Artificial intelligence shows unprecedented detail in global fishing activities

Researchers are learning more than ever before about the effects humans are having on global fish stocks.

It’s all thanks to a website — funded in part by actor Leonardo DiCaprio’s foundation — that tracks ships and uses a type of artificial intelligence to figure out incredible detail in worldwide fishing patterns.

Kristina Boerder, a PhD student in marine biology at Dalhousie University, is one of the researchers working with Global Fishing Watch and a co-author on a study published this week in the journal Science.

She said humans have been fishing for 42,000 years but we’ve been “rather in the dark” about where and how much fishing activity is happening. 

“This is really a problem because this is a resource that is not infinite,” Boerder told the CBC’s Mainstreet. “We need a better picture of what is going on globally on the oceans in order to … understand what’s happening.”

Launched in 2016, the website allows users to view a world map with tens of thousands of fishing vessels moving in “near real time,” which is 72 hours from the present time. The data are so detailed that individual vessels can be tracked hourly.

Vessels are tracked by on-board transponders known as Automatic Identification Systems or AIS that are picked up by satellites or ground-based stations. AIS was developed initially as a way to avoid ship collisions. 

Using only vessel movements, the website’s machine learning algorithm — a type of artificial intelligence — was able to identify more than 70,000 commercial fishing vessels.

Machine learning is a branch of computer science in which software uses a training set of data to teach itself to interpret large amounts of data — for example, how Google is able to tell with some degree of accuracy what is spam and what is email you want to receive. 

Global Fishing Watch’s sophisticated software, also called a neural network, can extrapolate the type of fishing the vessels are engaged in, when and where they are fishing and even the size of the engine powering the vessel.

For example, Boerder said the algorithm can distinguish purse seiners — which drive in a loop with their nets around schools of fish — from other types of fishing.