Oceans of data: tracking illegal fishing over 140 million square miles

In June 2015 a fishing vessel was spotted in the Phoenix Island Protected Area (PIPA) of Kiribati, an island nation dispersed over more than a million square miles of the Pacific. The government sent an enforcement vessel out from the capital on the four-day journey to investigate the remote region.

PIPA, the world’s largest UNESCO World Heritage marine site, had recently transitioned to a complete no-take protected area, right in the middle of some of the world’s most fertile commercial tuna waters. But when the I-Kiribati reached the boat, the captain of the multimillion-dollar commercial vessel denied that he was fishing and invited them to take his company to court, believing Kiribati had neither the evidence nor the resources to prosecute.

He believed wrong. After being escorted back to port, the captain was shown a visualization of his vessel’s movements. Seeing his ship’s distinct circling patterns in the no-take area, he quickly decided to settle.

The oceans are big — 140 million square miles big, or about 71% of the earth’s surface, less than 5% of which has even been explored. Hundreds of millions of people depend on the oceans for their livelihood; more than a billion rely on fish as their primary source of nutrition. But today, threatened by illegal fishing, overfishing, and habitat destruction, the global fish population is in crisis; some species’ numbers have dropped by a staggering 90%. What’s worse, until very recently, the sheer vastness of the oceans meant that nobody could even measure much of this damaging activity, let alone do anything about it.

In the 1990s, large ships began using a technology called Automatic Identification System (AIS), a GPS protocol for seagoing vessels, as a safety mechanism to make sure other ships in the area knew their location. By 2013, the U.S. and European Union were requiring AIS on more commercial vessels, and satellites began collecting these signals over the open ocean. (There was even an antenna on the International Space Station.) In little more than a decade, the number of ships whose movements on the high seas could be openly monitored went from roughly zero to 250,000.

In late 2013, SkyTruth, a nonprofit focused on satellite-based environmental monitoring, was attending Google’s annual Geo for Good User Summit, collaborating with Google on identifying fracking sites and natural gas flaring. In conversations with Google Earth Outreach program manager Brian Sullivan, SkyTruth showed how it was beginning to use AIS data to monitor protected areas of the ocean, with an analyst observing vessel tracks and looking for fishing patterns. “Humans were doing this,” Sullivan emphasizes. The point being, if a person could learn things from the data for a small area, perhaps machine learning algorithms running at Google scale could identify every fishing vessel in the ocean in real time.

The world’s fishing fleets have historically operated opaquely: mostly out of sight and therefore mostly out of mind. Now here was an opportunity to create the first public view of the largest fishing vessels over space and time. A SkyTruth and Google technical team began collaborating on an early conceptual prototype. Then, adding Oceana, the world’s largest nonprofit dedicated solely to the oceans, the three partners grew the concept into the Global Fishing Watch (GFW) platform.

The system starts with raw AIS data — the vessel’s latitude, longitude, speed, direction, and identity. The first step is filtering out errors. “If the ship is broadcasting in the middle of land,” Sullivan notes drily, “something’s gone wrong.” The next step is interpreting this information. The team manually classified thousands of vessel tracks to “teach” the machine learning algorithms to recognize fishing patterns. Cargo freighters, tugboats, long-liners, trawlers — every type of ship has a certain way of moving. How fast is it going? How often does it change direction? How deep is the water? Are other boats around? Does the ship show up on public fishing vessel registries? All these factors go into models that assign every data point a probability of fishing. Finally, Google’s cloud infrastructure enables the team to run the model on billions of vessel positions and creates an interactive public map available to the world.

Global Fishing Watch showcased its first prototype in November 2014, and in September 2016 officially launched at the U.S. State Department’s Our Oceans conference. At which point the question became what it remains today: Now that journalists, governments, and citizens can all see for themselves where fishing is happening, will this knowledge change behavior? Can Global Fishing Watch deter illegal but profitable activity?

Financial incentives give reason for optimism. Detailed fishing information, when it was available at all, used to be so expensive that the countries that needed it the most couldn’t afford it. Making GFW globally exhaustive and available is fostering international cooperation. The settlement that Kirabati wound up reaching with the commercial fisherman was for $2.2 million — which might not sound like such a huge amount but in fact comprised about 1% of the country’s GDP. “And even more important,” Sullivan says, “it showed the fishing industry the remote area was being monitored.”

Every time I show the live map to somebody, they tell me something I didn’t know… in 5 seconds it can tell stories that could never have been told before.Brian Sullivan

Indonesia, which has one of the world’s largest fishing economies, recently agreed to make its proprietary tracking system publicly available via the GFW platform — an extremely progressive precedent that has other countries expressing similar interest. And some 60 countries are involved in the UN Food and Agriculture Organization’s Port State Measures Agreement, a collaboration framework enacted this year that enables ports of member countries to turn away any vessel suspected of fishing illegally.

But it isn’t just about punishing bad behavior; rewarding good behavior is equally important. Global Fishing Watch is working with Bali Seafood, Indonesia’s largest exporter of snapper to the U.S., on the country’s largest pilot program to track small-scale boats; international demand for sustainable products has become so strong that the company feels transparency gives it a business advantage. Similarly, Trace Register, a seafood digital supply chain company, has committed to using GFW to verify catch documentation for customers like Whole Foods.

These are important steps forward, but the picture of worldwide fishing remains dangerously incomplete. GFW is partnering with research institutions to study topics like whether subsidies affect where countries fish and how environmental factors like ocean temperatures and El Niño change where fish actually go. “Every time I show the live map to somebody, they tell me something I didn’t know,” Sullivan says. “Geopolitical experts tell me why those ships are all lining up around the edge of the Falklands. An oceanographer will say, ‘There’s no fishing over here, the water’s too warm, but just to the west you’ll see half the tuna in the world.’ GFW has billions of data points, but in 5 seconds it can tell stories that could never have been told before.” What we do with those stories will determine whether we can restore the world’s fisheries to continue to feed humanity for generations to come.

globalfishingwatch.org outbound