Industries / Food, farming and forestry / Fishing and aquaculture
Fish farming
Fish farming is raising fish, shrimp, shellfish and other water animals in ponds, tanks and sea pens until harvest.
7
Milestones achieved, from About 6,600 years ago to 2020.
3
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
In depth
The same ground is charted in more depth, as an industry of its own.
The road
AchievedWatch forOpen
Achieved
About 6,600 years ago
Eel channels at Budj Bim
Gunditjmara people in western Victoria, Australia, built stone channels and pools at Lake Condah to catch eels and hold them alive. Carbon dating of charcoal from one trap gave an age of about 6,600 years. UNESCO added the Budj Bim Cultural Landscape to the World Heritage List in July 2019.
Lessonmanaging water to hold fish is far older than written records of fish farming.
Credit
Gunditjmara people
1763
Trout eggs fertilised by hand
Stephan Ludwig Jacobi (1711–1784) in Germany squeezed ripe eggs and sperm out of trout (called stripping), mixed them, and hatched the eggs in a wooden trough with a gravel floor. Fish hatcheries grew from this method. He first reported it anonymously in 1763 and under his own name in 1765; an FAO timeline dates it to 1767. It was largely forgotten until Remy and Géhin rediscovered it in France in 1842.
Lessona working method can sit unused for decades if nobody spreads it.
Credit
Stephan Ludwig Jacobi
1933
Shrimp larvae raised in tanks
In July 1933 Motosaku Fujinaga (published as Hudinaga) got kuruma shrimp to spawn and hatch in a culture tank in Japan. He went on to raise shrimp larvae in large indoor tanks, feeding them newly hatched brine shrimp, which let farms stock from hatcheries instead of the wild. He and colleagues opened a commercial shrimp farm in 1963.
Credit
Motosaku Fujinaga, whom the Global Seafood Alliance calls the "father" of shrimp farming
1970
Salmon grown in sea pens
On 28 May 1970, on the island of Hitra, Norway, brothers Ove and Sivert Grøntvedt put young salmon in a floating open net pen. The Norwegian Seafood Council calls these fish "the very first successful generation of farmed Atlantic salmon"; they were harvested in 1971. The trade site iLaks disputes the "first": it reports the Vik brothers kept salmon and trout in floating cages in 1959.
Credit
Ove and Sivert Grøntvedt
2007
A virus empties Chile's pens
Infectious salmon anaemia (ISA), a viral disease of farmed salmon, was confirmed at a Marine Harvest Chile site in central Chiloé in 2007 and spread through three regions. An industry analyst estimated it cut production by about 50 percent from mid-2007. Chile answered with farm zones, strict controls on imported eggs and biosecurity rules (rules to stop germs moving between sites).
Lessonfarms that share water share disease, so Chile now manages them by zone.
Credit
reported by Adolfo Alvial for the Global Seafood Alliance
2015
Engineered salmon approved, then shelved
On 19 November 2015 the US Food and Drug Administration (FDA) approved AquaBounty's AquAdvantage salmon, an Atlantic salmon carrying a growth-hormone gene from Chinook salmon so that it grows faster. The FDA called it its "first approval for a genetically engineered animal intended for food." AquaBounty sold its Indiana farm in 2024 and closed its last Canadian farms that December.
Lessonapproval did not make the fish pay.
Credit
AquaBounty Technologies; US FDA
2020
Cameras allowed to count lice
Sea lice are small parasites that feed on salmon skin. Norwegian farms must count them every week, which meant netting fish out of the pen and checking them by hand. In November 2020 the Norwegian Food Safety Authority (Mattilsynet) let Kvarøy Fiskeoppdrett and Seløy Sjøfarm rely on Aquabyte's underwater cameras alone at one site each. By 2024 about 30 businesses held such permissions, using 7 different systems. In December 2025 Mattilsynet said no counting method is formally approved and that not all systems in use count well enough.
Lessona camera's count is trusted only as far as it can be checked against a standard.
Credit
Aquabyte; Kvarøy Fiskeoppdrett; Seløy Sjøfarm; Mattilsynet
Watch for
Lice counts straight to regulator
Mattilsynet built and tested a working prototype that sends daily lice counts from farm cameras straight into its own systems through an API (a standard way for one computer system to send data to another). The test ran from week 46 to week 51 of 2025 with the technology firm OptoScale and the assurance firm DNV, and Mattilsynet concluded its systems can receive daily automatic counts safely.
Achieved whendirect reporting runs in everyday use and the rules let certified camera counts replace manual counts.
Claimed by
Mattilsynet, February 2026
Salmon farmed on a ship
Qingdao Guoxin Group, a Chinese state-owned company, says its 244.9-metre vessel Guoxin No. 1 2-2 sent its first 3,000 salmon (12 tonnes) to market after about six months on board. The ship has 15 tanks holding nearly 100,000 cubic metres of seawater pumped from 30 to 50 metres deep, and it can move along the coast to follow suitable water temperatures.
Achieved whenan outside party reports several full growing cycles with survival and harvest figures.
Claimed by
Qingdao Guoxin Group, reported by Fish Farming Expert, May 2026
Cameras set the feed rate
Scottish Sea Farms, feed maker BioMar and technology company Tidal run an AI camera system at the Wyre salmon farm in Orkney that watches how hungry the fish are and changes the feed rate as they eat. It is part of the site's daily feeding; no results have been published.
Achieved whenfeed conversion ratio (kilograms of feed per kilogram of growth), growth and welfare results against normal feeding are published and checked by someone outside the three firms.
Claimed by
Scottish Sea Farms, BioMar and Tidal, March 2026
Open
01
Open test for lice cameras
a written test procedure plus free scoring software. A farm or regulator feeds in a lice camera's counts and careful manual counts of the same pens, and the software reports, for each lice stage, the bias (how far the camera reads high or low), the correlation, and whether the camera passes stated limits.
The briefCredit · unclaimed02
Cause-coded register of die-offs
a public table of mass-mortality events (many farmed fish dying within days) in sea pens and land-based tanks from 2012 to 2025. Each row gives date, place, farm type, fish or tonnes lost, cause in fixed categories (disease, algae bloom, jellyfish, low oxygen, toxic gas such as hydrogen sulfide, equipment failure, handling during treatment, storm), and a link to the source.
The briefCredit · unclaimed
Open 01 · the brief
Open test for lice cameras
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
a written test procedure plus free scoring software. A farm or regulator feeds in a lice camera's counts and careful manual counts of the same pens, and the software reports, for each lice stage, the bias (how far the camera reads high or low), the correlation, and whether the camera passes stated limits.
Why it is still open
Mattilsynet said in December 2025 that no counting method is approved in Norway and that validation (proving a camera counts correctly) had been taken out of the draft Norwegian standard. Norway's Institute of Marine Research wrote in 2024 that "a validated standard is needed" to benchmark camera counts. In September 2026 a senior figure at equipment maker AKVA group was still calling for standard validation procedures.
Built with what exists
Python with pandas, SciPy and statsmodels; the Institute of Marine Research's standard manual counting procedure and its finding that 30 fish per pen is a workable sample; the BarentsWatch fish health API, which gives weekly manual lice counts for every Norwegian site under an open licence, to set realistic lice levels for simulated tests.
First step
read the Institute of Marine Research's 2024 reports on manual and automatic counting, list every measure and limit they use, and register with BarentsWatch to download one year of weekly counts for 20 sites.
Done looks like
a public repository with the procedure and scorer, plus a published table showing how often the scorer catches a planted 20 percent under-count at 10, 30 and 60 fish per pen across 1,000 simulated pens; the package filed as a comment to Standard Norge's committee on lice counting (SN/K 613), with any reply logged.
Open 02 · the brief
Cause-coded register of die-offs
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
a public table of mass-mortality events (many farmed fish dying within days) in sea pens and land-based tanks from 2012 to 2025. Each row gives date, place, farm type, fish or tonnes lost, cause in fixed categories (disease, algae bloom, jellyfish, low oxygen, toxic gas such as hydrogen sulfide, equipment failure, handling during treatment, storm), and a link to the source.
Why it is still open
the first global count of these events (Singh, Sajid and Mather, 2024) found reporting inconsistent between countries, stopped collecting in 2022, and called for standard reporting. A 2024 review by Sajid and colleagues says the causes "are not well understood". Scotland's Marine Directorate says its own mortality datasets "cannot be directly compared with one another".
Built with what exists
Singh and colleagues' published event list as a starting point; Scotland's monthly fish health inspection spreadsheets; the Canadian government's public list of British Columbia mortality events; company reports; a spreadsheet; and a low-cost AI model to pull dates, numbers and stated causes out of long reports, the repetitive extraction work Meta aims its Muse Ember model at (Note: "The jobs agents repeat").
First step
download the supplementary data from Singh and colleagues, write a one-page cause rubric, and code every 2021 event from one country.
Done looks like
at least 300 events from at least 4 countries, every row with a working source link; the coding rules published; a second person re-codes a random 20 percent without seeing the first codes, and the agreement score (Cohen's kappa) is published.