Blog

Somewhere in the years after the Second World War, a small number of brown treesnakes came off a ship or a cargo plane on Guam and slid into the forest. Nobody counted them. Nobody saw them arrive. Best estimates say it was a handful of animals, possibly fewer than you could hold in two hands, and by every rule biologists teach undergraduates, that should have been the end of the story.

A founding group that small carries almost no genetic variation. Cousins breed with cousins, the same faulty copies of the same genes keep meeting each other, fertility drops, immune systems get brittle, and the population fizzles out within a few generations. It is one of the most reliable patterns in conservation biology, and it is the reason rescuing a species from very low numbers is so hard.

The brown treesnakes did not fizzle out. They multiplied into one of the densest snake populations ever recorded anywhere on Earth, reaching an estimated 30,000 animals per square mile in parts of Guam at the peak. They climbed into the canopy and ate the island’s birds, nest by nest, until nine of the eleven native forest bird species were gone from the wild. They climbed power lines and bridged the gap between conductors, shorting the grid over and over, at an ongoing cost to the island’s economy of roughly 4.5 million dollars a year in outages and repairs. Eighty years on, they are still there, and there is still no known way to remove them entirely.

So the question that sat unanswered for decades was not how much damage they did. That part was obvious to anyone who walked into a Guam forest and noticed the silence. The question was how a population that inbred managed to do it at all.

A team led out of the University at Buffalo, working with scientists from the United States Geological Survey, has now published an answer, and it is the kind of answer that makes you rethink what you thought you were looking at. The paper appeared in Science Advances last summer, with Christopher Osborne as first author, now at SUNY Oswego, and Trevor Krabbenhoft at Buffalo as the corresponding author. Their finding is that the brown treesnake genome was never as impoverished as everyone believed. The diversity was there the whole time. The tools just could not see it.

Here is the part that is genuinely interesting. For most of the history of genome sequencing, the standard method chopped DNA into very short fragments, read them, and then reassembled the pieces by matching the overlaps, a bit like rebuilding a shredded book from strips of paper. That approach is superb at spotting tiny differences, the kind where one individual has a G and another has an A at the same position. It is close to useless at spotting the large-scale stuff. If a whole chunk of DNA has been duplicated, deleted, flipped around or picked up and moved to a different chromosome entirely, the short-read method reassembles it against a reference and quietly smooths it over. You get a genome that looks tidy and uniform because the messy parts were invisible.

Long-read sequencing reads much longer stretches of DNA in one continuous pass, so those big rearrangements stop hiding. When the team applied it to brown treesnakes, they found more than 19,000 structural variants across the population. These are not typos. They are whole paragraphs moved, repeated or removed. Taken together, they involve something close to eight times more DNA than all the single-letter differences combined.

That is the whole argument in one number. Everyone had been measuring the snake’s genetic diversity using the metric that happened to be measurable, concluded it was low, and moved on. The diversity that mattered was in a form nobody was counting.

What makes it more than a technical footnote is where those variants cluster. They are not scattered evenly. They pile up in two groups of genes in particular: the genes involved in immunity, and the genes involved in smell.

Immunity is the easy one to understand. An inbred population usually has a narrow immune repertoire, which is why small isolated populations get wiped out by a single disease that a diverse population would shrug off. If the brown treesnakes carried far more variation in those genes than anyone realised, they arrived on a new island with a better set of defences than their founder numbers suggested.

Smell is where it gets specific to this animal and to what happened on Guam. The brown treesnake is nocturnal and hunts largely by chemical cues, flicking its tongue and reading the air. It does not need to see a sleeping bird to find one. Guam’s forest birds had evolved over a very long time on an island with no snake that hunted like this, so their nesting behaviour offered no defence against a predator that could simply follow a scent up a tree in the dark. A predator with unusually sharp chemosensory equipment, arriving among prey animals with no evolved answer to it, is close to a worst case.

The researchers raised one further possibility that is almost unsettling in its neatness. In their native range across Australia, New Guinea and the Solomon Islands, brown treesnakes eat each other. Cannibalism is a normal part of how those populations regulate themselves. On Guam, they largely do not, and the team suggests the enhanced sense of smell may be why. Levi Gray, one of the researchers, put the idea in terms of recognition: the snakes on Guam may be able to smell that another snake is close kin, something more like a sibling than a meal. In a population descended from a handful of founders, almost every snake you meet is close kin. A behaviour that normally caps a population may have switched itself off precisely because the population was inbred.

That is a genuinely strange loop. The bottleneck that should have destroyed them may have removed the brake.

It is worth being careful about what the study does and does not establish. The researchers are explicit that they do not yet know whether these structural variants were already present in the founding snakes or arose after the invasion. Answering that requires sequencing native populations in Australia and the South Pacific and comparing them directly, which is the obvious next piece of work and has not been done yet. So the honest version is this: the diversity is real and measurable, its origin is still open, and the link to the invasion’s success is a strong inference rather than a proven mechanism.

Even hedged that way, the implications run in two directions at once, and the second one is the reason this paper matters far beyond Guam.

The first direction is uncomfortable. If invasive species carry reservoirs of adaptive potential that standard genetic surveys miss, then the reassuring assumption that a small invading population will collapse on its own is not safe. Risk assessments that estimate how likely a stowaway population is to establish itself have generally leaned on founder numbers and measured diversity. Those assessments may have been reading the wrong gauge.

The second direction is genuine good news, and Osborne has said as much. The same logic applies to species we are trying to save. Endangered populations get written off as genetically exhausted on the basis of exactly the measurements that turned out to be incomplete here. A tiny surviving population of some bird or mammal, declared too inbred to have a future, may be carrying structural variation that nobody has looked for, because looking for it requires sequencing technology that has only recently become routine. Flexibility in the genome that we could not detect is not the same as flexibility that is not there.

It also lands in a forest that has kept changing long after the birds went quiet. Ecologists comparing Guam with neighbouring islands in the same chain, where native birds still hold on, found that Guam’s jungle carried as many as forty times more spiders. The team behind that work, led by Haldre Rogers and published in an open-access journal well over a decade ago, did it the unglamorous way, walking transects for months with tape measures and counting webs. The explanation is simple once you see it. Birds eat spiders and birds tear through webs as they fly. Remove the birds and the webs stay up.

That is what an ecosystem missing a whole functional group looks like from the inside, and it is a reminder that the snake’s real damage is not the list of species it erased. It is everything downstream of that list, in a forest where nothing eats what the birds used to eat and nothing carries seeds where the birds used to carry them.

Which brings the story back to the birds, and to the only part of it that has anything resembling a happy turn.

Among the species the snakes took was the sihek, the Guam kingfisher, a small bird with a cinnamon head and a startlingly loud voice for its size. By the 1980s it was gone from the wild. What saved it was an unglamorous decision made just in time: a small number of birds were taken into captivity, and for nearly four decades the entire species existed only in breeding facilities, passed between zoos, its studbook managed like a bank account that could never be overdrawn. Every sihek alive descends from that handful. It is, in its own way, the same bottleneck problem as the snake, viewed from the opposite side.

Returning them to Guam was not possible, because the snakes are still there and a released kingfisher would simply be food. So conservationists did something sideways instead. A couple of years ago, nine captive-bred sihek were flown to Palmyra Atoll, a remote and snake-free scrap of land in the central Pacific, more than 3,000 miles from Guam. Six were released first, each carrying a small radio tracker, and the rest followed. It was the first time in roughly four decades that the species existed outside a cage.

They survived. All nine made it through their first year in the wild. Four pairs staked out territories and built nests, and the following spring three pairs laid eggs, which was the first wild sihek breeding since the species vanished from Guam. Those eggs did not hatch, which is not unusual for young birds on their first attempt, and the teams monitoring them treated it as a step rather than a failure. Palmyra was never meant to be the destination. It is a practice ground and a data source, a place to learn how these birds behave when nobody is feeding them, so that the knowledge exists for the day they can go home.

That day depends entirely on the snakes, which is why the genome paper is not an abstract curiosity to the people working on Guam.

Control efforts there have become inventive out of necessity. The most effective landscape-scale tool so far is startling when you first hear about it: helicopters fly over forest that is impossible to reach on foot and fire out thousands of small biodegradable cartridges, each containing a dead mouse with an acetaminophen tablet attached. The ordinary painkiller in most bathroom cabinets is lethal to brown treesnakes at a dose roughly equivalent to what you would give a child, while being comparatively low risk to Guam’s other wildlife. The cartridges are designed to snag in the canopy where the snakes hunt rather than drop to the ground where they would not. Trials run by federal wildlife scientists showed sharp falls in snake activity that persisted for about a year, with the notable caveat that the largest and best-fed snakes were the ones most likely to survive, which is exactly the kind of detail that suggests the dose needs work.

Alongside that sit detector dogs working cargo before it leaves the island, traps, hand capture, environmental DNA sampling that can pick up a snake’s presence from traces left in the environment, and citizen volunteers removing animals in park areas. It is a long, grinding, unfinished effort whose most important measure of success is not on Guam at all. It is Hawaii, where a single established brown treesnake population has been estimated to cost hundreds of millions of dollars a year in electrical damage alone, never mind what it would do to birds that have exactly the same evolutionary blind spot Guam’s did.

So the genome study lands in the middle of practical work, not on a shelf. Knowing that the snakes have unexpected depth in their immune genes matters if anyone ever seriously pursues a biological control agent, because a pathogen deployed against a population with a broad immune repertoire will not do what the models say. Knowing that the olfactory genes are where the variation piles up matters for trap design and lure chemistry, because the sense being exploited turns out to be the snake’s most variable asset rather than a fixed target.

The thing that stays with you, though, is the correction itself. For decades the brown treesnake sat in textbooks as a puzzle, an animal that succeeded in defiance of population genetics. It turns out it was not defying anything. The rules were fine. We were reading the animal with an instrument that could not see the thing that mattered, and the invisible thing was doing all the work.

There is a version of that sentence that applies to the kingfishers too, sitting in their nests on an atoll in the middle of the ocean, descended from a founding group as small as the snakes’ and written off for years as a species with nothing left in the tank. Nobody has sequenced them the way the snakes were sequenced. Somebody probably should.

The bird was a brown skua, and it was dead on a beach at the eastern end of a national park in southwestern Australia, a few hundred kilometres along the coast from the nearest city of any size. Someone found it, someone sent samples away, and a laboratory confirmed what a lot of people had spent four years quietly dreading. The official notification went out the following day, and it is a short, dry document that says one very large thing: highly pathogenic avian influenza H5N1, clade 2.3.4.4b, in a wild bird in Australia, for the first time.

That is the confirmed fact, and it is worth keeping it separate from everything that has been said around it, because plenty has.

For roughly four years the H5N1 panzootic had been working its way through the world’s wild birds with a thoroughness nothing quite like it had managed before. It moved through Europe, through Africa, through North and South America. It reached the Antarctic Peninsula. It killed birds in numbers that stopped being countable and started being estimated. Hundreds of millions of poultry have been culled or died worldwide since this thing started moving. More than 600 wild bird species and around 100 mammal species have been affected. Seabird colonies in the North Atlantic lost large shares of their breeding birds in a single season. In South America it ran down the Pacific coast and killed pelicans, other seabirds and sea lions in very large numbers, and then it reached the Antarctic, which is where the southern route into Australia began. And through all of it Australia held, the one inhabited continent where this particular virus had never been detected in a wild bird. Australian biosecurity officials were asked about it constantly, and the honest answer was always the same. Not yet.

There is an important footnote to the “last continent” framing that got lost in a lot of the coverage, and it deserves to be said plainly, because it changes how you read the mainland detection. Australia’s remote sub-Antarctic territory had already been hit. On Heard Island, a glaciated speck in the Southern Ocean thousands of kilometres from anywhere, H5N1 arrived around the middle of the previous year, most likely carried in by wildlife moving from the French-administered Crozet Islands roughly 1,800 kilometres away. Australian Antarctic Program scientists surveyed the damage, and the numbers are hard to hold in your head. Among southern elephant seal pups, average mortality across the island ran at about 76 per cent, and in some areas it reached 97 per cent. Out of a total pup production of around 17,000, roughly 13,000 died. Six vertebrate species tested positive, including king penguins, gentoo penguins and Antarctic fur seals. Working out the scale of it took drones, because Heard Island’s coastline is not something a small team can walk, and counting bodies from the air was the only way to turn an obvious catastrophe into a number anyone could publish. Surveys the following summer found several hundred dead adult king penguins, though that was a small proportion of a very large colony.

So the virus was not knocking on Australia’s door. It was already inside the outermost room. What fell in June was the mainland.

The skua matters as much as the location does. Brown skuas are big, powerful, scavenging seabirds that make their living partly by eating things that are already dead, and they move between the sub-Antarctic islands and the southern coasts of Australia and New Zealand. Epidemiologists had drawn that exact route on a map years earlier as the most likely way in. Scavengers eat infected carcasses, scavengers travel long distances, scavengers do not respect quarantine zones. The bird that brought it was the bird everyone said would bring it, which is a strange thing to be right about.

The detection site was Cape Le Grand National Park, near the coastal town of Esperance in Western Australia, about 700 kilometres southeast of Perth. A southern giant petrel at the same site was also suspected. Samples went to the national high-containment animal laboratory for confirmation, and the international notification followed immediately, which the World Organisation for Animal Health went out of its way to note approvingly. Australia has a long record of reporting fast even when reporting fast is inconvenient, and it reported this within a day.

Then came the part nobody can control.

Here is what has been confirmed since, because this is where it is easy to slide from fact into fear. Within about ten weeks of that first skua, Australian authorities had logged more than 350 confirmed detection events. The virus has turned up in six states and territories. More than twenty species of bird have tested positive, and the list reads like a survey of the southern Australian coastline: greater crested terns, silver gulls, Pacific gulls, black-faced cormorants, little penguins, Australasian gannets, several petrel and shearwater species, white-fronted terns. It also reached birds nobody associates with the sea at all, including Australian magpies, ravens, a peregrine falcon and a southern boobook owl.

It has crossed into mammals too, which is the part that gets virologists’ attention. Confirmed mammal cases include long-nosed fur seals and an Australian sea lion on Kangaroo Island, a common dolphin in South Australian waters, and a red fox in Adelaide. Every one of those is consistent with what H5N1 has done elsewhere in the world over the past four years, and none of them is evidence of the virus adapting to spread between mammals.

The first mass die-off came in early August, off the coast south of Adelaide. Aerial surveillance spotted dead greater crested terns at Cape Jaffa, and the count came in at 49 dead birds with another 35 visibly sick. That was small enough to describe precisely, which is exactly why it is worth recording, because two weeks later it stopped being that kind of number. Three separate mass mortality events were confirmed across South Australian islands, on Troubridge Island off the Yorke Peninsula and on the North and South Page islands near Kangaroo Island, and the combined toll passed a thousand greater crested terns. On King Island, off Tasmania, 28 little penguins and three terns were found dead in the same week.

The state premier in South Australia called it a significant escalation and said the loss of more than a thousand birds was deeply concerning. The federal agriculture minister was blunter about what comes next, telling people to expect more spread and larger numbers of wildlife affected. The environment minister put it in the flattest possible terms: you cannot stop wild birds flying and you cannot stop wild animals moving, and that means more events like this one.

Two things have not happened, and both belong in any honest account.

Australia’s commercial poultry industry has not recorded a single H5N1 infection. Farms have been locking down since the June detection, and so far the virus has stayed in wildlife. This is a genuinely important distinction and it is one people keep getting wrong, because Australia has had high pathogenicity bird flu in poultry before. Those were H7 viruses, a different subtype entirely, and they behaved differently. The most recent H7 outbreaks were contained and eradicated, and Australia was formally declared free of high pathogenicity avian influenza in poultry the year before the skua turned up. Blending the two together produces a story that is not true. The H7 outbreaks were a poultry emergency that was closed out. This is a wildlife event, so far.

And no human case has been recorded in Australia. The Australian Centre for Disease Control’s assessment is that most bird flu viruses do not spread easily from animals to people and that the risk to Australians is very low. That is not reassurance for its own sake, it is what the global numbers look like. Across roughly ten months spanning the arrival in Australia, twelve human H5N1 infections were reported anywhere in the world, in Bangladesh, Cambodia and India, and three of those people died. No person-to-person transmission was identified in any of them. Every case involved close contact with infected animals or contaminated environments. The public health advice in Australia is correspondingly simple and mostly about wildlife rather than people: do not touch sick or dead birds, keep your distance, keep pets away, and report what you find to the emergency animal disease line.

What is genuinely at stake here is not human health. It is a set of animals that exist in very small numbers in very few places.

That is why the most remarkable part of this story is not the virus at all. It is what Australia did with four years of warning.

On Phillip Island, off the Victorian coast, there is a little penguin colony famous enough that tourists come at dusk to watch the birds walk up the beach. Wildlife staff there have spent months catching those penguins one at a time, giving each one an injection, and then catching the same bird again three or four weeks later for a booster. The target is up to 5,000 penguins on Phillip Island and up to 800 more at the St Kilda colony in Melbourne. It is one of the largest wildlife vaccination programmes of its kind ever attempted anywhere, and it was possible only because the same organisation had already run the trials, years earlier, that showed little penguins mount a decent immune response to the vaccine without long-term ill effects. Somebody did that unglamorous preparatory work when the virus was still on another continent.

It is worth understanding how unusual this is. Vaccinating wild animals against a disease is something conservation biology almost never does, and for good reasons. You have to catch every individual, which for most species is impossible. You have to catch most of them twice. Handling stresses the animal and can itself cause deaths. The vaccine has to be tested on that particular species, because what is safe in a chicken is not automatically safe in a seabird. And you have to be confident that the population is small enough and concentrated enough for the effort to actually change the outcome, rather than being a gesture that makes everyone feel better while the virus goes around it. Penguin colonies that come ashore in the same place every night, and a critically endangered parrot that spends part of its life in an aviary, happen to be two of the rare cases where the arithmetic works.

At Cleland Wildlife Park in South Australia, around a hundred birds were vaccinated. And in Tasmania, where H5 spread across the state and had produced 54 confirmed cases by the start of September, keepers began vaccinating the orange-bellied parrot.

That bird deserves a paragraph of its own. The orange-bellied parrot is one of the rarest parrots on Earth. It breeds in one remote corner of southwest Tasmania and migrates across open water to the mainland, and the entire wild population is estimated at fewer than a hundred individuals. There is a captive breeding facility near Hobart holding several hundred more, and those captive birds are the reason the species still exists at all. Roughly 350 of them are being vaccinated, two shots each, over a programme running into the following year, before the birds are released at Melaleuca in the far southwest. The Tasmanian environment minister did not soften it: the virus threatens the species’ existence.

Beyond the parrot, conservation bodies have flagged which Australian species are most exposed, and the logic is uncomfortable but straightforward. The birds at greatest risk are the ones that nest in enormous dense colonies in a handful of locations, because that is a perfect setting for a respiratory virus. Christmas Island frigatebirds. Abbott’s booby. Gould’s petrel. Australian pelicans. Black swans. Species where a single bad season at a single breeding site could take out a meaningful share of the world population.

The other projected worry, and it is a projection rather than an observation, is ducks. If H5N1 establishes itself in Australia’s waterfowl, the virus stops being a coastal problem and starts moving inland along wetlands and river systems into the interior, where almost nobody is watching and almost nothing can be surveyed. That has not been confirmed. It is the scenario that keeps the modelling teams busy.

Ask the people actually doing this work how it is going and the answers are exhausted rather than dramatic. One veterinary pathologist working through the South Australian cases described the situation as pretty overwhelming. Wildlife vets have been telling rescuers and volunteers that biosecurity, the boring business of gloves and disinfectant and not carrying the virus home on your boots, is by far the most important thing anyone can do.

So did the drills work? Partly, and the honest answer is the useful one. Nothing was ever going to stop a scavenging seabird flying across the Southern Ocean and dying on a beach. Australia did not prevent the arrival, and no country has. What four years of preparation bought was a laboratory that could confirm the virus in days, a notification that went out within twenty four hours, a surveillance system that caught the first mass mortality from a helicopter, a public reporting line people actually use, and a vaccine programme that was ready to run on the day it was needed instead of being designed from scratch in a crisis.

It also bought something less tangible. Nobody had to argue about whether this was real.

Out on the islands the terns are still dying, and the counts are still coming in, and the season ahead will decide how much of this becomes permanent. But the picture worth holding onto is a smaller one. It is a wildlife officer on a windy beach in the dark, holding a little penguin that weighs about as much as a bag of sugar, while a colleague finds the right spot with a needle. Then letting it go. Then picking up the next one. Four thousand nine hundred and ninety nine to go.

They are not going to save every bird. They knew that before they started. They are doing it anyway.

Here is a number that should not be possible. Roughly a quarter of every insect observation held in the world’s biggest biodiversity database comes from the United Kingdom. Not from Brazil, which contains a fifth of the Amazon. Not from Indonesia, which has more islands than most people can name. From one cool, damp, thoroughly surveyed island group off the northwest coast of Europe, a place with fewer than sixty species of butterfly, in a world that has tens of thousands.

Meanwhile more than eighty per cent of all insect species on Earth live in the tropics.

That gap is the subject of a paper published in the journal BioScience earlier this month by a team of researchers spread across eight countries, led by Dr Shawan Chowdhury at Monash University in Australia, with Krushnamegh Kunte of the National Centre for Biological Sciences in India among the senior authors. Their argument is not that anyone did anything wrong in Britain. British naturalists have been counting moths and beetles in their gardens for the better part of two centuries, and the resulting record is one of the great unglamorous achievements of amateur science. The argument is about what happens when that record becomes the default picture of insect life for the entire planet.

To understand why that number is so lopsided, it helps to know how the British record came to exist. It was not built by a government programme. It was built by clergymen with butterfly nets, by factory workers who ran moth traps in the back garden, by schoolteachers and railway clerks and retired accountants who kept notebooks for fifty years and then left them to a museum. Britain industrialised early, it had a leisured middle class with a hobby, and natural history became something ordinary people did on Sundays. Two centuries of that habit, digitised, is what a quarter of the world’s insect record looks like.

It is easy to check the claim yourself, and I did. The Global Biodiversity Information Facility, usually shortened to GBIF, is the open database where museums, universities, government agencies and citizen science apps pool their records. It is the closest thing the planet has to a single ledger of what lives where, and it is what sits underneath an enormous amount of applied conservation work: extinction risk assessments, protected area design, environmental impact reports, the range maps in the back of field guides. Ask it how many insect occurrence records it holds worldwide and the answer comes back at around 340 million. Ask how many of those come from the United Kingdom and the figure is roughly 80 million. That is about twenty four per cent by my count, and the paper puts it at nearly twenty six per cent, a difference that comes down to which snapshot of a constantly growing database you are looking at. Either way, the shape of the thing is the same. One country supplies something close to a quarter of what science formally knows about where insects are.

Now run the same query for Brazil. Around three million records. Indonesia, one of the most insect-rich countries that has ever existed, comes back with well under a million. The entire nation of Indonesia has roughly one per cent of the British total.

This is not because there are fewer insects there. It is because there are fewer people being paid to write them down, fewer institutions to store the specimens, and fewer journals willing to publish the results in a language the database can read.

That last point turns out to matter enormously, and it is the first of four barriers the researchers identify. A great deal of tropical insect research does exist. It is simply invisible. It gets published in national journals, in regional languages, in university bulletins that never get indexed, in theses that sit on a shelf in a departmental library and nowhere else. To an international database, work that is not in English and not in a major indexed journal might as well not have happened. The knowledge is real. The pipeline that would carry it into global science is missing.

The second barrier is the raw data shortfall, and the team breaks it into parts that each have their own fix. There is almost no systematic long-term insect monitoring in tropical countries, the kind of repeated annual counting that lets you say whether a population is rising or falling rather than just that it exists. Citizen science, which now drives an enormous share of temperate insect records, has barely taken hold in most tropical regions. There are nowhere near enough trained taxonomists, the specialists who can look at a small brown beetle and say which of several thousand small brown beetles it is. This is a genuine global bottleneck and it is getting worse rather than better. Describing a new species properly is slow, meticulous, poorly funded work that produces very few citations, which in a modern university career structure makes it close to professional suicide. The people who can do it are ageing out, their positions are being converted into something more fundable, and the queue of undescribed specimens in the world’s collections keeps growing. There are drawers in museums right now holding insects nobody has ever named, collected decades ago, waiting for a person who no longer exists in sufficient numbers. And the specimens that have been collected are often physically inaccessible, sitting in drawers in institutions with no scanner, no database and no funding to make a catalogue.

The third barrier is infrastructure, which is the polite word for money. Tropical countries face financial constraints that have nothing to do with scientific ambition. Museum collections are scarce and frequently under-resourced. Computing capacity for the kind of modelling that modern conservation planning requires is limited. Some of the richest insect habitats on Earth sit in places where fieldwork carries genuine security risks.

And the fourth barrier is the one that makes the other three sting. Even where the knowledge exists, it very often does not turn into conservation action. Insects are almost absent from national conservation planning in most of the world. They rarely appear on protected species lists. They are rarely factored into how a reserve boundary is drawn. A tiny fraction of insect species have ever been assessed for extinction risk at all, and the ones that have tend to be the charismatic ones, the big showy butterflies and the dragonflies people like looking at. Everything else is classed as data deficient, which is a technical term that means nobody knows, and which in practice functions as permission to do nothing.

Somewhere inside all of this sits a practice the paper names directly, and the name has stuck because it is so exact. Parachute science. A research team from a wealthy country flies into a tropical nation, collects specimens or data, flies home, publishes in an international journal, and leaves behind essentially nothing. No local co-author with a leading role. No copy of the dataset in a national repository. No training, no equipment, no next step. Local researchers who did the actual finding, who knew which valley to walk up and which tree the thing sits on, are treated as field assistants rather than colleagues. The country supplies the biodiversity and receives a citation, if that.

The corrosive part is not just the unfairness. It is that parachute science produces a body of knowledge that cannot be maintained. A one-off survey tells you what was there on one afternoon. Conservation needs to know what is happening over decades, and only people who live there can produce that.

To show what the alternative looks like, the team worked through a single country in detail, using butterflies in Bangladesh as a test case. Butterflies are a reasonable bellwether because they are large, colourful, popular with amateurs and relatively well studied compared with, say, parasitic wasps.

The first thing they did was go looking for the invisible literature, and they found it. There were 111 published studies on the butterflies of Bangladesh. Only 34 of them had appeared in international journals. More than two thirds of the country’s own butterfly science had never reached the databases that global conservation planning draws on. And of those 111 studies, only three were focused on conservation itself. The research existed. It described, it listed, it recorded. It almost never reached the point of saying what should be protected and how.

Then they did something that sounds slightly absurd until you see the result. They went on Facebook.

There are active amateur naturalist groups in Bangladesh where people post photographs of butterflies they have seen, usually with a location and a date attached, often with an identification argued out in the comments by people who genuinely know what they are looking at. The researchers treated those posts as data. When they compiled the results, they found records for 167 of the country’s 169 threatened butterfly species.

One hundred and sixty seven out of one hundred and sixty nine. Sitting in plain sight, on a social media platform, uploaded for free by people who were not being paid and in most cases were not being counted as contributors to science at all.

That is the whole argument of the paper in one finding. The knowledge gap is not always a knowledge gap. Sometimes it is a plumbing problem. The observations have already been made. They are just not in the pipe.

With those records folded in, the team could do the kind of analysis that had previously been impossible for the country. They could map where species actually occur rather than where researchers happen to have visited. They could check how well the existing protected areas cover the ranges of threatened species, and the answer was not encouraging. They could model how habitat suitability is likely to shift as the climate changes, and the projection there is bleak: something in the order of forty two per cent of the country’s butterfly species face a meaningful risk of disappearing locally.

None of that could have been said a few years ago. Not because the butterflies were hiding, but because nobody had connected the amateur photographs to the museum drawers to the untranslated papers.

It is worth stopping for a second on why any of this matters beyond the satisfaction of a well-filled database, because insects are the kind of subject people nod along to without quite feeling. Insects pollinate around eighty per cent of wild plant species. They are the direct food supply for roughly sixty per cent of bird species. They break down dead material, turn over soil, control the populations of other insects that would otherwise eat crops, and sit at the bottom of food chains that end in things people do care about. Estimates of how many insect species exist run somewhere between fourteen and thirty million, and the overwhelming majority have never been formally described at all. When conservationists talk about biodiversity loss, insects are most of the biodiversity.

And every serious decision about them, every extinction risk assessment, every protected area boundary, every climate range projection, is built from data. If a quarter of that data comes from one temperate island and a tropical megadiverse nation contributes one per cent, then the models are being trained on a sample that does not resemble the world.

The fix the researchers propose is deliberately unglamorous, which is probably why it might work. Go and find the regional literature that already exists, including in languages other than English, and connect it to expert networks in the countries concerned. Expand citizen science in the tropics, and expand it past the photogenic species, because somebody has to count the flies too. Build and fund national biodiversity collections so that specimens stay in the country that produced them and can actually be consulted. Digitise the drawers that already exist. Treat social media observations, local naturalist knowledge and Indigenous knowledge as real evidence rather than colour.

And, running through all of it, change who holds the pen. The framework is explicit that international partnerships have to be structured so local researchers lead, that national regulations and benefit-sharing rules are followed rather than worked around, and that the country supplying the biodiversity retains sovereignty over the data about it.

Chowdhury’s framing of the problem is blunt, and it has less to do with discovery than people expect. The task is not primarily to go out and generate vast quantities of new information. It is to get much better at finding, connecting and using the knowledge that has already been produced by people whose names are not on the papers.

There is something quietly hopeful buried in that. A knowledge gap caused by there being nothing to know would be a hard problem, and an expensive one, and slow. A knowledge gap caused by the observations existing in the wrong format, in the wrong language, on the wrong platform, credited to the wrong people, is a solvable problem. Somebody in Bangladesh already photographed 167 of those 169 threatened butterflies. Somebody in a university in Dhaka already wrote the paper that nobody outside the country read.

The insects were never missing. The people counting them were just never invited in.

There is one guest nobody invites to a summer picnic and everybody expects anyway. The wasp arrives around the time the food does, hangs over the jam, dive bombs the glass of cider, and gets swatted at by six people at once who all miss.

Across Britain this year, it mostly did not come.

That is the observation, and it came from the one person in the country best placed to notice, because finding wasps is her job. Seirian Sumner is a professor of behavioural ecology at University College London, and she has spent her career on social wasps at a time when hardly anybody else wanted to. When she wrote up what she was seeing, the detail that stuck was not grand. It was that her students could not get hold of enough wasps to run their experiments.

And the ones that did turn up were small. Not slightly small. Spring small, in the middle of summer, which for anyone who knows how a wasp colony is supposed to work is the strange part.

Here is the thing she was careful to say, and it deserves to be said just as plainly here, because a lot of the coverage that followed lost it. These are anecdotes, not data. There is no peer reviewed finding. There is no confirmed population count for the year. What there is, at the moment, is a specialist noticing an absence that does not match anything she has recorded before, and saying so out loud while the counting is still going on.

Which makes this a story about the gap between noticing something and knowing it, and about a national experiment that happens to be positioned exactly in that gap.

The experiment is called the Big Wasp Survey. It started in 2017, and the method is about as unglamorous as science gets. Ordinary people put a small amount of beer in a trap, hang it in the garden for a set number of days during a fixed window in late summer, then send in what they caught so it can be identified to species. That is it. Beer, a bottle, a week, a post.

The first year, more than 2,300 people signed up and about 1,294 of them actually returned results, which produced 6,680 wasps identified to species. The breakdown was tidier than you might expect. Common yellowjacket and German yellowjacket split the catch almost evenly at around 44 per cent each, European hornets made up about 6 per cent, and a handful of other species accounted for the rest. Later, DNA work on those samples produced a genuinely useful result, which was that the common yellowjacket across the country behaves as one large interconnected population rather than a patchwork of separate local ones.

None of that sounds dramatic. It is, however, the only national wasp monitoring this country has. And when Sumner wrote up her observation in high summer, the survey had not yet started sampling for the year. The traps go out in a window in August and another at the turn of the month. So the person raising the alarm was doing it in the one part of the calendar where she could not yet check.

That is not a flaw in the story. That is the story.

The beer trap is doing something clever with all of that. It is cheap enough that thousands of people can run one, it is standardised enough that the results can be compared between a garden in the far north and a garden in the south, and it is timed to the exact stretch of the year when workers are out and easy to catch. It also produces physical specimens rather than sightings, which means species can be confirmed under a lens and, as the DNA work showed, interrogated years later for questions nobody was asking at the time. That is a lot of scientific return for a bottle of cheap lager.

To understand why small wasps in August are a signal rather than a curiosity, you have to know what a wasp colony’s year looks like, and it is more precarious than the picnic version suggests. A social wasp colony does not overwinter. Every single nest you have ever ducked away from was built from nothing that year by one female, working alone.

A queen comes out of hibernation in spring, finds a cavity, chews wood into paper, builds a few cells, lays into them, and then feeds and raises that first tiny batch of workers entirely by herself. There is no colony to help her. If she dies in that window, or if the weather keeps her from foraging long enough, that is the end of it. Nothing is left behind. The whole enterprise rests on one insect having a good few weeks in spring.

Once those first workers emerge, the arithmetic flips. They take over foraging, the queen switches to laying full time, and the colony scales up fast through the summer until it is the loud, confident, jam stealing operation everyone recognises. Workers that emerge early are naturally small, because there were few foragers to feed the larvae. By late summer they should be noticeably bigger, because by then there are hundreds of adults bringing food in.

So small workers in August mean one of two things, and possibly both. Either the colonies never built up enough foraging capacity to feed their larvae properly, or the larvae developed too fast in unusual heat and emerged underfed and undersized. Neither of those is the signature of colonies that simply started late.

There is a reason beer works, and it is the same reason wasps become a nuisance in the first place. For most of the summer a worker wasp is not interested in your food at all. She is hunting protein, caterpillars and flies and aphids, to carry back to the larvae in the nest. What she gets in return, from the larvae themselves, is a drop of sugary liquid. That exchange is the engine of the whole colony, and it is why a wasp ignores you in June.

Late in the season the queen stops laying, the larvae run out, and the sugar supply that kept the workers going simply ends. Thousands of adults are suddenly unemployed, hungry for sugar and with nothing to do. That is the wasp at your table. It is not aggression. It is an entire workforce made redundant at once, going looking for the only thing it still wants. They find fallen fruit, they find your drink, and they get clumsy and bold in a way that reads as malice and is closer to desperation.

Meanwhile a small number of new queens have been produced, mated, and left to find somewhere to spend the winter under bark or in a shed. Everyone else dies with the first hard cold. The nest is never reused. Next spring the whole thing starts again from those few hibernating females, which is why a bad spring does not dent a colony so much as delete a generation of them.

Which brings in the weather, and here the numbers need handling with care. Sumner points at a spring of extremes, warm dry stretches broken by cold damp spells and chilly nights, hitting precisely during the period when lone queens are most exposed. Then came the heat. Social wasps are reported to have a critical thermal maximum somewhere around 45 degrees Celsius, and during the summer’s heatwaves ground surfaces in the country reportedly exceeded 50. For scale, honeybees are generally put at around 49. Wasp nests, especially the ones started in shallow ground cavities and banks, sit right where that heat lands.

That figure of 50 degrees at the ground surface is the one to hedge, and it is worth saying why. It is a surface temperature, not an air temperature, and it comes through a comment piece rather than a published dataset. It is entirely plausible and it is not a measurement anyone has yet tied directly to a count of dead nests.

There is also a serious counterargument, and it comes from the same expertise. Social wasps are supposed to be tough. Nests are insulated by their own paper construction. Workers cool them behaviourally, fanning and carrying water. The diet is broad rather than specialist, which usually protects an animal when one food source fails. On paper, a wasp should ride out a bad summer better than a fussier insect would. That is exactly why a summer this quiet is interesting rather than expected.

And there is a longer pattern underneath it. Wasp numbers swing wildly year to year, far more than most insects. Last year was the opposite of this one, a boom year following a warm dry spring, with the predictable wave of newspaper stories about plagues of wasps. The absence of those same stories this year was, oddly, one of the first clues anybody noticed. An analysis published years ago that pulled together nearly four decades of records from a handful of English sites found the same thing in the long data: enormous swings, boom and bust, driven largely by what the weather does in a narrow spring window.

So the honest question is not whether wasps have crashed. It is whether the swings are getting wilder, and whether the window a queen needs is being squeezed from both ends, by heat that arrives too early and cold that returns too late.

There is one more pressure being raised, and it comes from people who deal with wasps commercially rather than academically. Pest controllers say they have had far fewer callouts for nest removal this year, which is its own kind of evidence, collected by accident. One of them has also flagged a different problem entirely: spring trapping campaigns aimed at invasive Asian hornets are not selective, and native wasp queens reportedly end up in those traps too, during the exact weeks when the loss of a single queen costs an entire colony. That claim comes from a pest control operator rather than a study, and it is worth knowing about rather than treating as settled.

Now the part that decides whether any of this matters to anyone who does not love wasps, which is nearly everyone.

Sumner and colleagues went through more than 500 academic papers a few years ago, covering roughly 33,000 species of stinging wasps, and put together the case that wasps do the same categories of work bees do while receiving almost none of the credit or the funding. Wasps are predators, and they are efficient ones. They take aphids and caterpillars in volume, which is to say they eat the things that eat crops. Insect predation as a global pest control service has been valued at somewhere around 416 billion dollars a year, and wasps are barely counted in that figure at all, mostly because nobody has done the work to count them.

They pollinate too, which surprises people. The review found evidence of wasps visiting 960 plant species, and 164 of those are completely dependent on wasps for pollination, meaning that without wasps those plants simply do not reproduce. Global insect pollination is valued at well over 250 billion dollars a year. Wasp venom and saliva have antibiotic properties, and yellowjacket venom has been studied for possible use against cancer cells. In parts of the tropics, wasp larvae are already food.

That is a substantial resume for an insect whose entire public identity is ruining a barbecue.

The funding gap is the reason this summer is so hard to resolve. Bees get monitored, mapped, modelled and mourned. Wasps get a citizen science project running on beer and goodwill, because the people who fund insect research fund the insects people like. The consequence is that when something unusual happens to wasps, the country’s ability to say what actually happened comes down to how many strangers hung a bottle in a hedge in August.

Two caveats before the end. The first is geographic. Everything here is Britain. There is no comparable set of reports from elsewhere, and it would be wrong to read this as a global event. The second is the one Sumner put first, and it still stands. Nothing here is a confirmed population decline. It is an absence, described precisely by somebody qualified to describe it, waiting on a count.

That count is happening now. The traps went out in the usual late summer windows, the returns come in afterwards, and the samples get sorted and identified species by species, the same way they have been for close to a decade. For most of those years the survey has been building a baseline that nobody had an urgent reason to look at. This year it has a question waiting for it, which is the situation every long running monitoring project secretly hopes for and never quite expects.

And if the traps come back light, the finding will not be that a nuisance went away. It will be that a predator, a pollinator and a decomposer went quiet at the same time, and that the only reason anyone can put a number on it is that ordinary people spent nine summers counting an insect they do not like.

The wasps may well be back next year. They usually are, and a boom summer would surprise nobody. But this is the first time the absence of something everybody swats at has been measurable, and that is worth more than the picnic it did not ruin.