Last year, the Australian government signed a global pledge to reverse biodiversity loss by 2030.
The pledge is one of a plethora of political drivers of recent years that have seen technologies like the Internet of Things, artificial intelligence, drones and other wizardry become commonplace on rewilding projects across Australia.
Rewilding aims to return threatened species to ecosystems where they’ve become extinct due to hunting, habitat loss or degradation, or other pressures.
Importantly, rewilded ecosystems may help mitigate climate change, by “increasing carbon removal from the atmosphere and protecting against climate change impacts by reducing soil erosion and flood risk”.
The Australian Wildlife Conservancy (AWC) is a donor-funded not-for-profit that owns, manages, or works in partnership with rewilding projects on more than 12.9 million hectares of Australia where they establish safe havens – fenced areas called ‘exclosures’ – from which feral predators such as cats and foxes have been eliminated.
As of 2023, there were 10 AWC safe havens in Australia where 16 mammal species have been reintroduced, with another 3 mammal species inhabiting 2 sites where predators are controlled but not eliminated.
Both types come under the care of Damien Kerr, chief technology officer for AWC. Kerr’s fresh-faced passion for the gadgets he hopes will achieve a sustainable future for the country is infectious.
“The best way of using technology for conservation is to be able to pump through the data as quickly as possible,” says Kerr, “to get the information straight back to the ecologists in near real-time.
“When there’s a flooding event that might affect a fence, we need to ensure our gates are locked. If a gate remains open, an animal [may] get through.”
Maintaining the fences to protect an AWC exclosure is of utmost importance, he says, because such a huge investment goes into building them.
That’s why Kerr keeps more tricks up his sleeve than just the checking of boundaries.
For instance, a remote camera that automatically switches on if it detects motion nearby, then without missing a beat takes a picture of the animal.
Clever, right? Trouble is, the camera swiftly accumulates data, and thousands of images might need processing within a few hours or over many days to help keep an exclosure predator-free.
“They can monitor over a period of time,” says Kerr, “[but] it can take months to see whether there is any indication of a remaining cat or fox within that area.
“And only after a long period of analysis of these many, many hundreds of thousands of images can we safely and confidently say that an area is feral predator-free, where we can then introduce the threatened species.”
Often when such a camera detects an intruder, a hunt follows, with experienced trackers or rangers brought in to trap or kill a predator.
Infrared cameras that detect movement at night then come in handy. Such technology was pivotal in the hunt for “Rambo”, a feral red fox (Vulpes vulpes) that avoided capture for more than 4 years at Pilliga State Conservation Area, hampering and delaying the multi-million dollar rewilding of the 35,000ha area of high-value forest and woodlands in northwest NSW.
Rambo’s persistent survival sparked a massive foxhunt lasting from mid-2018 to late 2022.
It can take months to see whether there is any indication of a remaining cat or fox within that area.Damien Kerr
During that time, 97 cameras were used to detect and monitor him in a hunt that clocked 10,400 nights of trapping and 73 of shooting, the setting of 3,500 baits and more than 55 days on which scent-tracking dogs were used.
After the flooding of the area last October, Rambo disappeared from the cameras. In March the area was declared predator-free.
Thankfully, a specially constructed breeding area within the wider fenced area had allowed AWC and National Parks and Wildlife Service to cradle 3 species – the bilby, the bridled nail-tail wallaby, and the brush-tailed bettong – in readiness for the declaration, when the animals were moved into the wider space. Populations are now reportedly booming.
Another must-have technology is the drone. A total of 34 western quolls (or chuditch in the local Noongar language) were translocated at Mount Gibson Wildlife Sanctuary in Western Australia during April and June using drones and VHF radio tracking collars, which together enabled ecologists to monitor the progress and movement of 16 individuals.
“In the past, ecologists would have to drive around the antennas, attempting to locate the signal of that tagged collar,” says Kerr. “We’re now able to do that with drones.”
And that’s a huge time saver for AWC Regional Ecologist Dr Amanda Bourne and her team, who conduct science programs at sanctuaries in southwest WA.
“We use a drone with a radio-tracking payload from wildlife drones,” says Bourne. “We can send it up for a 20-minute flight … and that can net more data than you might think.
“We did it yesterday and detected pings from every collared animal we had released in the translocation last week, and that’s very encouraging. It’s good to be able to quickly get some simple information that yes, they’re still on the property; yes, they’re still alive.
“And then we send teams out on the ground with the hand-held antenna [to] find them in the hollows and make sure they’re okay.”
Coupling the drones with thermal imaging, can also make other harder-to-reach spots accessible.
“We did some thermal imaging on one of our sanctuaries called Faure Island,” says Kerr, “up in Shark Bay on the northwest coast [of WA] … a natural sanctuary; cats are unable to get to it.
“There are a number of … burrowing bettong; we wanted to survey how many were active on the island. They’re nocturnal and … it’s very difficult to see at 2 or 3 o’clock in the morning, so we put up a thermal drone to track them.”
The thermal imaging illuminated a much wider area than the ecologists would have been able to canvas using torches, making it easier to know how many bettong were moving around.
“We were able to learn quite a lot about their behaviour.”
But it is artificial intelligence that seems to hold the greatest promise.
Adding AI to remote cameras has meant ecologists can now use them to detect not only animal movement, but to start to recognise the type of animal causing the motion (cat, fox, cow, dingo). They may even be able to identify an individual animal by its colour, striping, or number of spots.
The latest machine learning technology was used in just this way on the Species Classifier Model V4, which was trained to identify common species in a specific area by using over a million images.
According to AWC, the classifier was able to identify species with an accuracy of more than 95%.
As a result, scent-lured camera traps are now being set up to identify individual quolls from the pattern of their spots, a technology first developed by ecologists at Mornington Wildlife Sanctuary in the Kimberley in 2013, and based on PhD work at the University of Adelaide.
But again, the amount of data is daunting.
“At the moment it’s an optimistic research goal,” says Kerr, “identifying up to 42 target species that are specific to AWC’s sanctuaries. I believe we’ve processed over 9 million images to date, coming up to 10 shortly.
“That’s a massive amount of data and quite an extraordinary amount of sorting for the AI to do and to separate. One of our staff used the AI platform to process 590,000 images in bulk; the entire process from AI processing to human verification took about 72 hours, compared to the many months it would ordinarily have taken.”
And believe it or not, AI is even helping to manage cats.
I believe we’ve processed over 9 million images to date, coming up to 10 shortly.Damien Kerr
“Cat behaviour is difficult,” says Bourne, “they can be difficult to track, but will groom themselves if they get something sticky on them: The Felixer takes advantage of that.”
So, what’s next for rewilding tech?
“We’ve just had a couple of deliveries of equipment that represent an exciting new frontier for our AI,” says Kerr. “My colleagues and I are very keen to start using that; we’re going to be attempting what’s called AI on the edge …”
That entails “doing the processing of image recognition at the point of the camera, rather than afterwards”, in other words, right when and where you’ve taken the photo.
“In doing so,” he says, “we’ve been able to gather some rather unique equipment … with a little pair of cameras and infrared eyes …”
The Greenlight Project is a year-long look at how regional Australia is preparing for and adapting to climate change.