A US-funded initiative is using artificial intelligence algorithms used in online gaming to predict the movement of poachers in Africa’s nature reserves
The system, dubbed PAWS, is currently being tested in Uganda’s Queen Elizabeth National Park. The park’s rangers can obtain information from computer models that predict the poacher’ next steps.
The software, based on game theory, anticipates the behaviour of rational human beings in certain situations, be it a game of online poker or criminal behaviour.
“We’re trying to predict future poacher attacks or where poachers may strike next based on what we have observed in the past on our patrols,” Professor Milind Tambe from the University of Southern California, who is leading the initiative, told Reuters.
“We want to deliver the software so the local rangers can use it with minimum training.”
The researchers trained the artificial intelligent system using data collected by rangers over the past 18 years. Poaching is a massive problem in Africa, threatening many already endangered species. African elephant numbers, for example, fell by 20 per cent between 2006 and 2015 because of a surge in ivory poaching, according to conservationists.
There have been some successes. Uganda has seen its elephant population recover from a nadir of around 800 in the 1980s after years of chaotic rule and dictatorship to reach about 5,000 now. However, officials say poaching is back on the rise.
“Elephant poaching has started to increase in the last five years, which is a major concern and that’s linked to the global price of ivory,” said Andy Plumptre, director of the Albertine Rift Program for the Wildlife Conservation Society in Uganda.
Poachers can sell one kilogram of ivory on the black market for around $1,000. Even local villagers sometimes resort to poaching to obtain food or money to pay school fees.
Killing game animals, meanwhile, has knock-on effects. It destroys the prey chased by carnivores such as lions and leopards, causing their numbers to drop, too.
The PAWS programme has already delivered some benefits. In just one month, researchers managed to locate 10 antelope traps and elephant snares. Deploying the system in remote natural parks could prove challenging, due to poor mobile and internet connectivity. The connectivity problems threaten accuracy, but also make it impossible for rangers to obtain real-time data.