Under Antarctica’s winter ice, three in four temperature profiles of the ocean were taken by seals

#The lid of ice

Every winter, sea ice spreads around Antarctica. In September, at the end of the southern winter, the area with at least 15% ice (the sea-ice extent) averages 18.4 million square kilometres (1979–2025), nearly 1.9 times the area of the United States.

On the page, the white lid is the September ice averaged over those years; in winters with more ice, some profiles fall beyond its edge. Below it is the sea floor, with depths stretched 110 times so the top 2,000 metres of water can be seen.

The first people to spend a winter locked in this ice were the crew of the Belgica, in 1898. Among them was the Romanian biologist Emil Racoviță, who went on to found biospeleology, the study of life in caves.

#Ships rarely come in winter

Each thread under the ice is one profile: water temperature measured from the surface down, coloured from blue (water at its freezing point) to orange (warmer water, above 1 °C). Only the ships are shown now: 1,073 profiles in 47 winters.

An icebreaker can go in during winter, but rarely does. Ships brought more than 100 under-ice profiles in a single winter only five times (1986, 1989, 1992, 1997 and 2002). Since 2003 they have brought 118 in all.

#Floats cannot see the sky

Argo floats sink on their own to 2,000 metres and rise to the surface to send their data by satellite. Under the ice they cannot come up: they measure, store the data and send it when they reach open water. They brought 18,607 profiles.

Under the ice a float gets no satellite fix. The positions of many profiles are estimated later, usually by interpolating between the last fix before the ice and the first one after; that is why some rows of threads lie in straight lines.

#Then the seals arrive

In this selection, the first seal profiles appear in 2004. The tags went on Weddell seals, elephant seals and a few crabeater seals. In the programmes described by Roquet and colleagues, the tag is glued to the fur on the head and falls off at the moult; every few hours it keeps the profile of the deepest dive and sends it by satellite when the seal comes up to breathe.

Seals breathe through cracks and openings in the ice, so in winter they can measure places the other instruments rarely reach. They brought 61,648 of the 81,328 profiles: 75.8%.

“Under the ice” here means that the 25-kilometre cell where the profile falls had at least 15% ice on average that month. Requiring at least 50%, seals have 74.5%; at 80% or more, still 63.5%.

#Where the sea is shallow, almost only seals

Near the continent the sea floor rises to a few hundred metres on the continental shelf; on the map, the lit line is the 1,000-metre depth. Standard floats dive to 2,000 metres and rarely work in water this shallow, and ships rarely come in winter. Where the floor is shallower than 1,000 metres (from ETOPO, on a 14-kilometre grid), 41,849 of the 44,212 under-ice profiles are from seals: 94.7%.

This count was made after the main test and is exploratory, not a tested prediction. On the map, threads over deep water are dimmed.

#One Weddell seal’s winter

The seal with tag ct27-W3-07 sent 1,294 profiles between 13 February 2007 and 6 October 2007 in the southern Weddell Sea, between 74.1° and 76.0° south. Below, each column is one profile: date from left to right, depth downwards; each column goes as deep as the seal dived, and the colour shows salinity.

Between 10 and 60 metres (depth measured as pressure, in decibars), a profile’s mean temperature in June–September was between −1.93 and −1.68 °C, with a median of −1.86 °C, close to the freezing point of sea water. The monthly median of the same profile means for salinity rose from 34.03 in April to 34.55 in July.

When sea water freezes, the salt stays in the water below, which becomes heavier. On the shelf of the southern Weddell Sea this cold, salty water gathers and then flows down the slope to the deep, where it feeds Antarctic Bottom Water, one of the coldest deep-water masses in the world’s oceans. The seal also moved west during the winter, so part of the rise may come from where it went.

#Who measured, year by year

Up to 2003, 983 profiles in total were gathered under the winter ice. Seal profiles peaked at 6,996 in 2011.

Float profile counts generally increased, with year-to-year variation, and passed the seals in 2016, 2018, 2021, 2024 and 2025. In 2025 there were 2,083 float profiles and 316 seal profiles. The seal count swings a lot from year to year, probably partly with how many seals were tagged.

Pooled over 1979–2025, floats reached more places: 70% of the monthly grid cells with measurements (1° latitude by 3° longitude) had at least one float, against 29% with seals. In 27% of the cells, only seal profiles were recorded.