm, packing peak winds of almost 150 mph. By the end of the month it would become the largest recorded storm to travel so far northeast in the Atlantic.

About the U.S. Naval Research Laboratory
NRL is a scientific and engineering command dedicated to research that drives innovative advances for the Navy and Marine Corps from the seafloor to space and in the information domain. NRL headquarters is located in Washington, D.C., with major field sites in Stennis Space Center, Mississippi, Key West, Florida, and Monterey, California, and employs approximately 2,500 civilian scientists, engineers and support personnel.