Wildlife Tracking 2.0

New AI technologies are offering more humane alternatives to invasive animal monitoring practices.

Studies show that affixing a tag or a collar to an animal, like this mountain goat in Glacier National Park, can have long-term impacts on behavior and health. Photo by Tom Driggers.
Studies show that affixing a tag or a collar to an animal, like this mountain goat in Glacier National Park, can have long-term impacts on behavior and health. Photo by Tom Driggers.

A BLACK RHINO RACES ACROSS the savannah, desperate to escape the helicopter on its tail. A man hangs out of the craft, which hovers just a few feet above the ground, pointing a dart gun at the frightened mammal. He takes aim and fires, lodging a sedative-carrying dart in the rhino’s rear. The rhino continues to run for a time, then slows, and finally crashes to the ground, succumbing to the drugs pulsing through her veins. The helicopter lands, and several people disembark to surround the downed animal. They blindfold her and bore a hole in her horn, one just large enough to implant a transmitter. Then they inject an antidote to reverse the sedative. When she awakes, the transmitter will enable the team to track her whereabouts, until the battery runs out at least. The well-intentioned exchange takes less than an hour. But it may haunt the rhino for much longer.

Across Africa — and indeed, across much of the world — similar transactions play out every day. Faced with declining populations and mounting threats to animals big and small, scientists and conservationists are desperate for information. Where are rhino — and caribou, and eagles, and sharks — roaming? Where are they finding food? Where are they facing danger? Monitoring techniques, like radio collaring, banding, and trapping in nets, hold the promise of more data, data that could be used to stave off major threats. But these techniques can be invasive and can cause animals lasting trauma. Is there a better way?

A slew of new artificial intelligence technologies offers more humane alternatives.

THOUGH RESEARCH ON THE subject is limited, studies show that affixing a tag or a collar to an animal can have long-term implications for behavior, health, and even survival. It’s no wonder. Invasive monitoring involves chemical or physical immobilization of the animal. Subjects who aren’t shot with a sedative-carrying dart are typically caught in traps, facing stress and sometimes pain until released. However they are caught, the stress of the experience can, in extreme cases, lead to capture myopathy, a syndrome where the animal dies from hyperthermia resulting from extreme muscular activity within minutes to weeks of being captured. This syndrome has been reported in rodents, carnivores, primates, birds, marsupials, ungulates, and pinnipeds.

More commonly, the animal survives the stress of capture. But the experience could still change their behavior, which can have implications for both research results and for animal health and safety.

Take crows, and other intelligent birds, for example. Once caught, they tend to be cautious of the people and devices that caught them. (Yes, research indicates that crows can recognize and remember faces.)

“Crows do become much more wary of the trap and their behavior changes post trapping, but mostly with respect to how they behave the next time they see a trap, or the next time they see the person that trapped them,” says Kaeli Swift, a postdoctoral researcher with the University of Washington whose work focuses on crows.

In other cases, the tags or bands or collars fitted to animals can impact behavior. Penny Hawkins, a researcher with the Royal Society for the Prevention of Cruelty to Animals, noted in a 2004 review of the subject that the act of carrying a tracking device itself can cause stress and can require additional energy expenditures. Other research indicates that tags can impact reproduction and health. A 2014 study of fur seals, for example, found that bigger tags on mothers correlated with lower pup growth, and a 2011 study of king penguins found that those with tags were less likely to breed successfully in resource-poor years.

Other research, such as that involving big cats, suggests that after a short adjustment period, most animals tend to get used to their tracking devices. But overall, information on the subject is sparse. And as Steven Portugal, a professor at the Royal Holloway University of London, who deploys biologging devices on captive animals, told Mongabay: “If you’re significantly impacting the behavior, health, or welfare of the animal, that’s terrible from the welfare perspective, but also if that animal is not acting or behaving normally, the actual scientific data you’re getting would be useless as well.”

GOING WITHOUT the data might be a hard sell to those working in wildlife conservation. Getting similar data via less invasive means, on the other hand, would seem to be a win-win. At least that’s what a handful of burgeoning initiatives are hoping. These outfits, some of which are supported by large tech companies like Microsoft and Google, say simple images, the kind you might grab on your smartphone, can provide similar wildlife data to traditional tagging.

One such organization is WildTrack. The North Carolina-based international nonprofit was founded in 2014 by veterinarian Zoe Jewell and wildlife biologist Sky Alibhai. The pair had been monitoring black and white rhino populations in Africa in the early 1990s when they began to document that invasive monitoring techniques were reducing the fertility of female black rhinos. As they searched for alternative monitoring methods, they began to work with Indigenous trackers adept at identifying animal footprints. Through this experience, they developed a footprint identification technology that they branded “FIT.”

The technology essentially combines local knowledge with complex statistical algorithms — or machine learning — to identify species, age, and sex of an animal using an image of its footprint. It can even identify individual animals. Local community members, scientists, and tourists can all submit photographs to Wildtrack’s growing database, and progressive learning allows the algorithm to adapt and improve with time. So far, Wildtrack has developed algorithms for 15 endangered species, ranging from the black rhino to the Eurasian otter to South America’s tapir.

Another Portland, Oregon-based conservation nonprofit, Wild Me, has developed a similar program called Wildbook. This software uses photos to identify animals by unique characteristics, such as the pattern of a cheetah’s coat, a zebra’s stripes, or the shape of a whale’s fluke. Scientists can build their own image library by going out into the field armed with little more than a camera.

“Before a population assessment wasn’t something you could do in a weekend. It’s incredible,” Jenna Stacy-Dawes, a research coordinator with San Diego Zoo who has used Wildbook to count reticulated giraffes in northern Kenya, told National Geographic. “It’s been really helpful in allowing us to work faster and understand the population better than we ever really could have before.”

Programs like WildTrack and Wildbook are cost-effective. To grow a database, all that’s needed is a device to take photos, a GPS unit to determine field location, and a voice recorder, pen, or pencil to take down notes.

These tools can also help conservationists crowdsource information from citizen scientists around the globe, or use the Internet to grow their image database. Wild Me, for example, has been using a bot to scan YouTube for videos of whale sharks, which can be identified by their unique pattern of spots. The bot grabs a still image from the video, along with date and location information provided in either the video file or the caption, and uploads the information to Wildbook’s whale shark database. Its identification powers are constantly improving.

“Letting this thing loose on YouTube, especially with migratory species that are out in the ocean, you get this chance of finding outlier sightings of animals where researchers just simply aren’t going,” Jon Van Oast, a senior engineer at Wildbook who came up with the whale shark bot concept, told National Geographic. “It goes places where the researchers just can’t for logistics reasons and funding.”

AS WITH MOST THINGS, there are certain limitations to these technologies. Though WildTrack can inventory footprints left in mud, snow, dirt, or even pollen, not all footprints are created equal. Research on Bengal tigers in Nepal indicates that the best data comes from footprints left in wet soil or sand, and WildTrack’s application is inherently limited to animals that leave a literal mark on the Earth, which excludes marine mammals.

“They’ve got a very limited scope, and they’re not so good for birds just because of the size and movement of birds.” Kaeli says of non-invasive monitoring methods.

And of course, the use of digital monitoring and artificial intelligence raises broader ethical concerns about who owns the data, who controls the machine learning algorithms, and how the collected information is used.

Still, in a world where wildlife faces so much stress from deforestation, climate change, habitat encroachment, poaching, and more, it’s hard not to be enticed by the potential for less-invasive monitoring. Already, these types of technologies are being used to track animals in remote and inaccessible regions; to predict poaching vulnerability and direct ranger patrols accordingly; and to integrate imagery from satellites, camera traps, citizen scientists, professional researchers, and the web. They provide one more tool for conservationists to do their work. As Zoe Jewell sees it, these tools are the future of wildlife monitoring as they allow researchers to collect cost-effective, animal-friendly, and reliable data. “We need to know where endangered species are most at risk, and we have little time left,” she says.

Get the Journal in your inbox.
Sign up for our monthly newsletter.

Subscribe Now

For $15 you can get four issues of the magazine, a 50 percent savings off the newsstand rate.