Immense but not immutable: a decline in deep-pelagic life in the Gulf of Mexico

The open water of the Gulf of Mexico below 200 metres holds 90.4% of the Gulf's volume.^1,2 Its fishes, shrimps, krill and small squid rise from as deep as 1,200 metres to near the surface each night to feed, which makes them the link between the plankton and the tunas, dolphins, whales and seabirds that eat them.¹ Tracey Sutton of Nova Southeastern University and fifteen co-authors counted that community from the surface to 1,500 metres for thirteen years, and found it collapsed and still down.^1,2

The baseline is a nine-month campaign of continuous sampling in 2011, starting three months after the Deepwater Horizon well was sealed, on a grid of 46 stations in the northern and eastern Gulf.¹ The DEEPEND programme returned nine times between 2015 and 2023 with the same net, the same single unit of it in fact, fished the same way, and the same people sorting the catch.¹ Ships' echosounders ran alongside the nets, and the team tracked genetic diversity in three species of lanternfish.^1,2 Older counts from the 1980s and 1990s at one station inside the grid sit close to the 2011 numbers, so 2011 was the normal state, not an unusually rich year.¹

Within three and a half years, depending on the group, abundance fell 60 to 85%.¹ Fishes fell 63.5%, the larger crustaceans 62.0% and small cephalopods 59.7%.^1,2 Krill fell 72.8%, lanternfishes 62.9% and the bristlemouths of the genus Cyclothone, which may be the most numerous vertebrates on Earth, 69.0%.^1,2 Some of the crustacean and squid losses show up as early as August 2011.¹ The acoustic surveys fell too, by 40% near the surface at night.¹ The authors know of nothing comparable: the fish decline was larger than any in sixty years of the California Current record, the only other long series of its kind.¹ The dense patches of prey that predators need also went: in 134 full-depth samples since the decline, the biggest fish catch beat the smallest 2011 catch once.^1,2 Two of the three lanternfish species lost genetic diversity, one by up to 13% before a partial rebound.¹

The paper then works through the causes. The method did not change.¹ Temperature at depth held steady and the water stayed well oxygenated.^1,2 The plankton these animals eat was above average in 2013, in the middle of the decline.¹ What remains is the spill and the response to it: 4.0 to 4.9 million barrels of oil over 87 days, and 770,000 gallons of the dispersant Corexit 9500A injected at the wellhead, 1,500 metres down.^1,2 Other work has shown the oil's invisible footprint covering nearly the eastern half of the Gulf, far beyond what the satellites saw, and oil that settled on the sea floor being stirred back up for years.¹ Oil compounds turned up in the eggs of deep-sea fishes six years later at levels known to harm fish.^1,2 The authors call the spill the most parsimonious explanation, not a proven one.^1,2

Why no recovery? The authors point to the Allee effect: below some density, animals find mates less often, lose genetic diversity, and fail to rebuild, which matters most where animals are sparse and find each other by smell in the dark.^1,2 Lags are known from other spills.¹ Prince William Sound's herring collapsed four years after the Exxon Valdez and have not come back.¹

The losses run upward. Other surveys have found the Gulf's deep-diving whales down since the spill, 83.4% for Gervais' beaked whale, 49% for the goose-beaked whale and 31% for the sperm whale, and the paper notes these declines are larger than the models of the spill's damage predicted.¹ Tagged sperm whales now range widely instead of foraging in their usual patterns, endangered Rice's whales are seen feeding near the bottom and often failing, and stranded whales are described as emaciated.¹ Fewer migrating animals also means less carbon carried into the deep each night.^1,2

For a model of a sea, three things stand out. The damage was below the layer that satellites watch. The models of the damage underestimated it. And every number here is public: the counts, sensor logs, echosounder files and genetic data are in GRIIDC under 27 DOIs,¹ enough to test a food-web or recovery model against what happened. The use case on oil spills sets out what such a model would hold.