Three years after Australia’s most devastating fire season ever, new images from 1,100 motion-activated cameras placed across the country’s scorched forests are giving researchers an unequaled view into wildlife recovery — and the results are more promising than anticipated.
Project “Eyes on Recovery” — a collaboration between the World Wildlife Fund (WWF) and Conservation International, and supported by Google — has gathered more than 7 million photos, with many showing vulnerable species returning to their habitats. The cameras snapped koalas, a rare group of echidnas (also known as spiny anteaters), a female wombat and her joey, dingo pups, and critically endangered Kangaroo Island dunnarts (mouse-sized marsupials).
Female wombat and her joey. ©Grant Linley/Charles Sturt University
“We didn’t know what to expect — some of these species saw up to 90 percent of their habitat destroyed,” said Conservation International wildlife scientist Jorge Ahumada. “We’re finding that native species are more resilient than we thought.”
The “Black Summer” bushfires of 2019-2020 killed or displaced nearly 3 billion native animals. To understand how wildlife would recover, researchers turned to Wildlife Insights, a platform powered by an artificial intelligence (AI) model developed at Google and operated in partnership by Conservation International, the North Carolina Museum of Natural Sciences, the Smithsonian’s National Zoo, the Wildlife Conservation Society, WWF and other organizations.
A train of short-beaked echidnas on Kangaroo Island. ©Kangaroo Island Landscape Board
Initially, Wildlife Insights didn’t have data on Australian wildlife — meaning, the AI model had to learn to distinguish a wombat from a pig, or a kangaroo from a deer. Now, it can identify more than 150 Australian species with 90 percent accuracy on average, said Ahumada, Wildlife Insights’ executive director.
Without the AI technology, researchers would have had to review millions of images one by one — a tedious and time-consuming task. Wildlife Insights processed the images captured by motion-activated cameras five to ten times faster than a human — thanks in large part to its ability to filter out blanks, which can account for the vast majority of images.
The analysis not only proves the presence of vulnerable species — it also provides information on their recovery. Koalas, for example, were among the hardest hit species. An estimated 60,000 were killed, injured or displaced as wildfires tore through the eucalyptus forests where they live. The damage was so devastating that last year the Australian government declared koalas endangered across much of eastern Australia, citing impacts from the wildfires, drought and habitat loss. Field cameras captured koalas roaming the forest floor, which is unusual for a species that spends most of its time perched in trees. It’s an indication that the forest is still recovering, but wildlife is returning.
A koala carrying a joey in the New South Wales Blue Mountains. ©Science for Wildlife
The cameras also recorded images of invasive species, such as feral cats, foxes and cane toads, which can kill or compete with native species.
“In a place like Australia, where invasive species are pervasive, it’s really important to understand how fire can lead to an increase in invasive species, which often colonize areas that are disturbed,” Ahumada said.
Researchers say the response to natural disasters like wildfires is often driven by incomplete information. Satellite imagery shows the extent of the damage, but the impacts to wildlife can be less clear.
“Eyes on Recovery” could change that by offering new data into nature’s resilience — and where native species need additional support recovering. This could inform investment priorities and more proactive measures like protecting areas where fires are less likely to occur so they can serve as habitat refuges — or installing food and water stations after a fire for species that are slower to return to their natural habitats.
“We talk about resiliency in nature frequently, and how we must increase it in the face of climate change and more frequent natural disasters,” Ahumada said. “But what exactly does that mean? These cameras — and the AI we’ve developed — are giving us clues that can support species’ survival in future wildfires.”
Further reading: New tech could transform science of wildlife ‘selfies’