The dangerous places keep getting less lonely.
On the volcanic island of Vulcano off Sicily's coast, researchers from the Technical University of Munich have deployed an autonomous drone system that uses laser tomography to map volcanic gases in three dimensions — doing quietly and precisely what once required a scientist to stand at the edge of a crater. The system reads the chemical breath of an active volcano with roughly five percent accuracy, translating the fading of a laser beam into early warning signals that seismic instruments alone cannot provide. It is a small machine doing consequential work: placing a layer of intelligence between a restless mountain and the communities that live in its shadow.
- In November 2021, Vulcano's daily CO2 output leapt from 80 to 480 tons almost overnight, forcing 250 residents to evacuate in the dark while struggling to breathe — a reminder that the mountain's silence is never guaranteed.
- Traditional ground-level gas readings are contaminated by vegetation and soil, leaving volcanologists with blurred signals precisely when clarity matters most.
- A laser on a ground cart tracks a lightweight drone carrying only a reflector, measuring gas concentration through beam absorption across the plume — collecting up to 3,000 data points per fifteen-minute flight and assembling them into a 3D chemical map.
- The critical signal is a ratio: when CO2 climbs sharply relative to SO2 and then falls, the pattern has historically preceded volcanic activity — a warning curve that seismic sensors cannot draw.
- The team has proven the instrument works autonomously and accurately, but no eruption has yet been predicted with it — the next threshold is a permanent, self-flying monitoring station guided by AI interpretation.
On Vulcano, a small volcanic island off Sicily, a five-and-a-half-pound drone named Tina has begun doing work that once demanded a volcanologist's physical presence near an active crater. Researchers from the Technical University of Munich deployed the aircraft in the first fully autonomous use of laser-based gas tomography over a live volcano — measuring the chemical signature of the mountain's exhaled gases with roughly five percent error, and doing so without placing a human body in danger.
The method divides the labor elegantly. A ground-mounted laser automatically tracks the drone as it flies a pre-programmed grid, while the aircraft carries only a reflector. As the beam crosses the volcanic plume, it weakens — and that fading is the measurement. More gas means more absorption, means a stronger signal. In ten to fifteen minutes, the drone collects up to 3,000 readings, which an algorithm assembles into a three-dimensional map showing gas concentration at every altitude. Lifting the measurement into the air also solves a longstanding problem: ground sensors pick up CO2 from soil and vegetation, obscuring what the volcano itself is actually releasing.
What makes the system potentially transformative is not the measurement alone but the ratio it tracks. The relationship between carbon dioxide and sulfur dioxide follows a recognizable arc before volcanic activity — the ratio rises sharply, then falls. That curve, read over time, offers a warning window that seismic sensors cannot provide. The stakes are vivid on Vulcano: in November 2021, the island's CO2 output jumped sixfold in a matter of days, residents reported breathing difficulties, and the mayor ordered roughly 250 permanent residents to evacuate their homes at night.
The field has been moving toward this capability for years, with drones previously flying spectrometers through plumes to sample single points. Laser tomography maps the entire cross section instead — the difference, as one framing has it, between a thermometer and an MRI. The USGS has similarly flown uncrewed aircraft over Kilauea's erupting summit, reaching positions no crew could safely occupy.
Still, one campaign and a method paper do not yet constitute an eruption forecast. What the team has demonstrated is the instrument: autonomous flights converting laser absorption into gas maps that once required careers of risky fieldwork. The stated roadmap is full automation paired with AI interpretation — a monitoring station that flies itself, maps a gas cloud in fifteen minutes, and flags a dangerous ratio shift without human intervention. The day such a station issues a warning before an evacuation order, this stops being a research story.
On a volcanic island off Sicily's coast, a small drone named Tina has begun doing work that used to demand a volcanologist's life. The Technical University of Munich sent the five-and-a-half-pound aircraft over Vulcano's crater in what amounts to the first fully autonomous deployment of a laser-based gas mapping system—a setup that reads the chemical signature of an active volcano with roughly five percent measurement error, and does it without putting a human body anywhere near the danger.
The method is elegant in its division of labor. A laser mounted on a ground cart automatically tracks the drone as it flies a pre-programmed grid at distances up to 197 feet, carrying nothing but a reflector. The laser beam weakens as it crosses the volcanic gas between them. That fading signal is the entire measurement: more gas in the path means more absorption, means a stronger warning. The drone, which can operate at altitudes up to 9,800 feet, collects as many as 3,000 measurements per flight in ten to fifteen minutes. An algorithm then converts the absorption data into a three-dimensional map, showing not just where the gas is but how concentrated it sits at each altitude.
This solves a problem that has frustrated volcanologists for years. Ground-level gas readings pick up carbon dioxide from surrounding vegetation and soil, not just from the mountain itself. Lifting the measurement into the air isolates what the volcano is actually exhaling. The researchers—led by Marius Schaab and including Prof. Achim Lilienthal of the TUM robotics institute, volcanologist Nicole Bobrowski of Heidelberg University, and Prof. Thorsten Hoffmann of Johannes Gutenberg University Mainz—published their underlying method in IEEE Sensors Letters in September 2025.
What makes this system potentially transformative is not the measurement itself but the ratio it reveals. The relationship between carbon dioxide and sulfur dioxide in volcanic gas follows a recognizable pattern before activity: the ratio climbs sharply, then falls. That curve, read in time, gives authorities a warning window that seismic sensors alone cannot provide. The stakes are concrete on Vulcano. In November 2021, the island's daily carbon dioxide output jumped from eighty tons to 480 tons. Residents reported trouble breathing. The mayor ordered the island's roughly 250 permanent residents to evacuate their homes at night. The Grand Crater last erupted in the late nineteenth century, but the mountain never stopped exhaling.
The field has been building toward this moment for years. In November 2020, drones were still flying miniaturized spectrometers through volcanic plumes, sampling single points. The laser tomography approach maps the entire cross section instead—the difference between a thermometer and an MRI. The USGS Hawaiian Volcano Observatory flew uncrewed aircraft over Kilauea's erupting summit crater last year, mapping lava and tephra from positions no geologist could occupy. A DJI Matrice 600 with a thermal camera hovered directly over an active vent for close-up imaging that a helicopter crew would never attempt.
But one island, one campaign, a five percent error margin, and a method paper do not yet constitute an eruption forecast. Nobody has predicted an actual eruption with this rig. What the team proved is the instrument itself: autonomous flights turning laser absorption into gas maps that used to require careers of risky fieldwork. The CO2-to-SO2 ratio is established volcanology. The bottleneck was always getting clean, repeated measurements without walking scientists into a crater. A system that does it every day, in any weather the drone tolerates, changes the data density that forecasting models consume.
Prof. Lilienthal stated the roadmap plainly: automate the measurement and mapping processes, and have artificial intelligence interpret the data. A monitoring station that flies itself, measures an entire gas cloud in fifteen minutes, and flags the dangerous ratio shift without human intervention is the version of this technology a civil protection agency would actually deploy. The pattern extends beyond volcanoes. Drones have flown into hurricanes for NOAA, chased data around severe storms, surveyed the seafloor, and searched for survivors in earthquake rubble. The machines keep getting more useful and more adaptable, and the dangerous places keep getting less lonely. Watch whether the team's automation goal turns into a permanent monitoring installation on Vulcano or a sister volcano. The day one of these stations flags a ratio spike before an evacuation order, this stops being a research story.
Citações Notáveis
Our goal is to automate the measurement and mapping processes and have artificial intelligence interpret the data.— Prof. Achim Lilienthal, TUM MIRMI robotics institute
A Conversa do Hearth Outra perspectiva sobre a história
Why does the laser approach matter more than just putting sensors on the drone itself?
Because the laser reads the entire gas cloud at once, not just the single points a sensor on the aircraft touches. It's the difference between a thermometer and an MRI. You get a three-dimensional picture of concentration at every altitude, not a snapshot from one path through the plume.
And the CO2-to-SO2 ratio—that's the actual warning signal?
Yes. Volcanologists have known for years that this ratio shifts in a recognizable pattern before activity. It climbs sharply, then falls. But getting clean, repeated measurements of that ratio without sending people into the crater has always been the hard part. This system can do it every day.
So why hasn't this prevented an eruption yet?
Because one island, one campaign, and a method paper don't equal a forecast. What they proved is that the instrument works. The real test comes when a permanent station flags that ratio spike before authorities have to evacuate people. That's when this becomes operational, not research.
What happens to the volcanologists then?
They don't disappear. But the dangerous work—the fieldwork that used to demand someone stand at the crater rim—that gets replaced by a machine that flies itself. The human work shifts to interpretation, to building the models that turn data into warnings.
Is this the first time a drone has done anything like this on a volcano?
No. USGS flew uncrewed aircraft over Kilauea's erupting summit last year. Drones have been sampling volcanic plumes for years. But this is the first fully autonomous deployment of a laser tomography system that maps the entire gas cloud in three dimensions without a human in the loop.
What's the real bottleneck now?
Automation and AI interpretation. The team wants the station to read the data itself and flag dangerous patterns without waiting for a human to analyze it. That's the version a civil protection agency would actually buy and run continuously.