In British Columbia, removing salmon farms has not produced the clean drop in sea lice on wild juvenile salmon that many expected. That finding is now quoted worldwide as proof that farms do not drive lice on wild fish.

The evidence is real, but it is narrower than the headline. It says a lot about the Pacific, where wild salmon number in the millions and carry lice of their own. It says much less about the North Atlantic, where farmed fish outnumber wild ones by hundreds to one.

What the British Columbia data show

British Columbia emptied two of its most studied salmon-farming areas within four years. All farms left the Discovery Islands by 2022 after a federal order. In the Broughton Archipelago, farmed salmon fell from 21,645 tonnes at 17 sites in 2019 to 614 tonnes at two sites in 2024, under a 2018 agreement with three First Nations.

If farms were the main source of lice, lice on wild juveniles should have fallen and stayed down. They did not. In the Broughton, a peer-reviewed study of 2,868 juvenile chum and pink salmon from 2016 to 2024 found that prevalence peaked in 2022, at 54% on chum and 63% on pink, then dropped sharply in 2023. In 2024 it rose again to levels similar to or above those of 2016–2021, when the farms were still running.

The Discovery Islands told a similar story. A companion study of nearly 6,000 juveniles found 2024 lice levels among the highest in eight years, two years after the last farm closed. In spring 2026 about 22% of pink and 24% of chum juveniles there still carried lice.

The swing from one year to the next is larger than any change that tracks farm biomass. That is the industry’s core argument, and on its own terms it holds.

The evidence on the other side

The same coastline has also produced strong evidence that farms raise lice on wild fish. As the Discovery Islands farms were emptied between 2020 and 2022, Alexandra Morton and statistician Richard Routledge counted lice on 1,627 juvenile pink and chum salmon. They reported a 96% drop in lice per fish and found no significant link to salinity.

The industry’s own Discovery Islands series, read closely, does not show the farm-free years as the worst. Chum prevalence peaked at 26.9% in 2019 and pink at 17.1% in 2020, both farm years. In 2023, with no farms, no lice at all were found on pink salmon.

Older work points the same way. In 2007, juvenile Fraser River sockeye carried 4.83 lice per fish after passing farms, against 1.10 before them, while sockeye from the farm-free Skeena on the north coast carried 0.17. A multi-region study found lice on juvenile pink and chum significantly higher near farms in every farming region of BC, and highest in the Discovery Islands, where production was greatest. In the Broughton, a 2011 study found that fallowing farms reduced and evened out lice abundance, while active farms remained the main source.

Salmon Coast Field Station, which runs the longest independent series in the Broughton, says its data repeatedly link open-net farms to parasites on nearby wild juveniles.

Why the numbers don’t settle it

Both sides are reading real data. They disagree because the data are thinner than either side admits.

Start with who measures. The industry series comes from Mainstream Biological Consulting, contracted by Mowi, Cermaq and Grieg. Its peer-reviewed analyses are co-authored by a retired federal scientist and a University of Strathclyde professor, but the funding is not neutral. Neither are the advocacy groups behind much of the counter-evidence.

Then there is what gets counted. Most headline figures are prevalence, the share of fish with at least one louse, so one louse and fifty count the same. In 2026 the Broughton fish examined carried on average 0.67 lice each. A 2011 Broughton study put transmission about 100 times lower than during the epizootics of the early 2000s. Low-level infection can look “unchanged” in percentage terms while the damaging loads are gone.

It also matters which louse is counted. Of 340 lice found in the Broughton in 2026, only 93 were Lepeophtheirus salmonis, the salmon louse that farms amplify. The other 247 were Caligus clemensi, a generalist that also lives on non-salmonid fish. Pooling both species blurs the farm signal.

Finally, the Pacific has its own reservoir. On BC’s north coast, where there are no farms, returning adult salmon pass lice to juveniles every spring. Those natural levels were 10 to 100 times lower than on juvenile pink salmon in the Broughton when farms were active. Farms add to a natural baseline; removing them does not take the baseline away.

Other natural experiments

Outside British Columbia, most natural experiments point the other way: where farms are absent or fallow, lice on wild fish fall. The exceptions come from areas with little production, and from reviews of how regulators model the risk.

Norway offers the starkest numbers. Over two decades its farmed salmon came to dominate the louse’s hosts so completely that by 2017 they made up 99.6% of all available hosts and produced 99.1% of adult female salmon lice. The national salmon fjords, where farming is restricted, work as a natural control. An evaluation by NINA and the Institute of Marine Research found that lice on wild sea trout tracked larval output from farms up to 30 km away, and that the largest protected fjords had the least lice. In 2015, researchers sampled two reference areas with little or no farming, Porsanger and Risør, and found no wild fish carrying harmful lice loads. In outer Hardanger, one of the most heavily farmed areas on the coast, 80% of wild salmon smolts did.

Not everyone reads the Norwegian data the same way. A literature review in Reviews in Aquaculture argues that the traffic-light system, which limits growth where lice are judged to harm wild salmon, probably overestimates how much farm lice affect wild fish.

Scotland’s clearest evidence comes from a single sea loch. At Loch Shieldaig, one farm ran through two production cycles with fallow periods between 2002 and 2006. Lice larvae in the water were scarce while the farm lay fallow and rose and fell with the number of egg-bearing lice on the farmed fish. Sea trout returning early to the River Shieldaig between 1999 and 2009 carried significantly more lice in the second year of the farm cycle than in the first.

The largest experiment treated the fish rather than the farms. In 24 trials in Ireland and Norway, some 280,000 salmon smolts were tagged and released, some protected against lice and some left untreated. Lice explained on average 39% of marine mortality. An Irish analysis of overlapping data put the figure nearer 1%, and the two camps still argue over the method.

Iceland’s farming is younger and smaller, and its results are mixed. In Arnarfjörður, lice levels on wild fish were low overall but higher at the two sampling sites next to farms. In Patreksfjörður in 2019, monitoring of wild sea trout and Arctic charr found no consistent link between lice counts in the cages and on the wild fish.

Chile turns the Pacific case on its head. It has no native salmonids, but its own louse, Caligus rogercresseyi, lives naturally on native wild fish. Salmon farming did not bring the parasite to Chile; dense farms amplified one that was already there.

The farm signal is clearest where farmed fish dominate the host population, as in Norway. It is weakest where production is low or wild hosts are plentiful, as in early Icelandic monitoring and on the Pacific coast.

What the evidence supports

Three things are well supported. Sea lice are a natural parasite and will not vanish when farms go. Year-to-year swings, driven by temperature, salinity and wild host numbers, are large enough to swamp a farm signal in short series. And prevalence alone is too blunt to show how much farms add.

What the British Columbia data do not support is the leap to “farms don’t matter anywhere”. In Norway farmed fish are almost all the lice hosts, and where farming is absent, harmful loads on wild fish largely disappear. The North Atlantic has no Pacific-scale wild reservoir to hide the farm signal.

The useful lesson from Canada is about method. Count abundance, not just prevalence. Separate louse species. Run long, independent series with farm-free reference sites, before and after any change in farming. Without that, both sides will keep quoting the same data back at each other.