Researchers have developed an innovative algorithm that turns vultures themselves into a smart warning system and prevents mass poisoning events.
The situation of eagles in Israel is worrying. Eagles, nature's 'cleaners', maintain the ecological balance by cleaning up neglected open areas. When they disappear, the entire chain of life is affected. One of the main threats to eagles is poisoning. Researchers School of Zoology in the George S. Wise Faculty of Life Sciences, in collaboration with the Nature and Parks Authority, are fighting for every eagle. After returning dozens of eagles to the wild Transmitters that show where they are at any given time have now been developed in the laboratory of Prof. Or Spiegel from the School of Zoology and the TAD Center for Artificial Intelligence, in collaboration with researchers from Spain and the USA, and representatives of the Nature Reserves and Parks Authority, an innovative algorithm that turns the vultures themselves into a smart warning system. With the help of real-time analysis of the vultures' movements, suspicious feeding sites can be located and warnings given in advance, before mass poisoning events occur.
The study, published in the Journal of Applied Ecology, combines advanced GPS tracking with accelerometers and machine learning models to remotely monitor and classify the behavior of biblical vultures. According to the researchers, widespread use of the innovative technological approach will make it possible to protect a wide range of vulture species worldwide, about 70% of which are classified as endangered, often as a result of some type of poisoning.
The eagle landed.
In Israel, the biblical vulture population has plummeted from thousands to fewer than 200 individuals. Poisonings are responsible for about 57% of deaths, with poisoned carcasses causing the deaths of dozens of vultures at once. “Vultures are social animals, so one carcass could lead to the collapse of a significant percentage of the local population,” explains Gideon Adia, one of the study’s leaders.
The research team installed transmitters on biblical vultures, and used captive and wild individuals, in Israel and Spain, to train a Random Forest classification model to identify vulture behavior. The model was trained to identify various vulture behaviors – from flight to feeding – and succeeded with an impressive accuracy of up to 96%. A significant innovation in the study is the use and validation of the system’s “certainty level” mechanism, which allows for the distinction between correct and false identification, and can help inspectors distinguish between real feeding events and false alarms, thereby significantly improving the reliability of the alerts and preventing inspectors from being unnecessarily distracted.
Hoping all the stars align
"As always, the challenge is in the details," says Prof. Or Spiegel. "Machine learning models are 'hungry' – they need hundreds of examples of each behavior. Some of the rare behaviors, like eating, were of great interest to us. To catch the right moment, that is, a broadcast eagle, observed exactly as it is eating, requires a lot of luck and patience. For 'all the stars to align.' This doesn't happen every day, so the task required quite a bit of patience, especially in observations in nature."
Another challenge the researchers faced was "coordinating and creating a common language between the various parties, so that we could connect the dots between the data collection, the computational researcher who helped with the model, the student who analyzes other aspects, and the interface people who need to understand what is important and why it helps them," adds Prof. Spiegel. The researchers are now working to integrate the algorithm into RTG's alert system, to fully integrate the development.
To anticipate a cure for a blow
The technology has proven itself during field trials in the Judean Desert: the model successfully identified vulture feeding sites that occurred outside of safe feeding stations operated by the Israel Nature and Parks Authority, and most of the carcasses independently discovered by inspectors were found in the risk areas predicted by the algorithm. "Together with the alert system developed in recent years by the Israel Nature and Parks Authority, this system could improve the way vulture conservation works," notes Dr. Martha Acacio. "Instead of reacting after the birds have already been poisoned, we can now identify dangerous areas in advance, focus sanitation efforts, and remove carcasses that may be poisoned even before the vultures find them."
Prof. Or Spiegel's research team during fieldwork in the desert
"The Nature and Parks Authority, together with its partners in the 'Spread the Wings' project (the Society for the Protection of Nature and the Israel Electric Company), is investing great resources and efforts in protecting eagles and birds of prey from extinction. Among other things, the authority is investing through collaborations with researchers from academia, developing innovative technologies with the help of AI technologies, but also through hard work in the field of providing safe food at feeding stations and disposing of carcasses with a risk of poisoning, thus leading to environmental improvement, with the support of the Ministry of Environmental Protection," adds Ohad Hatzofe, avian ecologist at the Nature and Parks Authority.
Is there room for optimism?
When Prof. Spiegel was asked how he sees the future of vultures in Israel in a decade, he answered honestly: "I'm afraid we're still in a battle of attrition. The warning system that RTG has established is a significant leap forward, and the ability to identify feeding sites will also contribute and improve the fight, but there are so many challenges, both here and in neighboring countries, that it's hard to be optimistic about the local population. Vultures will not disappear from the world. In Western Europe, conservation programs are very successful, but here, in the Middle East, they may go the way of the buzzards and peregrines and disappear as nesting birds from the country."
"That doesn't mean that everything we do is in vain, quite the opposite. It's a long, Sisyphean war of conservationists who gain time, slow the decline, and sometimes even manage to stop the wheel. Let's hope that we can buy the eagles enough time until humans change their ways, stop poisoning, hunting, and disturbing, and extinction will be stopped not thanks to a single technological solution, but thanks to a broader change."
But even within this struggle there is hope and new initiatives. "Absolutely," Prof. Spiegel smiles. "We are already expanding the algorithm's capabilities, for example, to identify nests: who laid an egg, who is incubating, and where. This way we can help the eagles even at the beginning of their lives, and not just prevent their premature end. The eagles, in the end, do the monitoring for all of us."
To encourage adoption of the method in other places and species around the world, the research team has made the entire dataset, algorithms, and training materials freely available online, allowing conservationists around the world to adapt the technology to different species and regions. In an era where human activity threatens wildlife, the study demonstrates how technological innovation and knowledge sharing can help conserve nature in the 21st century.
More of the topic in Hayadan:
