In the past decade, ocean refuges have been multiplying faster than research can keep up. Under President Barack Obama, the area of U.S. waters considered “strongly protected”—meaning commercial fishing is prohibited but recreational fishing with permits is allowed—more than quadrupled. Today, fully-protected preserves known as Marine Protected Areas cover around 400,000 square kilometers, or 3 percent of U.S. waters, according to the state department website.

But these protected areas are often too massive to thoroughly patrol, leading some researchers to argue that they may not actually be protecting the species they’re meant to conserve. A new study combines newly developed satellite and machine learning technology to ask this question of one key predator species: the grey reef shark. By monitoring the interactions between two top predators—sharks and humans—the authors set out to find out whether these preserves did, in fact, have teeth.

Picture a shark, and the image in your mind is likely something akin to a grey reef shark. Preying mainly on fish, cephalopods and crustaceans, these sleek gray bullets can be found in the Pacific and Indian Oceans, where they tend to roam in shallower waters near coral reefs. The species is classified as near threatened, partly because they reproduce so slowly, making their populations vulnerable to over-fishing for food or medicinal purposes. 

For the new study, a team of researchers led by Tim White at Stanford University and Doug McCauley at the University of California at Santa Barbara headed out to the U.S. Palmyra Atoll National Wildlife Refuge, a protected area three times the size of California. There are few places better for tracking down these iconic predators than this preserve, which is located in the middle of the Pacific Ocean about 1,600 kilometers south of Hawaii.

“It’s one of the sharkiest places in the world,” says White. “You get in the water, and within seconds in most cases you’ll see a shark.” He has caught and tagged up to 60 reef sharks in a single day there, catching them with handheld fishing lines and barbless hooks that don’t injure the sharks.

White’s team captured and tagged six grey reef sharks with satellite tags that could track their movements, then released them. “When the dorsal fin breaks the surface of the water, there’s a little antenna that transmits to us,” White explains. They then tagged 262 more sharks with plain numbered tags, which they stuck into their dorsal fins like earrings. These tags let them track whether any local fishermen from the nearby island nation of Kiribati caught grey reef sharks that were tagged in the refuge.

The islands of Kiribati are several hundred kilometers from the edge of the refuge, and local fishermen generally don’t stray that far from their islands because their small aluminum skiffs and wooden canoes don’t have navigational equipment or enough gasoline fuel. Of the 262 tagged reef sharks, only five (2 percent) were definitely killed by Kiribati fishermen, presumably when sharks swam outside the bounds of the refuge. They were likely killed for their fins, which fetch a high price, especially in Asian markets, says White.

To track larger fishing boats, White and McCauley used satellite data and machine learning algorithms from Global Fishing Watch. This strategy takes advantage of the fact that about 70 percent of large fishing boats emit Automatic Identification Signals (AIS) to help avoid collisions between boats. AIS can be turned off, but that puts a boat at risk of crashing.

The new technology, only available in the past year, collects all AIS signals via satellite and uses them to plot the paths of fishing boats. Then, a machine learning algorithm flags when and where the boats are actively fishing. “Facebook’s algorithms can learn to recognize individual faces after being ‘trained’ on a set of labeled photos that a person has previously tagged,” explains White. “Similarly, the fishing score algorithm was trained to recognize fishing activity from known examples of fishing activity patterns.”

Throughout 2013 and 2014, White and McCauley’s team used this technology to track fishing activity from 193 boats belonging to 12 nations in the Palmyra Atoll refuge. “Until that very recently became possible, a lot of this fishing was fully invisible to us,” says White.

The data revealed a heartening picture. The researchers found that almost all fishing boats stayed outside the protected area, while two-thirds of reef sharks stayed completely inside. Over two years of observation, they only observed the equivalent of one full day of fishing within the refuge by boats emitting AIS signals—a vanishingly small amount, considering that the refuge is far too large to physically patrol.

White and McCauley’s research also revealed another surprise: Reef sharks need a lot more space than we thought. Previously, scientists had thought that these sharks tended to spend most of their time adjacent to reefs, where their predation helps maintain these diverse ecosystems. But when the Obama administration twice expanded the Palmyra refuge in 2009 and 2014, it turned out to make a big difference from a reef shark’s point of view.

The team found that some of the reef sharks traveled huge distances into the open ocean, as much as 1,000 kilometers. One shark spent 97 percent of its time away from the reef. Every one of the six satellite-tagged sharks left the original 2001 boundaries of the refuge, showing that it was originally too small to adequately protect the sharks.

In other words, as enormous as the current Palmyra Atoll refuge is, it doesn’t protect all reef sharks that pass through its waters. “It shows us that, yes, these big marine protected areas can work, but they may not work fully without additional protection,” says Mike Heithaus, a marine ecologist at Florida International University who was not involved in the study. He adds that larger shark species like hammerheads probably roam even farther and need larger protected areas.

Tim White says his next project will look at economic and social drivers of the shark fin trade among the Kiribati fishermen, which he says is critical to understand for future conservation policy. “My vision of an ideal scenario is definitely one that works for sharks and humans,” he says. “Studies like this are tools to reach that ideal point.”