In the 20 years since Hurricane Katrina devastated New Orleans, hurricane scientists have made great strides toward understanding how climate change influences tropical cyclones, at the same time as they have vastly improved hurricane forecasting. Better forecasts, in turn, save the country billions every time a storm makes landfall, according to a 2024 analysis published in the National Bureau of Economic Research.
But the progress didn't come out of thin air, says Gabriel Vecchi, a hurricane and climate scientist at Princeton University. In 2005, Katrina, and other damaging storms from that era, like Rita and Wilma, spurred a concerted push to get better at forecasting hurricanes. That energy was harnessed into a federally supported research effort, called the Hurricane Forecast Improvement Project, that gathered together the best scientists from across agencies, private universities and national laboratories in an effort that has continued through today.
A five-day forecast of a hurricane's future track, for example, is roughly as good now as a two-day forecast was when Katrina struck.
But that progress, and future improvements, are now at risk, Vecchi says, because of the ongoing disruptions to federal agencies and the country's science enterprise imposed by the Trump administration, which has slashed the size of the federal government in the months since Trump took office.
"Improvements in our hurricane forecast models and our weather forecasts and our climate forecasts…are best viewed as investments, not as expenses," Vecchi says. "They're not equivalent to going out to a fancy meal. They're equivalent to a 401k," he says — and when the investments stop, the progress stops too.
The rise of HFIP
The 2005 hurricane season included a record-breaking 27 named storms — including Katrina, among the costliest hurricanes in American history.
The season stunned Americans and forecasters alike.
Scientists across the country immediately began to ask how they could improve forecasts. They had been making steady improvements to track forecasts — or figuring out where, exactly, a storm would go — but they were struggling to forecast how much, and how quickly, any given storm could intensify. In fact, there had been essentially no improvement in intensity forecasting for two full decades.
To solve the intensification problem, scientists knew they'd have to get better at modeling the storms and the weather that spawned them. They'd also need better observations — from satellites, the atmosphere around the storm or from the ocean's surface. And they'd need the raw computing power to meld it all together.
In 2007, the National Oceanic and Atmospheric Administration introduced what came to be called the Hurricane Forecast Improvement Program. It included NOAA scientists, along with researchers from universities and national laboratories. In 2009, it took on a congressionally mandated goal: to make forecasts of hurricane track and intensity 50% better within a decade.
It was an ambitious goal, says Vecchi — and "an all hands on deck approach."
How exactly they would achieve it was another question.
Modelers go to work
Hurricane forecasts are created with computer models, often nested inside other, bigger models of global weather and climate.
"With any kind of hurricane forecast, you're limited by how good your computer models are," says Jeff Masters, a hurricane scientist who flew on NOAA's Hurricane Hunter planes for years and now writes for Yale Climate Connections.
An average hurricane is a few hundred miles in diameter. That's relatively small compared to the size of the planet and its big-scale weather, which is carried along in giant patterns like the jet stream. That big-scale weather often controls hurricane track, or where the storm will go.
But some critical factors that influence hurricane behavior are much smaller than the storm itself: a small patch of extra-hot water right at the ocean's surface could provide extra power. Turbulence just a few feet wide swirling on its edges could push it to grow, or help hasten its demise, depending on the conditions.
In the early 2000s, big weather models didn't have the computing power to create enough detail to accurately reproduce the physics controlling storm behavior — especially not the structure of the storm's inner core, which scientists thought was the key to understanding them.
So scientists needed to figure out how to make the storm-scale models nested inside the big ones better. And both needed to look at smaller and smaller-scale details.
In the mid-2000s, NOAA kicked off a project to develop a new high-power computing system at a laboratory in Boulder, Colo. Scientists worked in tandem: some developed models using the ever-increasing amount of computing power available at the Boulder site. Others worked on improving the models using the more limited computing resources available elsewhere. And every year, they'd trade learning back and forth, iterating to make both sets of models better.
Quickly, the different groups got better at taking in data from the satellites or other observations, understanding the smaller and smaller-scale physics controlling storm behaviors, doing the math on all of it faster and blending together the results from many different models to get one that was more likely to be right.
Computer chips, satellites and planes
Making the models better improved forecasts. But there was another critical piece to improve: the models needed good information to start from.
"It's garbage in, garbage out," says Andy Hazelton, a hurricane expert at the University of Miami who worked at NOAA's National Hurricane Center until the Trump administration's cuts to NOAA eliminated his job. Better observations of the real atmosphere and ocean, in real time, are critical to good forecasts.
Doppler radar measurements taken from aircraft that flew through the storm gave the models near-real-time information about the shape of whirling storms. Data from dropsondes, small parachuted devices dropped out of the aircraft that collect data as they fall toward the ocean's surface, added more insights.
"The improvement in forecast and track forecast has been shown to be up to 24% greater when a jet is flying," says Masters.
But hurricane hunter planes can't fly through every storm. So scientists needed other new sources of data to fill in some of the blanks.
In the early 2000s, satellites orbiting Earth could provide pictures of the top of the cloud layer. Those pictures were getting more detailed: by 2017, the GOES-16 satellite could beam down crystal clear images with a resolution of under 1 kilometer. But peering below that top layer, into the middle part of the atmosphere and the storm itself, was critical.
That started to change in the mid-2010s, when NOAA and other agencies began to get data back from several different microwave-focused satellites, including one run by the Department of Defense.
The microwave data is like running an MRI on the storm, explains Kim Wood, a hurricane scientist at the University of Arizona. Microwave-focused satellites "help us peer inside the storm and get a sense of structurally speaking, is it primed to get strong really fast?" Wood says — show how much energy and potential precipitation is hidden inside a storm.
HFIP, and the parallel improvements in computing and observations, worked. By 2023, forecast accuracy for hurricane tracks improved by 50%. In all, a five-day-out forecast was now as good as a two- or three-day forecast had been in 2005.
As for intensity, the improvements were similarly impressive, says Wood. Errors in intensity forecasts dropped by 56% from 2007 to 2023. By 2021, the National Hurricane Center had boosted its ability to forecast rapid intensification by about 25% over the center's mid-2010s performance, which was itself already better than 2005.
The vast improvements make a clear point, says Masters: "Funding works. If you want a better forecast, invest in the research to do that."
A good forecast is worth billions
But a question remained: Was all this improvement helping people and property?
An accurate forecast should, in theory, help decision-makers and families decide how to prepare: to evacuate, close businesses or not, tape down windows, buy emergency supplies or take other actions to protect themselves. Warning too many people could slow down economic activity unnecessarily. But warning too few, or too late, could lead to safety risks and infrastructure damage.
In 2024, a team of economists evaluated the value of the country's increasingly accurate forecasts. They found that from 2009 through 2019, the HFIP spent roughly $250 million.
The value of the improved forecasts, in contrast, adds up to roughly $2 billion for every single major storm that makes landfall
"It was one of those investments in research that paid off," says Vecchi. "That benefit is much, much larger than the total cost."
In fact, the amount of money saved because of improved forecasts for a single hurricane — let alone the dozens that hit the country each year — is larger than the cost of the entire federally funded weather forecasting operation.
Climate change increases hurricane risk
In 2005, there was a theoretical understanding that climate change would enhance some of the risks, says Andra Garner, a climate and extreme weather scientist at Rowan University. But by 2025, researchers are virtually certain that tropical cyclones are becoming more destructive.
A hotter planet means a hotter ocean. Hot water is like food for a hurricane, providing energy the storms can use to grow and strengthen, says Kerry Emanuel, a hurricane expert at Massachusetts Institute of Technology. Sea surface temperatures in the Gulf of Mexico have crept up by 1 degree Celsius since the 1950s and have smashed historical records in recent years.
A hotter atmosphere can absorb more water, like a thirsty sponge. A patch of air can hold seven percent more water than another patch just one degree Celsius cooler (the planet has warmed by more than 1.2 degrees Celsius, or 2.1 degrees Fahrenheit, since the late 1800s). That means there's more water suspended overhead, ready to be wrung out from the hurricane as rain.
"Water is by far the major killer and source of damage in hurricanes," says Emanuel. "And yet when we say the word hurricane, most people conjure up, even I do, an image of a wind storm. Well, of course it is a wind storm, but the thing that really kills is water."
Scientists have found that climate change made storms like Hurricane Harvey, which inundated the Houston area in 2017, drop an extra 15% of rain.
Storms are also traveling farther northward. The latitude at which they reach peak intensity, Emanuel says, has crept northward — leaving cities like New York, or regions like inland North Carolina, in the path of storms that were unlikely to reach them before.
Still more to learn — and major hurdles emerging
The improvements in hurricane science since 2005 have been vast, says Masters. But the work is far from over.
HFIP assigned itself new goals in 2017. It now aims to create functional models of a storm's inner core, critical to an accurate forecast, a full seven days out from landfall. And it plans to improve, and lengthen the lead time for forecasts not just of wind speed and rapid intensification, but storm surge, rain, and severe gusts — factors that are harder to model, but are valuable to the emergency managers helping people stay safe.
Crucially, says Hazelton, the new iteration of HFIP also incorporates social science — because the best forecast in the world doesn't matter if it's not received by those in harm's way.
But, Hazelton stresses, all the HFIP's goals — and the country's ability to track and forecast hurricanes more broadly — is now uncertain.
Earlier this year, the Trump administration aggressively slashed staff and budgets at agencies like NOAA. Many of the staff who fly and maintain the Hurricane Hunter planes and provide their forecast-critical information were fired in the early rounds of layoffs that swept the federal government.
And data from the microwave satellites that provide the MRI-esque view of the storm's innards were nearly halted abruptly this spring, though the decision was later reversed after outcry from scientists.
The president's proposed budget for NOAA would reduce support for researchers who work on many aspects of hurricane modeling and observations, along with one of the major modeling research centers at Princeton University.
Wood worries about the disinvestment in research "We didn't get this far by saying, 'Yep, we figured out hurricanes. We're done,'" they said. And even now, there are many more theoretical and practical questions to answer.
But the changes across the federal government and to NOAA not only endanger future progress, says Hazelton, they undermine the gains that have already been made.
"We're setting ourselves back years," says Hazelton.
Copyright 2025 NPR
