Across a Swiss meadow and into its forested edges, the drone dragged a jumbo-size cotton swab from a 13-foot tether. Along its path, the moistened swab collected scraps of life: some combination of sloughed skin and hair; mucus, saliva, and blood splatters; pollen flecks and fungal spores.
Later, biologists used a sequencer about the size of a phone to stream the landscape’s DNA into code, revealing dozens upon dozens of species, some endangered, some invasive. The researchers never saw the wasps, stink bugs, or hawk moths whose genetic signatures they collected. But all of those, and many more, were out there.
The researchers, from the Swiss Federal Institute for Forest, Snow and Landscape Research, were field-testing a new approach to biodiversity monitoring, in this case to map insect life across different kinds of vegetation. They make up one of many teams now deploying a suite of technologies to track nature at a resolution and pace once unimaginable for taxonomists. “We know a lot more about what’s happening,” Camille Albouy, an environmental scientist at ETH Zurich, and member of the team, told me, “even if a lot still escapes us.”
Today, autonomous robots collect DNA while state-of-the-art sequencers process genetic samples quickly and cheaply, and machine-learning algorithms detect life by sound or shape. These technologies are revolutionizing humanity’s ability to catalog Earth’s species, which are estimated to number 8 million—though perhaps far, far more—by illuminating the teeming life that so often eludes human observation. Only about 2.3 million species have been formally described. The rest are nameless and unstudied—part of what biologists call dark taxa.
Insects, for example, likely compose more than half of all animal species, yet most (an estimated four out of five) have never been recorded by science. From the tropics to the poles, on land and in water, they pollinate, prey, scavenge, burrow, and parasitize—an unobserved majority of life on Earth. “It is difficult to relate to nonspecialists how vast our ignorance truly is,” an international consortium of insect scientists lamented in 2018. Valerio Caruso, an entomologist at the University of Padua, in Italy, studies scuttle flies, a skittering family containing an estimated 30,000 to 50,000 species. Only about 4,000 have been described, Caruso told me. “One lifetime is not enough to understand them all.”
The minute distinctions within even one family of flies matter more than they might seem to: Species that look identical can occupy entirely different ecological niches—evading different predators and hunting different prey, parasitizing different hosts, pollinating different plants, decomposing different materials, or carrying different diseases. Each is a unique evolutionary experiment that might give rise to compounds that unlock new medicines, behaviors that offer agricultural solutions, and other adaptations that could further our understanding of how life persists.
Only with today’s machines and technology do scientists stand a chance of keeping up with life’s abundance. For most of history, humans have relied primarily on their eyes to classify the natural world: Observations of shape, size, and color helped Carl Linnaeus catalog about 12,000 species in the 18th century—a monumental undertaking, but a laughable fraction of reality. Accounting for each creature demanded the meticulous labor of dehydrating, dissecting, mounting, pinning, labeling—essentially the main techniques available until the turn of the 21st century, when genetic sequencing allowed taxonomists to zoom in on DNA bar codes. Even then, those might not have identified specimens beyond genus or family.
Now technologies such as eDNA, high-throughput sequencing, autonomous robotics, and AI have broadened our vision of the natural world. They decode the genomes of fungi, bacteria, and yeasts that are difficult or impossible to culture in a lab. Specialized AI isolates species’ calls from noisy recordings, translating air vibrations into an acoustic field guide. Others parse photo pixels to tease out variations in wing veins or bristles as fine as a dust mote to identify and classify closely related species. High-resolution 3-D scans allow researchers to visualize minuscule anatomies without lifting a scalpel. Other tools can map dynamic ecosystems as they transform in real time, tracking how wetlands contract and expand season by season or harnessing hundreds of millions of observations from citizen-science databases to identify species and map their shifting ranges.
One unassuming setup in a lush Panamanian rainforest involved a UV light luring moths to a white panel and a solar-powered camera that snapped a photo every 10 seconds, from dusk to dawn. In a single week, AI processed many thousands of images each night, in which experts detected 2,000 moth species—half of them unknown to science. “It breaks my heart to see people think science is about wrapping up the last details of understanding, and that all the big discoveries are done,” David Rolnick, a computer scientist at McGill University and Mila – Quebec AI Institute, who was part of the expedition, told me. In Colombia, one of the world’s most biodiverse countries, the combination of drone-collected data and machine learning has helped describe tens of thousands of species, 200 of which are new to science.
These tools’ field of view is still finite. AI algorithms see only as far as their training data, and taxonomical data overrepresent the global North and charismatic organisms. In a major open-access biodiversity database, for example, less than 5 percent of the entries in recent years pertained to insects, while more than 80 percent related to birds (which account for less than 1 percent of named species). Because many dark taxa are absent from training data sets, even the most advanced image-recognition models work best as triage—rapidly sorting through familiar taxa and flagging likely new discoveries for human taxonomists to investigate.
AI systems “don’t have intuition; they don’t have creativity,” said Rolnick, whose team co-created Antenna, a ready-to-use AI platform for ecologists. Human taxonomists are still better at imagining how a rare feature arose evolutionarily, or exploring the slight differences that can mark an entirely new species. And ultimately, every identification—whether by algorithm or DNA or human expert—still depends on people.
That human labor is also a dwindling resource, especially in entomology. “The number of people who are paid to be taxonomists in the world is practically nil,” Rolnick said. And time is against them. The world’s largest natural-history museums hold a wealth of specimens and objects (more than 1 billion, according to one study) yet only a fraction of those have digitally accessible records, and genomic records are accessible for just 0.2 percent of biological specimens. Many historical collections—all those drawers packed with pinned, flattened, and stuffed specimens; all those jars of floating beings—are chronically underfunded, and their contents are vulnerable to the physical consequences of neglect. Preservation fluids evaporate, poor storage conditions invite pests and mold, and DNA degrades until it is unsequenceable.
Today’s tools are still far from fully capturing the extent and complexity of Earth’s biodiversity, and much of that could vanish before anyone catalogs it. “We are too few, studying too many things,” Caruso, the Padua entomologist, said. Many liken taxonomy to cataloging an already burning library. As Mehrdad Hajibabaei, chief scientific officer for the Center for Biodiversity Genomics at the University of Guelph, in Canada, told me: “We’re not stamp-collecting here.” Taxonomists are instead working to preserve a planetary memory—an archive of life—and to decode which traits help creatures adapt, migrate, or otherwise survive in a rapidly changing climate.
The climate crisis is unraveling the life cycles of wildlife around the world—by one estimate, for about half of all species. Flowers now bloom weeks before pollinators stir; fruit withers before migrating birds can reach it. Butterflies attuned to rainfall falter in drought. Tropical birds and alpine plants climb toward cooler, though finite, mountaintops. Fish slip farther out to sea; disease-carrying mosquitoes ride the heat into new territories. Extreme weather at the poles stresses crucial moss and lichen, and shreds entire habitats in hours. Mass die-offs are now routine.
“Once you lose one species, you’ll probably lose more species,” Caruso said. “Over time, everything is going to collapse.” One in eight could vanish by century’s end—many of them dark taxa, lost before we ever meet them. Most countries—and global bodies such as the International Union for Conservation of Nature—cannot assess, and therefore cannot protect, unnamed organisms. As Edward O. Wilson told Time in 1986: “It’s like having astronomy without knowing where the stars are.”
Today’s machine-assisted taxonomy faces the same problem Linnaeus did: Nature’s complexity still far outstrips human insight, even with machines’ assistance. “We don’t perceive the world as it is in all its chaotic glory,” the biologist Carol Kaesuk Yoon wrote in her 2010 book, Naming Nature. “We sense a very particular subset of what surrounds us, and we see it in a particularly human way.” On the flip side, every new data point sharpens the predictive models guiding conservation, says Evgeny Zakharov, genomics director for the Center for Biodiversity Genomics. “The more we know about the world, the more power we have to properly manage and protect it,” he told me. With tools, the speed of taxonomists’ work is accelerating, but so is the countdown—they will take all the help they can get.
The post A Revolution in Tracking Life on Earth appeared first on The Atlantic.
