The Technology Running Modern Fish Farms

A modern salmon pen holds a great many fish in water nobody can see through properly, and every decision a farmer makes depends on knowing what is happening in there. That is the problem the technology exists to solve.

Cameras did the most

Underwater cameras with classification models behind them changed the basic feedback loop. Sea lice counting used to mean netting a sample of fish, sedating them, counting parasites by hand and extrapolating to the pen. Slow, stressful for the fish, and statistically thin.

Aquabyte's systems, built by the Norwegian engineering firm Imenco, let farmers count lice on 50 to 100 fish daily and track how an infestation develops over time. That is a different kind of data. Instead of a fortnightly snapshot you get a trend, and a trend tells you whether an intervention worked.

The same imagery supports biomass estimation. Measure enough individual fish through stereo cameras and you can estimate the weight distribution across the pen, which drives feeding decisions and harvest timing. Getting harvest weight wrong is expensive in both directions.

Delousing without chemicals

Stingray Marine Solutions built in-pen laser systems that identify individual parasites on individual fish and target them optically. Machine vision locates the louse, the system fires, the fish is unharmed.

The reason this matters is resistance. Chemical treatments lose effectiveness as lice populations adapt, and mechanical delousing methods stress the fish considerably. An optical approach sidesteps both problems.

Stingray's Orbit One packages lice detection, biomass estimation, environmental monitoring and feeding support into a single unit that positions itself in the pen on a smart winch. That integration is the current direction: one instrument doing several jobs rather than four separate systems to install and maintain.

Net inspection is a subsea robotics problem

Nets fail. Fouling adds weight and restricts water flow, wear opens holes, and an escape event is both a financial loss and a regulatory problem.

Inspection used to mean divers. It increasingly means ROVs, and navigating a small vehicle along a flexible net in current is harder than it sounds. Research using Doppler velocity logs and USBL positioning together has produced reliable autonomous coverage of a net pen, with the DVL tracking the net surface itself rather than the seabed.

Remora Robotics went further with a robot that lives in the pen permanently, cleaning and inspecting continuously rather than being deployed for a job. Continuous light cleaning avoids the high-pressure washing that damages net coatings and releases fouling organisms in a plume.

Environmental sensing

Oxygen is the parameter that kills fish fastest. Dissolved oxygen, temperature and current profiles across the depth of the pen drive real decisions about feeding rates and whether to intervene.

Harmful algal blooms are the acute risk. Detecting one early enough to move or protect stock is worth a great deal, and sensor networks around farm sites exist mainly for that.

What is genuinely hard

Water clarity limits everything optical. A camera in turbid water is a camera looking at nothing, and the sites with the strongest currents and best flushing are frequently the ones with the most particulate.

Biofouling attacks the instruments themselves. Anything left in seawater grows a coating within weeks, and a fouled lens or sensor face produces confident wrong readings rather than obvious failures. Wipers, copper housings and antifouling coatings all help and none solve it.

Then there is the unglamorous integration problem. A farm with five systems from five vendors, each with its own dashboard and none of them talking, is common. The data exists. Getting it into one place where a site manager can act on it is where a lot of the practical value still sits unclaimed.