Blog

When Anak Krakatau started throwing ash into the sky again this month, the first calls that went out in Indonesia were about airports and ferry routes and whether Jakarta could keep flying. Nearly three thousand flights were disrupted. Ash drifted across five provinces. The country’s volcanology agency pushed the alert to the second highest of its four levels and drew a three kilometre ring around the crater that nobody is allowed to cross.

But there was a second set of phone calls happening at the same time, quieter, and about something much smaller than an airport. Roughly fifty kilometres from that crater, across a stretch of open water, sits a peninsula of lowland forest and swamp called Ujung Kulon. Every wild Javan rhinoceros left alive on the planet lives there. Not most of them. All of them. One park, one population, one very bad day away from a species ending.

So while the ash was still falling, park staff went out to check on the animals, and the answer came back reassuring. Ardi Andono, who runs Ujung Kulon National Park, said the rhinos were unharmed. Riszki Is Hardianto, a wildlife researcher at the Indonesian watchdog group Auriga Nusantara, said no rhino deaths had been reported either. The volcano had not touched them.

That is the good news, and it is real, and it is also where the story gets complicated, because the harder question is not whether the rhinos survived the eruption. It is how many rhinos there were to survive it in the first place. And on that, the people who watch this species most closely do not agree at all.

The Indonesian forestry ministry’s figure sits in the range of eighty seven to a hundred animals. Save the Rhino International, one of the conservation organisations that has worked on this species for decades, puts the number at around fifty. Those are not two slightly different estimates of the same thing. That is a gap big enough to swallow the entire species twice over, and it exists because of an argument about what counts as a rhino you can count.

Here is the part that stays with you. In a report published a few years ago, Auriga Nusantara went through the official population tally and found something strange in the arithmetic. Eighteen animals in the official count, nine males and nine females, had not been photographed alive by a camera trap in years. They were still on the list. The system was built so that once an animal had been recorded, it stayed recorded until somebody produced proof that it had died. No body, no deletion. Of those eighteen, the report said, three were in fact already confirmed dead, a female some years earlier and a male and a female after that, and none of those deaths had been publicly announced, even though new calves were being announced regularly.

Sit with the shape of that for a second. It is not fraud in the dramatic sense. It is something more ordinary and in some ways more worrying: a counting rule that only ever pushes the number up. Births get added. Deaths get added only when somebody stumbles across a carcass in a rainforest that swallows things whole. Under that rule, a population can look like it is slowly growing at exactly the moment it is quietly shrinking.

The environment ministry did not dismiss the finding outright. It acknowledged at the time that fifteen rhinos had no recent camera trap sightings, and it acknowledged indications of poaching in an internal circular. What it did not do was open the underlying data for outside scrutiny, which is how you end up with two numbers in public and no way for anyone outside the system to reconcile them.

Then came the part that turned an accounting dispute into something much darker. Police investigating a wildlife trafficking network reported that poachers had claimed responsibility for killing twenty six Javan rhinos over a period of several years. If that claim is anywhere near accurate, it accounts for something close to a third of everything the official count said was out there. Six suspects were arrested. Others were still being sought, including at least one person who appeared to have access to information about where rhinos had recently been recorded, which is not a detail you want in a story about a park with a single population of anything. A collector who had paid roughly thirty thousand dollars for horns was picked up. One convicted poacher received twelve years, reportedly among the heaviest sentences Indonesia has handed down for a wildlife crime.

For thirty years before that, Ujung Kulon had reported essentially no poaching. It had been held up as the one place where this had been solved. The rhinos were behind a wall of remoteness and patrols and the simple fact that almost nobody knew how to find them. Then somebody did.

This is why the volcano matters even though the volcano did no harm. Anak Krakatau is not a hypothetical. It is the child of the 1883 eruption, a new cone that pushed itself permanently above the waterline in 1930 and grew for most of a century. Then one night, a chunk of its own flank slid into the sea. The tsunami that followed killed more than four hundred people along the coasts of the Sunda Strait and injured many thousands more. The volcano lost around two thirds of its own volume in the process, dropping from well over three hundred metres to barely a hundred. It has been rebuilding ever since, and it now stands somewhere around a hundred and fifty metres again.

Nobody planned for that. The warning systems of the time were built to detect tsunamis generated by earthquakes, and a mountain collapsing sideways into the water does not announce itself with a seismic signature anyone was watching for. The wave arrived at coastal towns with no siren in front of it.

Those same coasts include the shoreline of the peninsula where the rhinos live. And rhinos, for all their armour plating, are large animals in low, flat, coastal forest. Modelling of tsunami hazard for the park has been done and published. The scenarios are not comforting. You do not need the whole population to be swept away for it to be over, because a species this small does not have the numbers to absorb a bad year.

Which brings us to the one decision that would fix this, and to a rhino called Musofa.

The logic is obvious to anyone who has ever been told not to keep all their savings in one account. If every individual of a species lives in one place, then anything that can happen to that place can happen to the species. Disease. A tsunami. A fire. A shift in vegetation. So for years the plan has been to establish a second, managed group, somewhere those animals can be watched, treated, bred and protected without every risk being shared by every animal at once.

Late last year, for the first time in the history of Javan rhino conservation, they tried to move one.

Musofa was a male. He walked into a pit trap on a Monday evening. Two days later he was inside the Javan Rhino Study and Conservation Area, a managed and fenced site, where teams planned to monitor him intensively and, eventually, to support breeding. It was framed as a step toward stabilising the population and giving the species a more structured future.

He died two days after that.

The necropsy told a story nobody had known while he was out in the forest. Musofa had chronic disease affecting his stomach, his intestines and his brain, along with a heavy parasitic burden and extensive degeneration in his organs. He also carried old wounds from fights with other rhinos, though those were not identified as what killed him. He had, by the look of it, been a sick animal for a long time. The forest had simply never told anyone.

It is worth being careful about what that death does and does not prove. It does not prove that translocation kills Javan rhinos. One animal is not a trial. What it does show, brutally, is how little anyone can know about the health of individuals in a population that is monitored almost entirely through motion triggered cameras in dense vegetation. A camera trap can tell you an animal was alive at a moment. It cannot tell you that the animal has parasites in its brain.

And there is a correction worth making here, because the story has often been told as though moving Musofa was the attempt to give the species a second home. It was not, quite. The site he was moved to sits within the same national park. It is a fenced and managed area inside the existing range, not a separate landscape somewhere else on the island. It is a step toward the eventual goal rather than the goal itself. The single habitat problem that the eruption highlighted is still, at this moment, unsolved. All of the rhinos are still in one forest.

The forest itself has its own slow problem. An invasive palm called arenga has been spreading through parts of the park for years, growing thick and shading out the low plants that rhinos actually eat. Conservation teams have spent enormous effort cutting it back. It is the kind of threat that does not make headlines because nothing dramatic happens on any given day. The habitat just gets a little less edible, year over year.

There is also a problem that no amount of patrolling fixes, and it is written into the animals themselves. Geneticists looking at this species have suggested that something on the order of a hundred individuals would be needed simply to hold on to the genetic diversity that is left. If the true figure is closer to fifty, the population is already below the line where diversity can be maintained, which means that even a group that stops being killed and stops being crowded out of its food can keep losing something. Inbreeding does not announce itself. It shows up decades later as calves that do not thrive and immune systems that cannot handle a pathogen the population has never met.

That is the quiet reason the second site matters so much more than it sounds like it should. It is not only about spreading the risk of a wave or a disease outbreak. It is about creating a situation where animals can be seen, weighed, treated and paired deliberately, rather than hoped about from a distance. Musofa is the argument for it and the warning about it at the same time.

Everything anyone knows about these rhinos comes through cameras strapped to trees in humid forest, triggered by movement, producing frames that then have to be matched to individuals by the shape of skin folds and horn and scars. It works remarkably well for a species that will not tolerate being followed. It also means the data is a series of glimpses, and every conclusion drawn from it is a conclusion about glimpses. When a rhino stops appearing, there are two possible explanations, and only one of them is that the camera missed him.

There is a version of this species’ history that makes all of it feel inevitable. Javan rhinos once ranged across a huge sweep of Asia. There was a mainland population too, in Vietnam, hanging on in a national park, and for a while it looked like it might survive. Genetic work eventually established that the group had dwindled to a single animal. That animal was found dead, shot, its horn taken. The mainland subspecies was declared gone. That is what the end actually looks like: not a bang, but a slow narrowing until there is one, and then none.

So what is different this time, and is anything?

A few things, honestly. The camera trap network has recorded calves in recent years, which means animals are still breeding in the wild without any help. Patrols were stepped up sharply after the poaching case broke. The prosecution that followed carried real prison time, which matters in a business where the calculation has usually been that the penalty is survivable. Independent groups like Auriga are now auditing the official numbers in public, which is uncomfortable for everyone involved and is also probably the single most useful thing happening, because a conservation programme that cannot count its animals cannot know whether it is working.

And the response to the eruption was, in its own way, a good sign. The volcano flared, and within days there were people in the field checking on fifty or eighty or a hundred animals, and two separate organisations, one government and one independent watchdog, both said the same thing in public: no rhino deaths. Those two groups disagree profoundly about the population. They did not disagree about this.

The rhinos, for what it is worth, would have felt the eruption as sound and grit and a change in the light. They are not animals that panic easily. They are solitary, mud loving, thick skinned browsers that spend their lives moving quietly through dense cover, and they have been doing it in that peninsula since long before anyone was counting them.

The ash has mostly settled. Anak Krakatau is still on alert, still ejecting material in bursts, still watched constantly by instruments that did not exist the last time it did something catastrophic. Officials have been urging people to ignore the rumours circulating about imminent tsunamis and to follow the official channels instead, which tells you something about how raw the memory still is along that coast.

Somewhere on the other side of that water, in a forest with an invasive palm problem and a poaching problem and a counting problem, there are perhaps fifty rhinos, or perhaps a hundred, and nobody can tell you with confidence which. They survived the volcano. The thing they have not survived yet is the fact that there is only one place left where they can be.

The fires making the news in Indonesia right now are the ones you can photograph. Orange light in the trees, helicopters with water buckets slung underneath, firefighters in yellow standing in a field of black stumps. Those are real, and they are bad, and they are not the ones that scientists are most worried about.

The worst fires there are underground. They burn down through metres of dried-out peat, moving slowly sideways through the soil like a cigarette working its way along a mattress, and they put out almost no flame at all. You can stand on ground that is on fire. You can drive past a landscape that is burning and see nothing except haze. Satellites built specifically to spot fire from orbit look down at these and, a lot of the time, register nothing, because they are trained to find heat at the surface and a peat fire does not give them a hot spot to find.

And once one starts, there is more or less no putting it out. Water goes onto the surface and the fire is not at the surface. Firebreaks stop a fire that travels above the ground and this one is travelling below it. The practical answer, the one the people who study this give without much softening, is that it burns until the rains come.

Right now the rains are not coming, and that is the whole crisis in one sentence.

A strong El Nino has settled over the Pacific, and it has been joined by a positive phase of the Indian Ocean Dipole, which is the equivalent weather pattern in the ocean to Indonesia’s west. Those two things together do something very specific to the region: they make the dry season drier and longer. Robert Field, a researcher at Columbia University who built the Global Fire Weather Database, has described the current fire activity as tracking sharply upward in a way that looks a lot like 2015, and has pointed at the El Nino as the reason the dry season is biting harder than usual.

In early August, roughly ninety percent of the country got little or no rain at all. On one day at the end of that month, satellites logged nearly a thousand active hotspots. By the first days of September, only about three weeks into the fire season, the fires had already released something in the region of ten percent of the total greenhouse gas emissions that came out of the entire catastrophic 2015 season. Three weeks in. Ten percent of a full disaster.

To understand why this particular kind of fire is so much worse than a normal forest fire, you have to know what peat actually is. It is not soil. It is thousands of years of dead plants that fell into waterlogged ground and never fully rotted, because the water kept the oxygen out. Layer on layer on layer, stacked up over millennia, in some places many metres deep. It is, functionally, an enormous underground store of carbon that has been sitting safely underwater since before recorded history.

Wet peat does not burn. Dry peat burns extremely well. It is essentially fuel, and it has been accumulating for a very long time.

Indonesia holds about thirty-six percent of the world’s tropical peatlands, concentrated across Kalimantan, Sumatra and Papua. That number alone tells you why what happens there is not a local story.

The thing that turns a safe carbon store into a fuel store is drainage. Through the 1990s, huge areas of peat swamp were cut through with canals to drain the water off, initially in an effort to convert the land to rice growing, later to make room for oil palm and other plantations. The rice project failed. The canals stayed. A drained peatland is a peatland with its fire protection removed, and in a dry year it becomes something you could not design better if you were trying to build a fire that cannot be extinguished.

So when a fire is lit at the surface, and most of them are lit deliberately as the cheapest way to clear land, it does not stay at the surface. It goes down.

It is worth being precise about why land gets burned, because the lazy version of this story blames whoever is holding the match and stops there. Fire is used to clear land because it is close to free. Machinery is not free. Labour is not free. Burning off a field costs a box of matches and an afternoon, and it also leaves behind ash that acts as fertiliser. That calculation is the same for a family with two hectares and for a company with twenty thousand, and it has been the same for a very long time. What changed was not the tool. What changed was the ground underneath it, when drainage turned wetland into something that catches.

The physics of what happens next has its own name. Smouldering combustion is fire without flame: a slow, oxygen-starved reaction that creeps through porous fuel at relatively low temperature and sustains itself for as long as there is dry material to work through. It is the same process that keeps a lit cigarette going in an ashtray with nobody drawing on it. Scale that up to a carbon deposit several metres deep and hundreds of square kilometres wide and you have a fire that can run for months, drop below the reach of anything poured on top of it, travel under a road and come up on the other side, and produce vastly more smoke per unit of fuel than open flame does, because incomplete burning is exactly what makes smoke.

That is where the measurement problem comes in, and the measurement problem is not a technicality. It is arguably the heart of the story.

Fire detection from space works by looking for thermal anomalies, patches of ground that are much hotter than they should be. The instruments doing that work over Indonesia include MODIS, on NASA’s Aqua satellite, and VIIRS, on newer platforms. Both are very good at what they were built to do. Neither was built for this. A smouldering peat fire under an intact forest canopy can be hot enough to be destroying an ancient carbon store and cool enough at the surface, and hidden enough beneath leaves, that the sensor passes over without flagging it.

Then there is the second layer of the problem, which is almost perverse. Mark Cochrane, an ecologist at the University of Maryland Center for Environmental Science, has made the point that the very worst smoke events are the hardest ones to see from orbit. The smoke is thick enough to block the satellite’s view of the fires producing it. The situation gets worse, and the data gets thinner, at exactly the same time.

Sit with that for a second. The official numbers, the hotspot counts, the burned-area figures, the emissions estimates, are all downstream of instruments that systematically lose sensitivity as conditions deteriorate. Which means the official numbers are close to certain to be an undercount, and the undercount is worst in the worst moments.

Shi Jun Wee, a graduate student at the University of Maryland, is working on the fix. Together with NASA and the MapBiomas mapping initiative, the project is building detection algorithms aimed specifically at fires burning under canopy, the ones the standard method misses. It is a good example of how unglamorous a lot of climate work actually is. No dramatic rescue, no new machine. Somebody sitting with satellite data trying to teach software to find a fire that does not look like a fire.

The reason it matters is that you cannot manage what you cannot count. A government cannot send crews to a fire it does not know exists. An emissions inventory that misses the underground component gives a country a cleaner-looking ledger than reality supports. And a peat fire that nobody has logged is one that gets to keep burning down through a carbon store for weeks without anybody deciding whether to intervene.

The emissions difference is not marginal. Compared with other tropical forest fires, peat fires have been found to produce roughly three times as much fine particulate matter, five times as much sulphur dioxide, three times as much organic carbon and about twice as much methane and carbon monoxide. Everything bad about smoke, multiplied.

Fine particulate matter is the number that reaches into people’s lungs. These are particles small enough to pass through the body’s filters and into the bloodstream, and they are the reason a peat fire hundreds of kilometres away is a medical problem rather than an inconvenience.

Across Sumatra, Kalimantan, Sulawesi, Java and Papua, that is what has been happening. The Indonesian health ministry has reported that around 3.4 million people are at risk of respiratory infection. Clinics in affected areas have described handling fifty to seventy patients a day with persistent coughs and raw throats. Pregnant women and young children are the groups flagged as most vulnerable, which is the usual and grim pattern with air pollution, because small bodies process more air relative to their size and developing lungs do not recover from insult as reliably as adult ones.

Millions of schoolchildren have been pushed onto remote learning. Nine national parks have closed. Flights have been delayed. In West Kalimantan, which has had some of the worst of it, residents have described conditions unlike anything in their living memory, with fire reaching the edges of everything they had.

The domestic detail that has stuck with people is the ash. One mother in an affected area described sweeping every single day and the ash simply returning, and her children being ill on and off for a month straight. That is what a peat fire season is at ground level. Not a wall of flame. A grey film on every surface, over and over, and a child who does not get better.

The volunteers fighting it have been open about the conditions. One described the situation as very serious and listed the problems plainly: not enough water, visibility ruined by haze, and shortages of the protective equipment and basic supplies crews need to work in smoke all day. The national government has moved, deploying dozens of helicopters and aircraft, water pumps and thousands of personnel, and distributing more than 260,000 face masks. A mask is a reasonable thing to hand someone. It is also an admission that the air itself is now the hazard and that nobody can switch it off.

The haze has not stayed inside the country’s borders, either. It has drifted across the region, which is what has made Indonesian fire seasons a recurring diplomatic problem for decades rather than a domestic one.

All of this has happened before, which is the part that makes 2026 feel less like an accident and more like a test.

1997 was catastrophic. 2015 was catastrophic. The 2015 season released around 1.75 billion tonnes of greenhouse gas equivalents and did an estimated 16.1 billion dollars of economic damage. A study by researchers at Harvard and Columbia estimated that the haze from that season was linked to roughly 100,300 premature deaths across the region, a figure that regional governments disputed at the time and which remains a modelled estimate rather than a body count, but which gives a sense of the scale at stake when the air over a populated region becomes unbreathable for weeks.

After 2015, Indonesia did act. There was a serious push into peatland restoration, and the centrepiece of it was blocking the canals, which is exactly the right instinct: if drainage is what made the peat flammable, then rewetting it is what makes it safe again. There was work on wetland restoration, on building firefighting capacity, and on preventing accidental ignitions. Some of it demonstrably worked. Peat restoration has been shown to reduce fire risk where it has been carried out properly.

The trouble is the scale of what had to be undone. Decades of canals cut across an area the size of a small country cannot be reversed in a decade, and rewetting peat is slow, expensive, politically awkward work that runs directly against the interests of whoever is farming the drained land. In the first half of 2026, before the season even properly started, around 100,000 hectares burned, and roughly half of that was peatland. That is the reform’s report card, and it is a mixed one. The system is better than it was. It is not better enough for an El Nino.

There is one more thing about peat fires that makes them different from almost every other disaster in the news, and it is the timescale.

When a normal forest burns, the carbon released came out of the atmosphere over the lifetime of those trees, decades, maybe a century or two, and the forest can pull some of it back in as it regrows. When peat burns, the carbon coming out went into the ground over thousands of years. It cannot be put back on any timescale that means anything to a person alive today. Peat accumulates at something on the order of a millimetre a year. A fire can take a metre of it in a season.

That is not damage. That is deletion.

So what happens next is, uncomfortably, out of anybody’s hands in the short term. The fires below the surface will keep going until the wet season arrives in force and the water table comes back up, and then most of them will stop, and the smoke will clear, and the flights will resume, and the children will go back to classrooms. The headlines will end. The canals will still be there.

The longer game is the one being played by people like Wee, sitting with satellite data trying to teach an algorithm to see a fire that gives off no visible flame, and by the restoration crews damming canals one at a time, and by the researchers who keep making the same point to anyone who will listen: the ground itself is the problem, and until the water goes back into it, every strong El Nino is going to do this again.

Somewhere under a stand of trees in Borneo right now there is ground that looks completely ordinary. No smoke, no light, nothing to photograph. A metre down, something laid there before any of this was a country is quietly going up, and it has been doing that for weeks, and nobody has written it down.

The submarine was going home when it died. It had given up on its mission, turned back towards Germany, and was crossing a stretch of shallow water off the Danish coast when it ran into a British mine. Twenty-six or twenty-seven men went down with it, depending on which record you read. So did eighteen sea mines and six torpedoes that had never been used.

For ninety-nine years, nobody knew exactly where it was. Then a Danish diver found it, sitting on the seabed in water shallow enough that you can reach it and cold enough that it has not rotted away. And a team of scientists went out to it with steel cages full of mussels, because they wanted to answer a question that nobody had thought to ask for most of a century.

The question was simple. The explosives are still in there. Where are they going?

The answer, published in a marine pollution journal by researchers from Aarhus University working with colleagues in Germany, is that they are going into the water, into the seabed, and into the animals living on the hull. Not in dramatic quantities. Not in a way that would hurt anyone swimming past. But measurably, continuously, and in a way that is going to get worse rather than better, because the thing holding it all in is a steel shell that has been dissolving since the year it sank.

The wreck is UC-30, a German minelaying submarine of a type built in a hurry during the First World War. It was a small boat by any standard, under fifty metres long, with a crew of around twenty-six and a job that was mostly invisible. Minelayers did not hunt ships in the way the famous U-boats did. They crept into shipping lanes, dropped mines through vertical tubes in the bow, and left. Whatever happened afterwards happened to somebody else, days or weeks later, with no submarine anywhere near.

UC-30 was commissioned in the summer of 1916 and lasted less than a year. Over four patrols it sank five vessels, three merchant ships and two requisitioned trawlers. Its commander, a man in his early thirties named Heinrich Stenzler, went down with the boat. His body came ashore in Denmark and was eventually identified and taken home to Germany for burial, which makes him one of the very few of that crew to have a grave anywhere at all. The rest are still where they were.

That matters for what happens next, and it is worth saying early. This is a war grave. It is not a dumping ground that somebody can simply lift and take away, and any conversation about cleaning it up has to sit alongside the fact that the men who loaded those mines are still inside the hull.

The wreck was rediscovered in 2016, off the Danish coast near the island of Romo, by Gert Normann Andersen, a diver who has spent a working lifetime locating lost ships in these waters. It lies in an area of the North Sea that is busy with fishing, shipping and, increasingly, offshore construction, and it was catalogued as a historical find rather than an environmental one.

The study that changed that was led by Kathrine Juul Andresen, a geoscientist at Aarhus University, with collaborators including a toxicologist at the University Medical Center Schleswig-Holstein and researchers connected to Germany’s Alfred Wegener Institute. The fieldwork was not straightforward. Civilian diving on the wreck is not permitted, for reasons that become obvious when you remember what is stacked inside it, so the sampling was carried out by divers from the Royal Danish Navy.

They took water. They took sediment. They took starfish and common mussels off the hull. And then they did something cleverer, which is the part of this study that makes it more than a spot check.

They took healthy blue mussels, the ordinary edible kind, put them in steel cages, and had the divers fix those cages to three different wrecks. Alongside UC-30 there was SMS Mainz, a German light cruiser from the same war, and a Second World War transport ship. The mussels stayed down for around eleven weeks, filtering seawater the way mussels do, which is to say constantly and indiscriminately. A mussel is a very good instrument. It cannot be told where to look, it does not need power, and whatever is in the water ends up in its tissue.

When the cages came up, the mussels from UC-30 had accumulated the highest concentrations of TNT of the three sites.

That is the finding, and the way Andresen has put it publicly is about as plain as science gets: they found traces of TNT in the wildlife, in the seabed and in the water column, and pollution is definitely happening. In the seawater around the wreck the measured TNT was roughly 72.6 nanograms per litre, which is a number worth translating. A nanogram per litre is a part per trillion. This is not a spill. It is a seep, at a concentration you would need very sensitive equipment to notice at all.

But the team also found the compounds TNT turns into once it is out in the world, the ones labelled 2-ADNT and 4-ADNT, which are what you get when bacteria or seawater start breaking the original molecule apart. Finding the breakdown products alongside the parent compound is how you know this is an ongoing process rather than a one-off release. Something is leaking now.

And the mussels were not merely carrying the stuff. The researchers reported significant differences in the activity of the enzymes those animals use to process foreign chemicals, the biological equivalent of a liver working harder than it should. That is a sign of stress, and it is the difference between saying a pollutant is present and saying it is doing something.

There is a quiet irony in using a mussel for this. The animals were not placed there as victims of an experiment so much as recruited as instruments, and the reason they work is the same reason they colonised the wreck in the first place. A submarine hull on a sandy seabed is a rare hard surface in an otherwise featureless place, and marine life takes any hard surface it can get. Within a few decades a wreck like this becomes one of the richest patches of habitat for miles, crusted with mussels and patrolled by starfish and fish that have nowhere else to shelter. The thing that makes it a good home is the same thing that makes it a slow leak.

None of this means the North Sea is poisoned, and the researchers have been careful not to say that. The contamination appears to be local, clustered around the wreck itself rather than spread across the region, and Andresen has said openly that more work is needed to establish how far from a wreck the effect actually reaches. Nobody has suggested that seafood from these waters is unsafe.

What makes the study unsettling is not the concentration. It is the direction of travel.

Steel in seawater corrodes. It has been corroding since the day the boat went down, and every year there is a little less of it between the explosive filling in those mines and the open sea. The mines are inside vertical tubes in the hull, which is both a blessing and a problem: they are contained, for now, by two layers of metal that are both running out. Andresen’s own warning is that this will worsen as the wreck continues to deteriorate and exposes more of the material inside. The clock does not tick down. It ticks up.

And UC-30 is one wreck.

A survey commissioned by Denmark’s environmental authority a couple of years ago counted 10,127 wrecks in Danish waters alone. Around fifteen per cent of them date from the two world wars. Not all of those were carrying explosives, and some are buried deep enough in sediment that they are probably sealed off from the water. But the proportion that are both war-era and armed is not small, and Denmark is one country with one coastline.

Widen the frame and the figure becomes genuinely hard to absorb. Experts estimate that around 1.6 million tonnes of conventional munitions are lying in German waters of the North Sea and the Baltic. A large part of that was not lost in combat at all. It was deliberately dumped after the wars ended, when the fastest way to make a defeated country’s ammunition disappear was to load it onto barges and tip it over the side into shallow water. Nobody at the time was thinking about what the steel casings would be doing eighty years later, because nobody expected anyone to still be thinking about it.

It is worth understanding why so much of it ended up in the sea in the first place, because it was not carelessness so much as a decision that looked reasonable at the time. At the end of both wars there were mountains of shells, bombs and mine casings that somebody had to make safe in a hurry, in countries with wrecked infrastructure and no appetite for storing high explosive near towns. The sea was deep, it was close, and it was assumed to be effectively infinite. Barges went out, cargo went over the side, and the paperwork, where there was any, was often vague about exactly where. That is part of why mapping is now the first job: a great deal of this material is not where the records say it is, if the records exist at all.

So there is now an entire slow-moving problem sitting on the floor of some of the busiest seas in the world, and almost nothing about it is easy to solve.

The obvious answer, blowing the munitions up where they lie, is the one that has historically been used and the one the researchers are most wary of. Detonating old explosives underwater does not make the chemistry go away. Depending on currents and the nature of the seabed, it can take material that was concentrated in one place and distribute it over a much wider area, while also killing anything nearby with the pressure wave. Recovering the munitions instead is safer for the environment and far more dangerous for the people doing it, and it is expensive on a scale that has kept most governments from starting.

Germany has begun to spend real money on recovery, and there is European funding behind work on robotic systems that could map where munitions are and eventually lift them without putting a diver next to a hundred-year-old detonator. That is the direction the field is moving: find everything first, prioritise what is actually leaking, and leave alone the wrecks that sediment has already sealed. Some of these sites are genuinely less of a risk left undisturbed than dug up, and knowing which is which is most of the work.

There is also an argument for a kind of humility here. One of the toxicologists involved in this research has made the point that with a carcinogen, there is no threshold at which you can declare a dose harmless in principle, only doses too small to measure a harm from. That is a scientist’s framing rather than a public health warning, and it should be read as the former. The practical risk to a person eating mussels from the North Sea today is not what this study is about. It is about whether we should be comfortable with a curve that only goes one way.

What lifts this above an environmental audit, though, is what the wreck actually is.

UC-30 was a machine for killing people at a distance, and it failed at that more often than it succeeded, and then it killed its own crew on the way home. Those eighteen mines were built to destroy ships full of men who would never have seen the thing that got them. The submarine sank before it could use them, which at the time would have been counted as a small mercy.

A hundred and nine years later, the mines are doing the only thing left available to them. They are leaking, slowly, into a patch of seabed covered in mussels and starfish, and into the tissue of animals that have made a home out of the hull. The submarine has become a reef. It is also still, chemically, a weapon that has not finished going off.

The team at Aarhus is not calling for the wreck to be raised or the site to be closed. What they are asking for is that these thousands of sites stop being filed as maritime history and start being treated as infrastructure that needs monitoring, because that is what they now are. Andresen’s own summary of the task ahead is modest and a little bleak: they are looking for the gentlest solutions available, because there is unfortunately no easy way to clean any of it up.

That is probably the honest ending. Nobody is going to fix this in one go. The best realistic outcome is that the wrecks that are leaking get found, the worst of them get carefully emptied, and the rest get watched.

The men inside UC-30 have been dead for more than a century, and the war they fought in ended before their grandchildren were born. The cargo they were carrying is still out there in the dark, still doing its job in the only way it has left, one molecule at a time, into the mussels.