This app could help turn the tide on poaching

The newest weapon in the fight against wildlife poaching and illegal logging doesn’t rely solely on boots or bullets, but on a computer software application driven by artificial intelligence.

Called Protection Assistant for Wildlife Security, or PAWS, the app uses algorithms – similar to the mathematical models used to power computer games – to devise strategies for defeating those who seek to destroy nature.

Thousands of animals are illegally slaughtered each year for their ivory, skin and bones for use in traditional medicines and feeding black market demand, and others meet their death at the hands of illicit trophy hunters, in what the World Wildlife Fund refers to as “a global poaching crisis.”

Last year, about 30,000 elephants and a record 1,338 rhinos were killed in Africa, and tens of thousands of other animals were poached and trafficked, “feeding an illegal wildlife trade that is increasingly being driven by international organized crime,” the international conservation group said in a recent statement.

Developed in 2013 and still in the test phase, the PAWS software analyzes data on terrain, topography, routes most frequently used by animals and paths traveled by poachers, said Milind Tambe, a computer scientist and professor of industrial systems engineering at the University of Southern California who developed the technology with his doctoral students, including Fei Fang and Thanh Nguyen.

The Army Research Office and the National Science Foundation funded the research.

“We can do pattern recognition, essentially building a model and predicting poachers’ activity based on their past actions,” Tambe said.

“We can build up predictions of where poachers may strike. Then that allows us to generate patrol strategies that would be effective against those types of poachers.”

For example, the system can generate random patrol routes for rangers in order to keep potential poachers guessing about the wardens’ whereabouts.

The technology has been tested in Malaysia and Uganda, where it was successfully loaded onto hand-held GPS devices that teams of rangers could use.

Patrols will be launched in Queen Elizabeth National Park in western Uganda in the coming months. Uganda is home to some of the world’s most threatened ecosystems. Andrew Plumptre, director of the Albertine Rift Program for the Wildlife Conservation Society in Uganda, said he was eager to see the app put to the test.

Plumptre said ranger patrols account for the greatest portion of the Uganda Wildlife Authority’s budget for all of its parks and wildlife reserves, and recent studies show an average of 42 percent of households around two of Uganda’s key elephant strongholds acknowledged hunting in the parks in the last year.

Although many of the hunters do so out of a desperate need for money to pay for essentials such as health care or school fees, “there are regular poachers who can earn significant amounts of money from the hunting,” Plumptre said in an email.

“Making the patrolling more efficient and effective could greatly improve funding for other activities that are also needed, such as removal of invasive plants, and programs with the local communities,” he added.

In Malaysia, where Tambe and his researchers teamed with the wild cats conservation group Panthera to test guided patrols using the artificial intelligence technology, the program yielded positive outcomes.

A research paper written by the USC team, Panthera and other researchers noted that PAWs was able to guide rangers toward poaching hot spots by suggesting feasible and detailed routes. The technology also proved beneficial “at finding good ridgelines that are taken by animals and humans” and easier for the rangers the follow, according to the report.

Having the software do the work could potentially reduce the burden of patrol planning, the researchers found.

But they also found limitations. For example, one of the routes PAWS suggested for rangers on foot patrol in Uganda traversed a body of water, forcing the wardens to walk along the water’s edge and slowing them down. Extreme changes in elevations on routes suggested by the software could stop patrols in their tracks. The technology was unable to provide detailed patrol routes in large conservation areas, and questions remain about maintaining and upgrading the software, Tambe said.