“Drones Fight Poachers" has an undeniable sexiness to it as a news narrative. Who doesn’t want to read about flying killer robots battling machete-wielding criminals chasing innocent animals on the wild African plains? The instant appeal of a high-tech solution to a pervasive low-tech problem is also why Silicon Valley giant Google has given the World Wildlife Fund (WWF) $5 million for drones to stop poaching. But to actually stop poachers, WWF should focus less on drones and more on math—and some lessons learned in Iraq and Afghanistan.
University of Maryland computer scientist Thomas Snitch is applying a mathematical forecasting model he developed for use by the military in Iraq and Afghanistan to Africa. Snitch is trying to overcome poaching networks’ advantages in money, opportunity, and manpower using his military model to put park rangers in the right places to intercept rhinoceros killers.
Illegal poaching and terrorism don’t seem to have much to do with another. In fact, the relationship is deeply interwoven. First, there’s the money. The State Department recently warned that the proceeds from illegal poaching go to African terrorist organizations like the Lord’s Resistance Army, the Janjaweed and al-Shabaab. But that’s not the only connection. The tools of poaching also look like tools of war. In places like South Africa’s Kruger National Park, here’s how that war appears. On one side are park rangers who make $150 per month protecting endangered animals such as the black rhino. Opposing them are networks of poachers who are more numerous, better armed, and better funded. Today, one rhino horn sells for $50,000 a kilogram on the black markets of Vietnam, according to commonly cited figures from researcher D. Graham-Rowe. The price can reach more than $1 million. It’s one reason why there were more than 1,000 illegal rhino killings in South Africa last year according to South Africa’s Ministry of Environmental Affairs.
The problem with the way groups like the WWF use drones is that they treat unarmed drones as a silver bullet, says Snitch. They succumb to the temptation to spend too much money on systems that are difficult for the rangers to work with, that are costly to repair, and that won’t actually stop poachers. They also don’t know how to use the drones correctly. “I ask these people all the time, ‘You have this device. Where do you plan to fly?’ ‘Well,’ they say ‘we'll fly back and forth.’ OK, Kruger is the size of Israel.” It’s a math problem. One he’s faced before.
In 2010, Snitch’s day job was trying to predict where insurgents in Iraq were going to plant roadside bombs. He approached the problem as one of incremental elimination. If he could figure out those spots where the insurgents were definitely not going to place IEDs then he could make sure that the most vulnerable areas were covered.
“We looked at every IED explosion over the last five years and pinpointed [the explosions] on the maps,” Snitch explained to a group gathered at the University of Maryland in the spring of 2013. “On top of that we overlaid where the U.S. troops were when they were hit and what roads they were on. We then took drone intelligence and human intelligence. We came to the conclusion that when an IED blast goes off there’s a 90-percent chance that the bomb factory is between 685 and 750 meters from the explosion. Does that tell you the exact house? No. But the commander knows the perimeter to start looking. We looked at the patterns of where the insurgents moved, what coverage they had, and where they can hide.”
His biggest asset in anticipating roadside-bomb sites was high-resolution satellite imagery. “We were able to develop a map of Baghdad down to 16 inches … very precise location and terrain.” This data made it easier to understand the factors that made one IED site more attractive than another. Though Sunni-on-Shiite violence was common, the primarily Sunni insurgents weren’t likely to plant bombs in Shiite areas because the risk of being in the wrong place at the wrong time was too high.
Neighborhood dynamics also played a key role in bomb placement. Which families lived on which streets? Which areas got foot traffic, and at which times of day? How much sunlight or shade does a particular street corner receive? How close is the nearest escape route? Where does it go? Understanding how all of these variables relate mathematically to influence the behavior of someone looking to plant a bomb was once unfathomable. It becomes possible with lots of satellite data, which, in 2010, was also opening up for civilian use. Now, a lower price and greater number of publicly available pictures from space is what may save endangered species like the black rhino.
Snitch had been to Africa several times as a tourist and was aware of the poaching problem. He recalls one day sitting in a meeting with U.S. military planners, discussing how to bring down the number of IED attacks in Baghdad, when he reached an important realization. “I said to myself, wait a minute, we’re looking at insurgents that use IEDs at U.S. military targets. Would this same model work for poachers?” In fact, it does.
Just as poaching goes to fund terrorism and poachers are outfitted like armies, so the battle against them looks like the battle against insurgents.
While much of the poachers’ behavior can be analyzed independently, Snitch’s model attempts to figure out where the rhino will be at those times when poaching is most likely—that is, around 8 or 9 p.m., when the full moon is out during the dry season. (The poachers prefer to operate soon after dark to maximize the amount of time for escape on foot before dawn.) If you know where the rhinos are, you can anticipate where the poachers will strike. But predicting where a single animal is going to be, at night, in a 7,523-square mile game preserve roughly the size of Wales is a riddle of extravagant complexity.
GPS-enabled tags, collars, and chips on the rhinos or even inside the horns themselves is one tactic that the World Wildlife Fund is supporting in Kenya, which has announced plans to implant chips into the horns of the more than 1,000 rhinos in that country. But low-jacking wildlife is hardly an ideal solution. Predicting where the rhino will go while putting trackers on the fewest number possible requires a lot of environmental data. For his model to function, Snitch needs as much information as possible about where the rhinos have been spotted. He also needs to understand the landscape on an extremely minute level, ideally down to every wild patch of prickly pear (a favorite rhino treat). In essence, he’s trying to quantify one of the most random environments imaginable: the wild. As uniquely difficult a math problem as this sounds, it’s one he has experience with in a different context.
With current technology, Snitch can see objects 16 inches in diameter, or “about the size of home plate,” from space. But satellite photos don’t run in real time. For the model to perform correctly, he needs park rangers to report rhino sightings, fence breaks, tracks, unusual plants or animals, or other abnormalities using hand-held devices and drones.
“Those are your eyes,” says Snitch of the drones. “If I've got a pretty good idea where these [poachers] are going to be, [the drone] helps me fine tune the deployment of the rangers a little bit…. They protect rangers because they see if there's anyone out in the bush waiting for them. It basically gives you the edge as these guys are coming to you.”
He’s extremely skeptical of the WWF program and the way that most groups and governments use drones in anti-poaching efforts. The export of military-grade equipment to countries like South Africa for use in anti-poaching efforts is illegal. The U.S. State Department also restricts the quality of camera equipment aboard drones. The drones that can fly over Kruger will much more closely resemble the unarmed Falcon, a favorite of many police departments. They’re cool and easy to launch but toothless. The actual enforcement of anti-poaching laws will still fall on the shoulders of people—the rangers. Kruger National Park, like most of the game reserves where rhinos are under threat, is too large and the poaching opportunities too numerous for drones alone to make a big difference. The park rangers can only move 10 to 12 kilometers a minute from one spot to another, so even if a drone flying a random flight path happens to see a poacher, the information is useless unless rangers already are within striking distance.
Snitch believes that he can cut poaching in Kruger Park, at least, by 70 percent with just $450,000—a fraction of what Google gave to WWF for drones. If he accomplishes that in South Africa he hopes to convince leaders in other countries that the model works. A simple mathematical formula (supported by mountains of data) could give authorities a badly needed advantage to stop poaching and, with it, maybe even terrorism.