Researchers are using satellite imagery and artificial intelligence to map human activity at sea more accurately than ever before. The work uncovered a host of previously unknown industrial activities, from questionable fishing operations to the explosion of offshore energy development.
The maps are published today in the journal Nature. Research led by Global Fishing Watch, a non-profit organization backed by Google, shows that up to three-quarters of the world's industrial fishing vessels are not publicly tracked. Up to 30% of shipping and energy vessels are also not publicly tracked.
Researchers say these blind spots could hinder conservation efforts around the world. To better protect the world's oceans and fisheries, policymakers need a more accurate understanding of where people are exploiting resources offshore.
According to the Kunming-Montreal Global Biodiversity Framework adopted last year, almost all countries on the planet have agreed to a common goal of protecting 30% of the earth’s land and water areas by 2030. So, the question is which 30% should we protect? "Without this map, you can't discuss where the fishing activities are, where the oil platforms are," said David Kroodsma, co-author of the Nature paper and director of research and innovation at Global Fisheries Watch.
Until now, Global Fisheries Watch and other organizations have relied primarily on automatic maritime identification systems (AIS) to understand what is happening at sea. The system can track ships carrying radio transmitter boxes, and the data has been used in the past to document overfishing and forced labor on board ships. Even so, the system has significant limitations. Requirements for carrying AIS vary by country and vessel type. And when someone wants to avoid detection or cruise somewhere with a weak signal, they can easily turn off the box.
To fill in the gaps, Kroodsma and his colleagues analyzed 2,000 terabytes of images provided by the European Space Agency's Sentinel-1 satellite constellation. Sentinel-1 uses advanced radar instruments to observe the Earth's surface instead of taking traditional optical images (like taking pictures with a camera). Radar can penetrate clouds and "see" in the dark - it can detect offshore activity that AIS misses.
Because 2,000TB of data is quite large, the researchers developed three deep learning models to classify each detected vessel, estimate its size, and classify different types of offshore infrastructure. They monitored about 15% of the world's oceans, where 75% of industrial activity takes place, looking at the movements of ships and the development of fixed offshore structures such as oil rigs and wind turbines between 2017 and 2021.
Fishing activity declined at the start of the COVID-19 pandemic in 2020, but they found intensive vessel traffic in areas of the public tracking system that had "previously had little vessel activity," particularly in South and Southeast Asia and along the northern and western coasts of Africa.
The data also shows a boom in offshore energy development. By the end of 2020, there will be more wind turbines than petroleum structures. By the following year, turbines accounted for 48% of all marine infrastructure, while oil structures accounted for 38%.
Almost all offshore wind development is taking place off the coasts of Northern Europe and China. In the U.S. Northeast, clean energy opponents have tried to falsely link whale deaths to upcoming offshore wind development, despite evidence that ship strikes are the problem.
There are more ships surrounding oil structures than there are wind turbines. Tankers are sometimes used to transport oil to shore as an alternative to pipelines. Over the five years studied, the number of oil structures grew by 16%. By 2021, offshore oil development will cause five times more global ship traffic than wind turbines. "The actual amount of ship traffic caused by wind turbines around the world is minuscule compared to other traffic," Klozma said.
On the world map, blue circles indicate the location of offshore wind infrastructure, mainly in Europe and China. The yellow circle indicates the location of the oil platform. Purple circles indicate the location of "other structures".
When asked whether such research would be possible without artificial intelligence, "the short answer is no, I don't think so," said Fernando Paolo, the study's lead author and a machine learning engineer at Global Fisheries Watch. "Deep learning is good at discovering patterns in massive amounts of data."
Another article published today in Nature reviews Paolo and Kroodsma's research, saying that as open source software for processing global satellite imagery, the new machine learning tool "democratizes access to data and tools, allowing researchers, analysts and policymakers in low-income countries to take advantage of tracking technology at low cost." The article was written by Konstantin Klemmer, a postdoctoral researcher at Microsoft, and Esther Rolf, an assistant professor at the University of Colorado Boulder.
These technological advances provide an important basis for documenting rapid changes in maritime activity at a critical time when countries are trying to halt climate change and protect biodiversity before it is too late. "The reason this matters is because [the sea] is getting more and more crowded, more and more used, and all of a sudden you have to decide how to manage this huge global commons. It can't be the Wild West. That's the way it has been historically," Crozma said.