Friday, October 9, 2026

On a remote island, these GPS hackers are preparing for an invisible war

Example of spoofed smartphone

From BBC by Katherine Dunn

Energy grids, data networks, stock markets: how easy is it to shut down our modern world by attacking the systems that synchronise time and space?
Meet the jammers finding out.

Harald Hauglin is a toweringly tall man with a salt and pepper beard, wire-framed glasses, and an impish grin.
When I met him in September 2024, the pocket of his mandatory hi-vis vest displayed a patch showing a white rabbit on the run with a clock.

Hauglin is chief engineer of time and frequency metrology at the Norwegian Metrology Service.
He is also a "Nerd of the Northern Lights" – a nickname adopted by attendees at the 2024 Jammertest.


Andoya island with the GeoGarage platform (NHS raster chart)
 
Hosted annually by the Norwegian authorities, Jammertest is an open-air hacking festival set on a remote island 300km (186 miles) north of the Arctic Circle.

Its aim is simple but mind-bending: to use satellite signals to mess with time itself – and see if receivers in everything from clocks to drones can withstand the pressure.

At the heart of this experiment is a technology that runs much of our modern digital world: Global Navigation Satellite Systems (GNSS).

We tend to think of satellite networks such as GPS as tools for navigation.
Yet they also do something else just as important: they transmit ultra-accurate time.
Power grids, telecommunications networks, stock markets, transit services and more all depend on the synchronisation provided by GNSS.

But the importance of GNSS to infrastructure also creates an Achilles' heel.
Satellite transmissions are weak, and can be overwhelmed or hijacked by stronger, nefarious signals.

Since Russia's full-scale invasion of Ukraine, such interference has risen dramatically.
The most obvious impacts have been on aviation.
In May of this year, an RAF plane experienced GPS jamming near the Russian border, while in September 2025, the Swedish Transport Agency said that interference was a daily occurrence.

Norway's government is now one of several preparing for a world in which GNSS could become a strategic vulnerability.

So, is it possible to prepare for an attack on our sense of time and place itself? Each year, dozens of engineers, physicists and officials head to the Arctic Circle to find out.

Jammertest: The world's largest open resilience-check for satellite jamming


The setting for Jammertest is the village of Bleik on Andøya, a tiny, remote island often used to study the Aurora Borealis.
The picturesque venue resembles an engineering summer camp, the epic Arctic scenery and daily waffle buffet all part of the appeal.
Yet its purpose is deadly serious.

Jammertest was only in its third official year when I visited, but the idea of testing commercial GNSS receivers to see how they react under intense, if artificial, interference is not new.

For years, the US military has been running tests to show how receivers react when GNSS is manipulated, held at the White Sands Missile Range in New Mexico.
These tests are tightly controlled, however, including what data can be shared after the fact.

The Norwegians take a diffrent approach, modelling the test event on the white-hat hackathons beloved in the tech industry, where hackers probe for chinks in cybersecurity armour to make software better.
Crucially, these tests are done openly, in full view of any visitors to Andøya, and attendees share the results of their tests with other government and commercial attendees on the final night of the event.

 
Katherine DunnJammertest, an open GNSS resilience test, is held annually at Andøya, Norway
(Credit: Katherine Dunn)

Over the several days I attended the 2024 Jammertest, I sat next to Hauglin in the log cabin-style Bleik Community House, trying not to spill coffee on expensive electronic equipment as he and his colleagues unleashed staged attacks.
The job involved feeding false information to receivers in everything from planes to cars to industrial equipment: about where they were located, and often, at what time.

Think of it like this.
If you are reading this story in London in September 2026, GNSS manipulation would feed you false information that, for example, you're in Lima, Peru – and it's September 2017.
 
The risks of extended and slow-burn 'jamming'

There are two main kinds of GNSS interference.
The first is jamming, or overwhelming signals so that the service is no longer available.
The second is spoofing – a hijacking of the radio waves, where a false location and time is offered to a GNSS receiver.

The most dangerous kinds of GNSS attacks are two-fold.

The first major risk is extended jamming: a sustained, total wipe-out of GNSS, long enough to run down back-up timing and cause scores of knock-on issues as the digital synchronisation stitching together our modern world begins to fray.

How long a power grid – or any modern digital system – can remain synchronised without access to GNSS comes down to how much money a government or company is willing to spend installing the physical stuff that GNSS otherwise makes irrelevant.
Largely, this is a combination of land-based clocks and fibre optic cables to transfer time with extreme precision.

The second, even worse scenario, is a slow-burn, sophisticated spoofing attack that isn't noticed or contained, and therefore never allows back-up systems to kick in.
This is essentially a timing-based sneak attack that could dislodge digital systems in one fell swoop.

Both these kinds of attacks go after GPS time, and both involve the most sophisticated and powerful levels of manipulation or disruption – comparable to elaborate cybersecurity attacks.

They are also the kind of attacks that Hauglin and the Metrology Service, alongside numerous companies who provide industrial time as a service, meet on Andøya to test out.

The aim is to explore how vast digital systems may cope in the short-term following an attack, multiple people told me.
For example, by establishing exactly how long the energy grids and mobile networks have before things start to go seriously wrong.

How to peel away space and time

On the first afternoon at the 2024 Jammerfest, there were two test-flights: a Norwegian rescue helicopter, and a small plane, flown by a pilot from Eurocontrol, the European aviation safety body.

I watched as manipulation of the GNSS signals fed false information to the receivers on both flights, creating location tracks that didn't reflect where the pilots had actually flown. 

The helicopter's location on the flight tracker appeared to move in countless overlapping, panicked swirls, like an anxious doodle on a page – manoeuvres a helicopter wouldn't, and couldn't, pull off in real life.
Later, the tracker showed Eurocontrol's test plane making a series of odd, paperclip like patterns in the sky, full of extreme, gut-churning turns.
In reality, the plane had been doing wide, graceful loops above the town.

On the second morning, I watched as Hauglin undertook more sophisticated disruption.
Over 40 minutes, he and his colleagues gradually raised the power of the spoofing equipment, strengthening the decoy signal: the electromagnetic equivalent of slowly overwhelming a system before it realises anything is wrong.
Like boiling a frog.

As the transmission from the sneaky antenna ramped up, a screen on a monitor mounted on the wall above us showed the area on Google Maps.
Steadily, a beacon marking our location moved, misleadingly, out into the Norwegian Sea.
 
What does it look like when we 'hack' time?

The impact on clocks was harder to see without complex equipment.
To track the "drift" caused by interference – the way a GNSS-synchronised clock can be forced to gradually peel away from the real time – it was necessary to compare GPS time to an uninterrupted reference time.
This reference time came via fibre-optic cables unspooled all the way to a GNSS-receiver on the other side of a mountain, where it was safe from interference.

I watched over Hauglin's shoulder as he showed me a chart that tracked the spoofed clock's drift, measured in nanoseconds, away from the reference time. 

 
attended Jammertest in 2024 (Credit: Katherine Dunn)

As Hauglin said, the issue is that we tend to take time for granted – and we certainly don't like to think of it as a weapon.
 
Can hacking be stopped?

New options to replace, or strengthen, GNSS are also being proposed or developed all the time.

Do you remember the white rabbit on Hauglin's shirt?
As well as a reference to the character in Lewis Carrol's Alice's Adventures in Wonderland, the motif also represents White Rabbit – a precision timing system developed at Cern in Switzerland, which can deliver time accurately to less than a nanosecond.

This and other timing systems usually replace the reliance on satellites with fibre optic cables, and atomic reference clocks.
And they are much harder to hack.

But nearly everyone I spoke to agrees that there is no single alternative system that can match every service GNSS has come to provide.

"There is no silver bullet, which is why you need a combination of systems," says Dana Goward, who has campaigned for policies to protect GPS since 2013, as the co-founder of the Resilient Timing and Navigation Foundation.

In 2024, Goward told me he believed spoofing was only going to get worse – "it's cheap, it's easy" – and that one of the biggest risks could come from the other GNSS systems themselves.

In June 2026, those fears were realised when a group of experts warned that Russian satellites have beamed interference from above, disrupting GPS signals across Europe for brief periods.

In the past, GNSS interference was limited to a region or a border area, as radio signals from Earth can only travel so far.
Now, with interference also coming from above, nowhere and nothing will be off-limits.

Satellite navigation is essential to everyday life, from tracking your morning jog to landing air ambulances.
But as reliance on satellite navigation grows, so do the risks associated with its interruption, natural or intentional.
To strengthen European resilience in navigation, the European Space Agency (ESA) takes part annually in Jammertest.
Organised on the remote island of Andøya, Norway, Jammertest is the world’s largest open air testing campaign for jamming and spoofing resilience.
In September 2025, ESA engineers attended Jammertest with ESA’s mobile navigation lab to test how different systems respond to interference.
After this, the data are analysed to check which technologies perform the best against jamming and spoofing.
By bringing together academia, industry and governmental organisations, Jammertest helps make satellite navigation better for everyone and protects European assets.
 
More like this:

Given that so much of the Jammertest event for the Scandinavians was, implicitly, about preparing for the risk of some kind of attack by Russia, I was slightly curious that the 2024 event was being held so openly.
The level of security, and the presence of Norwegian intelligence, was discrete.

The openness, however, is also part of its point.

"Is it also to show: 'we are becoming more resilient? We're dealing with this?'" I asked Tomas Levin, the senior engineer for the roads administration, and one of the organisers of Jammertest.
"Of course," he said.
"So come on, Russia, go ahead."
 
Links :

Thursday, October 8, 2026

Captain's decision led to yacht grounding, says NTSB

6 Saturdays in the boatyard after the grounding, on November 22, 2024.​​NTSB

From Baird by Alan Bosworth

The US National Transportation Safety Board (NTSB) has determined that the probable cause of the grounding of the recreational vessel 6 Saturdays near St. Augustine, Florida, was the captain's decision to navigate through an area known to have shoals instead of the marked entrance channel.


The incident occurred on June 19, 2024, when the 55-foot (16.8-metre) vessel grounded on shoals at the north side of the St. Augustine Inlet.
After the grounding, the captain manoeuvred into deeper water where the vessel began taking on water and partially sank.
The captain and one passenger abandoned the vessel and were rescued by local first responders; there were no injuries.
Damage to the vessel was estimated at $1 million.

According to the NTSB's report, the navigation chart for the area contained a cautionary note warning of frequent changes in depth due to shifting shoals.
 

 Visualizations of the area (NOAA raster & vector ENC) with the GeoGarage platform
 
The captain stated he had been through the inlet "many times" previously but was unaware of any local navigational warnings and could not recall seeing any navigational aids.

The NTSB found the captain did not navigate toward the lighted whistle buoy to follow the buoyed channel, instead proceeding directly through the shoal water at about 15 knots.

 
After the grounding, the Florida Fish and Wildlife Conservation Commission issued a civil penalty to the captain for, "reckless or careless operation of a vessel".
 
Links :

Wednesday, October 7, 2026

Why is this monster El Niño so strong?


Strong winds have contributed to the current El Niño’s strength
CSU/CIRA & JMA/JAXA
 
From New Scientist by Michael Le Page


Though the El Niño is still months from its peak, it is already breaking many records.
Two main factors are behind how it became so powerful


Mind-blowing.
Unprecedented.
Off the charts.
Jaw-dropping.
Godzilla-level.
Superlatives have been flying about the current El Niño climate pattern, which is already the strongest on record months before it is expected to peak – and unusually, they are coming from scientists, rather than the media.
So, what has made this such an exceptional event?
This will no doubt be the subject of many studies to come, but the broad picture is clear.
Two main factors have driven the formation of this record-smashing El Niño, the first of which was set in motion long before it began.

To understand this El Niño, we have to understand La Niñas, the opposite phase of the climate pattern.
During La Niñas, trade winds blow west across the equatorial Pacific, pushing surface waters warmed by the sun towards Australia and Indonesia.
This piles up a layer of water known as the “warm pool”.
And it really does pile up: sea level in the western Pacific rises as much as 20 centimetres.

The bigger the warm pool gets, the more unstable it becomes.
If the trade winds holding it in place weaken or reverse direction, the warm water spills back across the Pacific, making surface waters along the equator abnormally warm – an El Niño event.
So the extent to which an El Niño raises global temperatures and causes extreme weather depends on how big and hot the warm pool gets, and on how much of it moves east across the Pacific.
The warm pool gets hotter during La Niñas and colder during El Niños, but temperatures have been increasing over the past decades due to global warming, says Michael McPhaden at the US National Oceanic and Atmospheric Administration.
By late 2025, the heat content of the western Pacific was much higher than normal – although it wasn’t the highest on record.
“Based on that alone, one might not have expected this event to become as strong as it has,” says McPhaden.
“The highest heat content developed prior to the last El Niño in 2023 to 2024 due to the preceding ‘triple dip’, or three-year La Niña, associated with stronger-than-normal trade winds that piled up warm surface waters in the western Pacific for a prolonged period.” La Niñas typically last nine to 12 months.
The second key factor was a series of strong bursts of wind blowing east, pushing the extra-hot warm pool towards South America.
This isn’t just a surface phenomenon, says McPhaden.
Each wind burst creates a downward bulge in the layer of warm water, known as a Kelvin wave, which travels across the Pacific over the following six to eight weeks.

Every Kelvin wave increases the thickness of the warm layer in the eastern Pacific.
Put another way, the bottom of the warm layer, or thermocline, becomes deeper.
“The cumulative effect of successive westerly wind bursts and associated Kelvin wave responses led to an exceptional deepening of the thermocline in the eastern Pacific,” says McPhaden.
“The thermocline is currently more than 80 metres deeper than normal in some areas.”
The deeper the thermocline, the harder it is for colder water to well up and reach the surface.
So we now have a huge amount of extra warm water spreading across the surface of the Pacific and transferring stupendous quantities of heat to the atmosphere, with repercussions across the globe.
One strong wind burst in April was driven by three tropical cyclones forming in the western Pacific around the same time.
Cyclones rotate anticlockwise north of the equator and clockwise south of it, so the winds closest to the equator always blow eastwards.
“The stronger the westerly wind nudge, the easier it is for El Niño to develop and strengthen,” says meteorologist and author Bob Henson, who pointed out at the time that these cyclones would boost the strength of the then-nascent El Niño.

In principle, it is possible that global warming played a role here, too, by increasing the odds of three cyclones forming around the same time, he says.
More widely, the question of whether global warming is making El Niños stronger is unresolved, says Michael Mann at the University of Pennsylvania.
“It’s honestly an open question from my stance.”
The number of observed El Niño and La Niña events in modern times is too small to reveal any link.
Some teams having been studying things like corals to try to extend the record back many centuries.
A recent study of corals in the Galapagos suggests there is a link, but Mann says there are crucial gaps in the coral record that mean no firm conclusion can be drawn.
“We still do not have a reliable, continuous record of El Niño prior to the instrumental era,” he says.
Climate models are little help either, he says.
“The models [are] nearly equally split as to whether the amplitude of El Niño events is likely to increase or decrease with anthropogenic warming.” 
What is clear is that warming amplifies the damage done by El Niños.
The current super El Niño will also lead to a big spike in global warming, says Mann.
“It gives us a glimpse of what is coming down the pike if we continue to warm the planet with carbon pollution.
Warmer oceans, more evaporation, bigger flooding events, hotter continents and worse droughts leading to more extensive and destructive wildfires.”
 
Links :

Tuesday, October 6, 2026

U.S. shipwreck, laden with World War II munitions, is a time bomb near London

 
The final mast of the SS Richard Montgomery is removed on Sept. 30.
The shipwreck lies in the Thames estuary with 1,400 tons of explosives onboard. 
(Dan Kitwood/Getty Images)

From WashingtonPost by William Booth
 
A salvage crew has been working to stabilize the SS Richard Montgomery, which sank a few months after D-Day in 1944 with 1,500 tons of explosives in her holds.

SHEERNESS, England — In the mouth of the River Thames, just downstream from London, a team of elite marine salvors has been busy trying to stabilize one of the world’s most dangerous shipwrecks.

The SS Richard Montgomery, a U.S. military cargo ship, dragged anchor, broke her back on a sandbar, flooded and sank in late summer 1944, a few months after D-Day, with 1,500 tons of explosives still in the holds.
 
 Visualization with the GeoGarage platform (GB raster & ENC)
 
The Monty, as the locals call her, has been disintegrating, slowly but surely, on the shallow seafloor for the last 82 years.

Though a precise manifest is elusive, there are still thousands of World War II munitions inside the Montgomery’s broken hull, including 2,000-pound blockbusters, fragmentation bombs, white phosphorus smoke ordnance and crates of pressure fuzes.

Old bombs, even underwater, can still detonate under the right conditions and one thing certain about the Monty is that her conditions constantly change in the fast-moving tides and shifting sands of the estuary.
“It’s not just the explosives on the wreck that we’re worried about,” said Nolan Conway, who is leading the salvage operation from a high-tech Dutch barge that sits, ever so gently, alongside the wreck.

“It’s the stuff around the barge that we want to be careful about,” Conway said.
“Things could create a sympathetic detonation at the wreck.”

The wreck has stood for all these years as a kind of war memorial, its three masts visible above the waterline, and clearly marked on nautical charts.
You can see it from land.
Tour operators take day-trippers out to circle it.
Local communities remember the sacrifices — the bravery and the terror — of the earlier generations.

The wreck serves as a reminder, too, that the lethal detritus of war often lingers on, whether those ships and bombs are sunk off seaside towns on the English coast or in the Persian Gulf.

The Montgomery is a time capsule and a potential time bomb.
And, after much delay, the British government decided that it was time to do something to prevent a catastrophic collapse of the ship’s degrading steel hull.

On a recent afternoon, Conway, the project leader for Resolve Marine, a company based out of Fort Lauderdale, Florida, was aboard a launch vessel that shuttles crew out to the barge.

His team members are masters of disaster.
A recent assignment had them dealing with a fire in a torpedo room of a submarine operated by the Indian navy.
In 2024, they were hired to disentangle the container ship that struck the Francis Scott Key Bridge in Baltimore Harbor.
They routinely deal with groundings, capsizes, cargo losses, spills and fires.
The trip from the dock to the wreck takes just 20 minutes.

A first impression? How crowded the scene.
The wreck lies a couple miles from shore and just a few hundred meters from active shipping lanes.
Thousands of vessels pass by each year, including bulk freighters, container ships, tugboats, liquefied natural gas carriers and cruise ship bound for London, plus all many of fishing craft and pleasure boat.
The wreck is marked by buoys denoting an off-limits exclusion zone.
The masts were the remains of the derrick cranes that once were used to load and unload its lethal cargo.

The British government initiated the $12 million project to remove the three masts after experts concluded the steel structures might collapse onto the decomposing decks below — and that could be very bad.

Asked if the work on the Montgomery was dangerous, Conway explained how carefully they approached their task — the extensive surveys using radar, magnetometers and sonar to form a deep understanding of the wreck and seafloor around it, which is littered with UXO and PXO, terms of art for unexploded ordnance and potential unexploded ordnance.
Over the years, the wreck has collected additional dangers.
The seafloor around it is dotted with half-buried objects, possibly defensive mines and antiaircraft ordnance left by the British navy and unexploded offensive mines and bombs dropped by the Germans.

To put it very simply, the salvors really don’t want to bump into anything.

A shock wave could set off an explosion, which could be lethal to a diver, or could spark a larger conflagration.
Though some British government assessments remain classified, experts with the Royal Military College of Science estimated in 1970 that a “top event” — meaning essentially the worst case, in which the whole cargo goes kaboom! — could produce a blast of water and debris 3,000 meters high and set off a five-meter tsunami-style wave that could threaten local towns and industry, including nearby oil and gas facilities.

 For 82 years, the masts of the SS Richard Montgomery, a US Liberty ship that broke her back and sank off Sheerness in 1944 while loaded with munitions, stood above the Thames Estuary as the visible warning sign for roughly 1,400 tonnes of unexploded ordnance still sealed in her holds below.
The UK government formally announced plans to remove the masts in 2020, since their continued decay risked destabilizing the wreck and increasing the chance of the cargo shifting or detonating on its own.
Engineers finally began the careful, remotely monitored cutting operation in 2025, lifting sections away from the hull without disturbing the explosives underneath.
The recovered masts are now headed to Chatham's Historic Dockyard, where conservators will stabilize the corroded metal before putting them on public display, turning the most visible symbol of one of Britain's longest-running unexploded ordnance threats into a museum piece.
The wreck remains surrounded by an exclusion zone in the Thames Estuary [Getty Images]
 
The British government, in its reports, has calculated the risk of a top event as “remote.”
Remote, however, does not mean unfathomable, let alone completely safe.

In 1967, salvors attempted to stabilize munitions inside the wreck of another American-built ship, the SS Kielce, that was being used by the Polish navy, which had sunk soon after the war near the English port of Folkestone.

Using mines, the salvors somehow ignited the hold, sparking an explosion equaling a 4.5-magnitude earthquake.

“If the Montgomery were typical salvage job, we would simply rig to the mast, cut the mast using traditional techniques and pull it free — simple and done,” Conway said.
“But because of the sensitive nature of the wreck, we went with special tools.”

Those special tools include a barge that dropped three legs into the silty sand to position itself in place, just inches above the waterline and inches from the wreck.
Divers enter the water one at a time, careful not to step onto the fragile deck below.

“Visibility is measured in inches,” said Jeff Cornish, one of the divers.
“We work by feel.”
To cut the mast, the salvors deployed a diamond wire saw, “because it produces lower vibrations, with a lower potential for something to go wrong,” Conway said.

“We don’t do anything that would upset our mothers,” he added, meaning they plan to come home from work.
The Montgomery was a Liberty ship, designed for convoy duty in the North Atlantic.
She was built at the U.S. Navy’s direction by the St. John’s River Shipbuilding Company in Jacksonville, Florida, and took on her last load of high explosives at Hog Island, today the site of Philadelphia International Airport.

After surviving the ocean crossing and German U-boats, she was destined for an American air base in newly liberated Cherbourg, France.
The Montgomery arrived at what became her grave in late August 1944, two months after the Allied landing on the beaches of Normandy.
People in nearby towns — and those following a live stream online — watched as the masts were removed this week.
The cut masts, covered in rust and sea growth, now lie aboard the salvage barge.
They were in surprisingly robust shape.
But the structures around them were not.
Some of the steel could be knocked apart with a single bang by a hand scrapper.
“It’s paper thin,” Conway said.

David Alexander, an emeritus professor of emergency planning at University College London, has studied the wreck and its cargo.
“It is a rare case,” Alexander told The Washington Post.
In a research paper he wrote that the Montgomery “exemplified a very visible, alarming risk, but one which cannot be analyzed with any degree of sophistication and accuracy because there are simply not enough reliable, accurate data.”

“It’s impossible to say with any certainty what the explosion risk is,” he said, but that the risk is far from zero.
Alexander worried aloud that the wreck lies close to active shipping, that windsurfers have been recorded in the exclusion zone, that divers likely have sneaked out to the site and in years past, fishermen were photographed on the wreck at low tides.
Locals have reported seeing strange lights on the water above the wreck at night, likely phosphorous munitions that shook loose from their crates and ignited on the surface, he said.

In a debate in the House of Lords, and in the London newspapers, fears also have been raised that terrorists or saboteurs, using drones or simple explosives, could target the wreck.
But removing 1,500 tons of explosives might require emptying out local towns or building seawall defenses just in case.
So far, the British government appears satisfied that cutting down the masts — and watching and waiting — are enough.
The masts are now headed to a nearby museum and conservatory at the Historic Dockyard in Chatham, where the structures will be restored and displayed to the public, and the story of the Monty will continue.
“There’s a lot of interest in that ship,” said Paul Barnard, deputy CEO of Chatham Historic Dockyard Trust.
“A lot of history, and lore, and speculation, and to be honest, rumor, about what could happen next.”
 
Links :

Monday, October 5, 2026

Why AI has trouble predicting the intensity of hurricanes

A Pacific Ocean wave crashes into a seawall protecting beachside homes on Sept.
7 in Long Beach, California.
(Mario Tama/Getty Images)

From WashingtonPost by Chanh Kieu (associate professor of atmospheric science at Indiana University)

Artificial intelligence now produces excellent weather forecasts on a broad scale.
But several factors complicate its ability to predict how fast a hurricane will ramp up.
 
 Artificial intelligence has revolutionized weather forecasting in just a few years, with global AI weather models now able to produce forecasts that rival some of the world’s best physics-based prediction systems.

This remarkable progress has been driven by three factors: massive amounts of weather data, advances in AI models and unprecedented computational power.
While most of the current discussion on improving AI for weather focuses on the models or new hardware, the data is crucial.

At the global scale, AI has benefited from decades of climate and weather records covering the entire Earth.
These datasets contain millions of examples of how atmospheric conditions evolve with time, allowing AI models to learn patterns in a way that would have been impossible a decade ago.

But when you zoom in from the global scale to the regional scale, forecasts start to be much more challenging for AI.

That difference matters for forecasting hurricane intensity.

The world has seen many hurricanes rapidly intensify in recent years, strengthening from relatively weak storms into destructive monsters in a matter of hours.
Hurricane Polo did it off Mexico’s Pacific coast as it rapidly strengthened from a tropical storm on Sept.
21 to a powerful Category 5 hurricane in 24 hours.
Polo quickly became one of the strongest Pacific storms in decades, with winds reaching 180 mph.

When rapid intensification surprises forecasters — as Hurricane Michael did in 2018 when it grew into a destructive Category 5 hurricane right before hitting Tyndall Air Force Base and Mexico Beach, Florida, in 2018 — communities can be left with too little time to evacuate and prepare.

Why AI struggles with detail
 
Damage is shown in the Baja California Sur state of Mexico after Hurricane Polo, one of the strongest Pacific storms in decades.
(Eduardo Verdugo/AP)
 
Unlike global weather forecasts, hurricane intensity forecasts are often considered a regional forecasting problem.
Regional forecasts are often concerned with extreme events, such as heavy rainfall, squall lines, severe thunderstorms or hurricanes.

These extreme events often develop rapidly or move quickly over short periods of time.
Capturing such behavior in AI models requires data in much greater detail than current global datasets can typically provide.

When scientists train AI models to predict hurricane intensity, they usually rely on two sources of data.

The first is observations, which include measurements of rainfall, near-surface temperature, wind speed and other weather variables collected from weather stations, radars, buoys and satellites.
Such direct observations can be detailed, but they are often limited to near-coastal regions and unevenly distributed, while many of the most important stages of hurricane development occur over the open ocean where direct observations are sparse.

Modern satellites can help fill some of these gaps in the open ocean, but they can help estimate only part of the rainfall, surface winds or cloud-top temperatures due to limits in satellite coverage.
In particular, they cannot simultaneously scan a complete three-dimensional structure of every hurricane around the globe.
At present, even the best observational systems provide only a partial view of hurricanes at any point in time.

Scientists used satellite data to map aspects of Hurricane Matthew’s intensity as it gained strength on its way toward the Florida Panhandle in 2016.

The second source of training data comes from weather model simulations, which combine atmospheric conditions and knowledge of physics to provide the most complete three-dimensional picture of the atmosphere at high resolution.

However, this simulated data is not perfect either, because all computer models contain approximations and uncertainties arising from incomplete knowledge of the Earth’s atmosphere.
As such, there are always fine-scale processes that model simulations cannot capture.

So, we simply don’t have a good full, three-dimensional dataset to train an AI model for hurricane intensity prediction at present.

The other challenge: Chaos

A satellite view of Hurricane Erin tracking westward across the Atlantic in 2025. 
(Gallo Images/Gallo Images via Getty Images)

But data is not the only issue for AI hurricane prediction.

Suppose in the future scientists could measure the entirety of a storm, measured every second, for thousands of storms around the world.
That still might not allow AI to predict hurricane intensity perfectly.
The reason: chaos.
Tiny differences in the initial state of a hurricane can quickly grow over time.

My recent research with colleagues suggests that hurricanes may contain some element of chaos that can prevent AI models from accurately predicting hurricane intensity at long forecast times.

Once embedded in a favorable environment, a tropical storm can intensify toward a maximum possible strength.
Scientists call this upper limit the potential intensity.
It is determined primarily by the surrounding environment.
For example, warm ocean water can fuel a hurricane’s intensity, or wind shear can slow down a hurricane’s development.
If the ocean temperature rises, the potential intensity of a hurricane increases, too.

Any small disturbances will also cause the hurricane intensity to fluctuate.
The warmer the ocean surface, the more fluctuations.

Computer simulations show how forecasting a hurricane’s intensity becomes more uncertain the further out the forecast is.
Hurricane intensity is measured as the maximum wind at the height of 33 feet (10 meters).
The images show how much two hurricanes can differ in their physical appearance, even when they have the same Category 5 intensity.

Recent studies have proposed that these fluctuations are not purely random but occur within what is known as a chaotic attractor — a set of possible storm states within which the hurricane can evolve unpredictably.
Although the existence of such a chaotic intensity attractor has not yet been fully established, it presents a fundamental dilemma for training AI models to predict hurricane intensity.

On one hand, scientists want AI models to make the most accurate predictions possible.
Thus, during training the goal is to minimize the difference between the forecast and what actually happens until an AI model achieves the smallest possible error.

On the other hand, we also want the AI model to capture the hurricane’s intrinsic chaos.
But if an AI model can capture this chaos, then its error cannot be reduced indefinitely.
So, an AI model trained to minimize forecast error may learn the most likely evolution of a hurricane while smoothing out unpredictable fluctuations.
In this regard, these two goals compete with one another.

With data always containing some uncertainty, the rules an AI model learns are only approximated.
The accuracy of hurricane intensity forecast will therefore get worse after just a few days.

What will it take to make better forecasts?

The challenge for AI models to predict hurricane intensity is not just about obtaining more data, building better neural networks or deploying faster computers.
It is also about understanding hurricane behaviors and how chaos in intensity emerges.

Both dictate whether AI models can learn what is predictable and what is unpredictable.
That distinction not only puts a cap on our current hurricane intensity forecast accuracy but also determines the next generation of weather forecasting and evaluation systems, which should focus on a range of possible hurricane intensities and their probabilities instead of a single intensity number.