Counting Fish: Making Sealife Sustainable for Future Generations

Around 18,000 species live in the world’s oceans and researchers estimate total fish numbers at 3.5 trillion.

But managing global fish populations and stopping overfishing is an urgent necessity. Nations often break the rules and fish outside their waters after depleting their own stocks. Global Fishing Watch estimates that 34 percent of the world’s major commercial fish species are overfished and researchers from Stanford University found that fish stocks could collapse entirely by 2048.

One Fish, Two Fish...

This stark warning has led the United Nations Fishing and Agricultural Organization (FAO) to gather as much vital information about fish species around the globe as it can. The FAO looks closely at how local fish populations are managed, records catches internationally and monitors fishing fleets, refrigeration, and supply vessels.

Despite this activity by the FAO and organizations like the National Oceanic and Atmospheric Administration (NOAA) in US waters, counting fish to maintain sustainable populations is not easy. Traditional techniques rely on trawling for samples, then donning rubber gloves and physically recording species by size, weight, and sex.

Fishy Data

NOAA’s Fisheries Scientific Computer System brings together automated systems to cut down the time taken to gather fish data, using onboard sensors and then recording it in real time along with weather and oceanographic conditions.

Now researchers are trying out an imaging system linked to machine learning models and artificial intelligence (AI) to vastly increase the count. CamTrawl counts fish as they enter trawl nets and they are then released without harm. During one summer season, CamTrawl collected up to three million still images that were then automatically identified and added up using species identification systems.

This technology-led era has also seen the NOAA adopt controlled and autonomous underwater vehicles (AUVs) to help in counting fish. SeaBED is a torpedo-shaped AUV that coasts a few meters above the sea floor taking pictures of fish and their habitats. Thousands of pictures are then blended into photomosaics that provide a unique insight into how fish live at depth.

There are a range of vehicles collecting vital sea life data wherever it may be. Wind and solar powered surface vessels–or ‘saildrones’ – automatically gather sea and atmospheric data, while conducting a survey of one of the US most valuable domestic species: the Atlantic pollock. Aerial drones are launched from seagoing ships to count rapid migratory fish like the bluefin tuna.

Added to this fleet is HabCam, a large ship-towed seabed imaging system that can take six images every second to provide a real-time picture of the ocean floor. By linking the millions of images collected by HabCam to neural networks used in AI, analysts are teaching computers to identify species and speed up the census.

In the future, AUV swarms may take up the bulk of the deep, dangerous and time-consuming work that is currently done by single unmanned vehicles. Smaller, less expensive underwater robots will work in sync to scan the seas. “This will give us better spatial and temporal coverage in the areas we’re trying to study, and provide us with much richer and robust data sets in far less time,” said Erin Fischell, an AUV developer at the Woods Hole Oceanographic Institute.

The Future of Fish

Ongoing FAO efforts to keep fisheries and sea life populations viable include an international code of responsible conduct alongside measures to tackle illegal fishing and activities like transhipment–where vessels offload their catch at sea onto refrigeration vessels to avoid detection.

Globally, the FAO is looking to use each fishing vessel’s automatic identification system alongside AI to allow authorities to track and analyze catches wherever they happen. Their aim is simple: to keep fish numbers healthy for generations to come and stop rogue operators from causing lasting damage.