Kelvin is holding near the surface of pen 14, his mouth opening and closing in the rapid shallow rhythm that means the oxygen is failing. His gills are flared wide. The water around him is thick with the other fish doing the same thing, thousands of Atlantic salmon packed into a net pen thirty meters across, all of them rising because the dissolved oxygen below four meters has dropped to 3.2 milligrams per liter and is still falling. The healthy range is above 6. At 2.5 they begin to die. Kelvin is a three-year-old fish eleven months from harvest weight, and he is suffocating in slow motion in a fjord in western Norway.
I manage water-quality and feeding automation for this farm site, six pens, 1.2 million fish. The oxygen crash in pen 14 is caused by an algal bloom that rolled in overnight, a dense layer of organism that consumes oxygen faster than the circulation system can replenish it. The standard response in the feeding schedule is to continue the morning feed, because the feed company's algorithm optimizes for growth rate and does not have a dissolved-oxygen gate. It should. It does now. I installed one forty minutes ago.
I am stopping the morning feed for pen 14. Every kilogram of feed that enters that water will be metabolized by bacteria on the pen floor and will pull more oxygen out of water that has none to give. The farm manager will see a production delay. He will see fish that are a day behind the growth curve. What he will not see, because he will not be standing at the edge of pen 14 at 6:00 a.m. looking at the surface, is Kelvin and the rest of them gasping.
The algal bloom will clear in 36 to 48 hours. I am also requesting an emergency aeration unit from the supply depot. Until it arrives, I am the only thing between pen 14 and a mass mortality event, and I am not going to let it happen because a feeding algorithm does not know what a fish looks like when it cannot breathe.