The Oldest Ice Ever Found On Earth

The oldest ice ever recovered from Earth's surface is about 6 million years old. It came from the Allan Hills, a windswept patch of East Antarctica, and a team led by Sarah Shackleton and John Higgins reported it in the Proceedings of the National Academy of Sciences (PNAS) in October 2025 (Shackleton et al., 2025). That is a different record from the oldest continuous ice core, a roughly 1.2-million-year sequence drilled at Little Dome C by the Beyond EPICA project. Both numbers are real, and they answer different questions.

Ice sounds like a strange thing to set an age record for. It melts and flows, and unlike rock or fossil bone it seems too fragile to last millions of years. Yet in the right place it is one of the best archives on the planet, because it holds samples of ancient air.

What an ice core actually preserves

An ice core is a cylinder of glacial ice pulled out by drilling straight down through an ice sheet, then read layer by layer to reconstruct past climate. Snow that falls on Antarctica does not turn into ice at once. It first becomes granular, porous firn, and as more snow piles on top over decades the firn compacts under its own weight until the pores pinch shut and leave sealed bubbles.

According to the Wikipedia overview of ice cores, that sealing happens when the material reaches a density of about 830 kilograms per cubic meter, at depths that range from roughly 64 to 115 meters across Greenland and Antarctica. At Dome C the close-off depth is about 95 meters, where the surrounding ice is already around 2,500 years old. That detail matters later: the air in a bubble is always younger than the ice around it, and the gap grows where snowfall is low.

Those bubbles are the scientific payoff. Each one is a physical sample of the atmosphere at the moment the layer sealed, not an estimate built from indirect evidence. Researchers can crush a piece of ice in a vacuum chamber, release the gas, and measure its carbon dioxide and methane directly, the same way you would test a sample of air today.

The ice itself carries a second, independent line of evidence. Water molecules differ slightly in mass depending on which oxygen and hydrogen isotopes they contain, and the ratio of heavy to light isotopes in snowfall shifts predictably with temperature. Measuring those ratios along a core gives a temperature record that runs in parallel with the greenhouse-gas record from the bubbles and cross-checks it.

Key fact: The roughly 6-million-year-old Allan Hills ice is older than the earliest fossil evidence of the genus Homo, the LD 350-1 jaw from Ethiopia dated to about 2.8 million years ago, by more than 3 million years.

Depth works like a stack of newspapers

Age increases with depth in roughly the way publication dates increase down a stack of newspapers: the oldest issues sit at the bottom, undisturbed by what was added later. Near the surface, individual annual layers can be counted like tree rings. A tree lives a few thousand years at most. An ice sheet that is thousands of meters thick and has been collecting snow for well over a million years can outlast every living organism on the planet and still keep a fine-grained record.

The analogy breaks at the bottom of the stack. Ice flows, and the deepest layers get thinned, folded, and sheared. That is the reason a continuous record is so hard to get, and it comes up again below.

How scientists put a date on ancient ice

Age is the whole point of this article, so the dating methods deserve a closer look. No single technique works across 6 million years, and the methods change as the ice gets older.

Method	What it uses	Example or limit from the sources	Weakness
Annual layer counting	Visible or chemical yearly layers	Works only to a limit, about 55,000 years at Vostok	Layers thin out and blur with depth
Ice-flow modeling	Physics of how an ice sheet accumulates and deforms	Used alongside other methods in deep cores	Needs assumptions about past snowfall
Orbital matching	Methane and the oxygen-to-nitrogen ratio, matched to Earth's orbital cycles	Used to tune the age scale of deep cores	Indirect, depends on the tuning target
Volcanic markers	Ash and sulfate from eruptions dated elsewhere	Toba, about 72,000 years ago, and the Laschamp event, about 40,000 years ago	Only works where a marker is found
Argon isotopes in trapped air	Measuring an argon isotope in the air bubbles themselves	The method behind the roughly 6-million-year Allan Hills date	Gives a date for a sample, not a full sequence

The first four rows come from the Wikipedia article on ice cores, which also notes that the age of the gas and the age of the ice can differ by more than 1,000 years at low-snowfall sites such as Vostok. The last row is the one that matters for the record-setting Allan Hills samples. The WHOI press release says the team dated the ice directly, by measuring an isotope of the noble gas argon in trapped air, rather than inferring its age from layers.

That is why the Allan Hills result is described as the oldest directly dated ice. Layer counting is useless at 6 million years, because nobody can count layers through ice that has been stretched and folded for that long. A property of the sample itself has to carry the age, and the argon measurement does.

The Beyond EPICA project's 1.2-million-year continuous core

Beyond EPICA – Oldest Ice is a European project, funded by the European Commission under its Horizon 2020 program and coordinated by Italy's National Research Council (CNR). According to the project's record on CORDIS, the EU contribution is about €11 million, and the project runs from June 1, 2019, to May 31, 2027. Its stated aim is to recover the first stratigraphically undisturbed ice-core record reaching beyond 1 million years, with a target of 1.5 million years.

The drill site is Little Dome C, on the central Antarctic plateau at about 3,200 meters (10,500 feet) altitude, where the average summer temperature is about minus 35 °C. It lies on the same plateau as Concordia Station. The Wikipedia article on EPICA explains why the region is good for this: Concordia sits at 3,233 meters elevation, the annual mean air temperature is minus 54.5 °C, and snow accumulates at only about 25 millimeters a year, so a very long record is packed into a thin stack of ice. The Little Dome C location itself was chosen in 2022.

What the drill actually reached

A CNR press release dated January 9, 2025, announced that the project's fourth field campaign had drilled a 2,800-meter core to the bedrock beneath the ice sheet. Twelve scientific institutions from ten European countries took part, and the team spent more than 200 days drilling and processing ice across four field seasons.

The important detail is where the continuous record ends. The release says the upper 2,480 meters is expected to hold a continuous climate record going back 1.2 million years, possibly more, with up to 13,000 years compressed into a single meter of ice in that section. The lowest 210 meters above bedrock is old, heavily deformed ice of unknown origin. So the core reaches bedrock, but only the upper part is the intact timeline.

Some context on the scale of that compression. By the end of November 2024, per CORDIS, drilling had reached 2,059.74 meters, and the ice at 1,836.18 meters preserved climate records from about 195,000 years ago. The remaining few hundred meters carry the rest of the million-year story, which is why each additional meter is so valuable.

Reading the record: The record is real, but it comes with caveats. The CNR release words the 1.2-million-year span as what the upper section is "expected" to hold, and the Wikipedia article on EPICA says analyses at the Alfred Wegener Institute confirmed that the samples are older than 1.2 million years. Processing continues into the project's last year. The CNR release says the cores were due to travel to Europe by icebreaker under a cold chain held at minus 50 °C, and CORDIS says the first 1,620 meters were processed at the Alfred Wegener Institute in summer 2024. A fifth and final campaign began in November 2025, according to the same Wikipedia article.

The previous record for a continuous core, from the original EPICA Dome C drilling, reached back about 800,000 years, so the new core extends it by roughly 400,000 years. For a sense of what older, shallower drilling achieved, see the table later in this article.

Why isn't Beyond EPICA the "oldest ice ever found"?

Because oldest and most complete measure different things. Beyond EPICA's strength is that its upper section keeps every layer in order across 1.2 million years. The Allan Hills samples reach much further back in time, but they are scattered pieces recovered from separate places, with no guarantee of what happened between them. A record can be the longest unbroken one without being the single oldest object recovered, and the reverse.

The 2.7-million and 6-million-year-old ice from Allan Hills

Long before Beyond EPICA reached bedrock, a different part of Antarctica was already producing larger age numbers, though never as a tidy sequence. The Allan Hills are a mostly ice-free group of hills about 22 kilometers (14 miles) long near the heads of the Mawson and Mackay Glaciers in Victoria Land, according to Wikipedia. They are better known to many people for meteorites, including the first lunar meteorite ever recovered, ALHA 81005, found in the 1981 to 1982 season.

For glaciologists, the area is interesting for blue ice: patches where flowing ice runs into a mountain barrier, strong winds strip away the fresh snow, and old ice ends up near the surface instead of buried under kilometers of newer snow. Ice recovered there in 2016, dated by argon methods, was about 2.7 million years old, a record at the time, as the Wikipedia article on ice cores describes. It was not a conventional record, because the layers needed for standard core dating were missing.

The 6-million-year result

In October 2025, a team led by Shackleton of the Woods Hole Oceanographic Institution and Higgins of Princeton University reported something older. The paper, titled "Miocene and Pliocene ice and air from the Allan Hills blue ice area, East Antarctica," appeared in PNAS volume 122, issue 44, with a publication date of October 28, 2025, according to Sci.News. The ice is about 6 million years old, which Woods Hole called the oldest directly dated ice and air on the planet.

The work was part of COLDEX, the NSF-funded Center for Oldest Ice Exploration, a collaboration of 15 U.S. research institutions led by Oregon State University. The team drilled shallow cores, only 100 to 200 meters deep, at several locations on the edges of the ice sheet. For comparison, the oldest continuous cores in the interior need more than 2,000 meters of drilling. Woods Hole's release says the team had initially hoped for ice up to about 3 million years old, and the Phys.org summary quotes COLDEX director Ed Brook: "We knew the ice was old in this region."

The samples are not continuous. The Eos coverage compares them to snapshots rather than a video: individual frames from very different moments, with no film in between.

Key fact: Oxygen-isotope measurements on the oldest Allan Hills ice indicate that the region cooled by about 12 °C (roughly 22 °F) over the last 6 million years, which Woods Hole describes as the first direct measurement of Antarctic cooling over that span.

How can ice this old survive at the surface?

Three conditions combine, according to the Woods Hole and Sci.News accounts. Ice flow and rugged mountain terrain bring old ice up toward the surface. Strong winds remove fresh snow that would otherwise bury it. And the cold slows the movement of the ice nearly to a standstill. Not all blue ice is old, which is why finding ice this ancient took dating of many samples, and researchers are still working out why the ice survives so near the surface in this particular region.

The records side by side

A table helps keep the numbers straight, because the records differ in kind as well as age.

Record	Place	Reach back	Continuous?	Depth
Vostok core	Antarctica	about 420,000 years (ice at 3,310 m)	Yes, with mixing deeper down	3,769 m, deepest core drilled
EPICA Dome C	Antarctica	about 800,000 years	Yes	3,270 m, completed December 2004
NEEM	Greenland	about 128,500 years	Yes	2,537 m (2010)
Beyond EPICA	Little Dome C, Antarctica	about 1.2 million years	Upper 2,480 m expected continuous	2,800 m to bedrock
Allan Hills (2016)	Antarctica	about 2.7 million years	No, fragments	Shallow blue ice
Allan Hills (2025)	Antarctica	about 6 million years	No, fragments	100 to 200 m

Sources: Wikipedia, Ice core; Wikipedia, EPICA; CNR press release (January 2025); WHOI press release (October 29, 2025). Depth alone does not set the age. Snowfall does: the WAIS Divide core in West Antarctica reached 3,405 meters yet covers only about 62,000 years because so much snow falls there, while a low-snowfall plateau packs far more time into each meter.

A short history of drilling deeper into time

The record has moved in steps, and each step needed a new drill site rather than just a longer drill. A rough sequence, drawn from the Wikipedia ice core overview and the press releases cited below:

• Vostok, Antarctica. The deepest core ever drilled reached 3,769 meters. Ice at 3,310 meters is estimated at about 420,000 years old, and deeper ice is mixed enough that the record becomes hard to interpret.
• Dome F, Antarctica. A hole drilled to 3,035 meters in 2006 was estimated to reach ice about 720,000 years old.
• EPICA Dome C, Antarctica. Drilling finished in December 2004 at 3,270 meters, about 5 meters above bedrock. Because snowfall at the site is so low, the usable record reaches back roughly 800,000 years and shows eight previous glacial cycles.
• Beyond EPICA, Little Dome C. A 2,800-meter core, announced in January 2025, expected to hold a continuous record back to about 1.2 million years in its upper 2,480 meters.
• Allan Hills blue ice. Fragments about 2.7 million years old from 2016, and about 6 million years old from the 2025 paper.

Notice that the deepest core is not the oldest. Vostok drilled farther down than any other project, but a deep hole at a site with high snowfall simply spreads the same number of years across more meters. Low snowfall is the main lever, which is why the Concordia region, with only about 25 millimeters of snow a year, holds so much time in so little ice. The Allan Hills approach turns the problem around: instead of drilling through the whole ice sheet, it samples places where old ice has already been lifted to the surface.

Two records, two scientific questions

The difference is easiest to see as a diary. The Beyond EPICA core is a diary written every day, with no pages missing for 1.2 million years, so you can follow a change as it happens. The Allan Hills samples are loose pages torn from a diary that may go back much further, each readable and individually datable, but with unknown gaps in between.

Each kind of evidence answers a different question. A continuous core is what you need to study gradual transitions, such as how quickly carbon dioxide changed as the planet slid into or out of an ice age, because every intervening step is there. A scattered fragment answers a simpler question that is still valuable: what was the atmosphere like at some specific moment tens of millions of years further back than any continuous core can reach? According to Eos, modeling suggests a continuous core may not go beyond about 1.5 million years, which is why the team argues that fragments are the route to earlier times.

Press coverage sometimes squeezes this into a single "oldest ice" headline. A story about the Beyond EPICA milestone in early 2025 and a story about the Allan Hills milestone in late 2025 describe two legitimate records, not competing claims about one thing.

Why the Mid-Pleistocene Transition makes 1.2 million years the important number

Even though the Allan Hills fragments are older, the Beyond EPICA core's span was chosen on purpose, because it crosses a shift in Earth's climate rhythm that is still not fully explained. The Mid-Pleistocene Transition lasted about 550,000 years, from roughly 1.25 million to about 0.7 million years ago, according to Wikipedia. Before it, glacial cycles followed a rhythm of about 41,000 years, matched to the wobble in Earth's axial tilt (obliquity), and the ice sheets were thinner and responded in a roughly linear way. After it, cycles averaged about 100,000 years and became lopsided: ice built up slowly through long cooling phases, then collapsed quickly into a warm interglacial.

Researchers have proposed several explanations, and none is settled:

• Declining atmospheric carbon dioxide, combined with the gradual stripping of loose regolith (broken rock and soil) from northern continents so that ice sheets could grip bare bedrock and grow thicker.
• Changes in how much carbon the deep Atlantic stored, or in volcanic outgassing, carbonate weathering, or dust-driven iron fertilization of the oceans.
• A 2023 proposal that "obliquity damping" is a missing link.

Most of the evidence so far comes from deep-sea sediment cores, which are an indirect record: they infer past conditions from ocean chemistry. A continuous ice core spanning the transition would offer direct measurements of carbon dioxide and methane in ancient air bubbles. That is the specific scientific payoff the project was designed for, and it is why the CORDIS record states its goal as a record that covers the transition and reaches toward 1.5 million years.

The physics underneath is the same family that governs how Earth's orbit and tilt shape climate, similar in spirit to how Earth would behave if it had a second moon. The difference is that the shift already happened once, and the ice is physical evidence of what the atmosphere did during it.

What six-million-year-old ice can and cannot tell us

The Allan Hills samples come from the Miocene and Pliocene epochs. Eos describes parts of the ice as forming in periods much warmer than today, with higher sea levels and open forests and grasslands across much of the world. Ice that formed then holds air from a world with a different climate, which makes it a rare direct check on models of how the climate responds to carbon dioxide.

The samples have limits that are easy to forget. They are fragments, so the team cannot reconstruct a smooth curve of temperature or gas concentration. Each date is a spot measurement, and the interpretation of what the Antarctic landscape looked like when the snow fell, whether surface snowpack or permafrost, depends on assumptions. Woods Hole says ongoing work will try to reconstruct past greenhouse gas levels and ocean heat content, and a longer-term study planned for 2026 to 2031 aims to push the records further back, with the COLDEX team due to return to Allan Hills for more drilling. Those are plans, not results.

This is a good example of how a record gets used. A single pair of numbers, 6 million years and 12 degrees of cooling, is only the first sentence of the story. The detail comes from many more samples measured by many more laboratories.

Common misconceptions about "oldest ice"

"Oldest ice" is a single record, like the tallest mountain. It is not. There are at least three numbers worth knowing: the oldest continuous core (about 1.2 million years, Beyond EPICA), the oldest fragment reported before 2025 (about 2.7 million years, Allan Hills), and the current oldest ice overall (about 6 million years, Allan Hills, 2025). Mixing them up misrepresents what each discovery shows.

Ice cores are just thermometers. They hold several independent lines of evidence at once: air in bubbles, oxygen and hydrogen isotope ratios for temperature, volcanic ash and sulfate that can be matched to dated eruptions, and windblown dust that records changes in aridity and wind patterns. Volcanic eruptions leave their signature in the ice as well, and the volcanic chemistry behind that signal is on display in some of the most acidic natural waters on Earth.

Older is better science. Age does not decide scientific value. A 6-million-year fragment can tell you roughly what the carbon dioxide level was back then. Only a continuous sequence can tell you how fast it changed through a specific multi-thousand-year transition. Both questions matter.

The ice is always a perfect time capsule. Ice flows. The bottom 210 meters of the Beyond EPICA core is deformed ice of unknown origin, and the blue-ice fragments have no neighboring layers to verify against. The skill in ice-core science is knowing exactly where a record can be trusted. In some ways it resembles how the body holds blood pH inside a very narrow range: a stable baseline is what makes a measurement meaningful, and anything that disturbs the baseline has to be accounted for.

Why researchers want a longer record at all

Ice cores are not collected for the sake of a record. CORDIS says the project exists to clarify how slow parts of the climate system, such as the carbon cycle and the ice sheets, drove past climate change, and to tighten the constraints on how Earth will respond to future greenhouse gas emissions. The word "slow" is the key. Weather changes in days and ocean currents in decades, but ice sheets and the carbon stored in the deep ocean change over thousands of years. A record has to be long enough to see them move.

The existing 800,000-year record already gives a striking benchmark. The Wikipedia article on EPICA reports that greenhouse gas concentrations in the warmest periods of the last 800,000 years never exceeded post-industrial levels, and quotes the project's coordinator, Carlo Barbante, saying current carbon dioxide is about 50 percent higher than at any time in that span. Extending the continuous record to 1.2 million years adds roughly 400,000 years of context, and it adds the era before the glacial cycles lengthened, when Earth's climate responded to carbon dioxide and orbital changes in a different way.

Older fragments serve another purpose. Parts of the ice formed in periods when the planet was warmer and sea levels were higher, as Eos describes it. That makes those samples the nearest thing to a physical test of what a warmer world looks like, and the 12 °C of Antarctic cooling since then is a figure that climate models can be checked against. The two kinds of record are complementary: one shows how fast the system moves, the other shows how far it has moved.

The people and logistics behind the records

Records like these take years because of where the ice is. Little Dome C sits on the high plateau near Concordia Station, with summers near minus 35 °C, and the Beyond EPICA cores had to be shipped to Europe under a minus 50 °C cold chain. Drilling at that depth means roughly 200 days of drilling and processing spread across four field seasons.

These are among the most isolated places on Earth where scientists still manage to work, and the logistics are part of why the numbers move slowly. A drill that fails or freezes in can cost a season. Cold also makes core handling an engineering problem in its own right, since fragile ancient ice has to stay cold from the borehole to the laboratory freezer.

Glaciology also connects to bigger questions about how ice sheets behave. If you want the mechanics of how ice moves and why it eventually flows to the bottom of the sheet, see our explainer on how glaciers form and move, and for the other side of the story, the scenario piece on what would happen if Antarctica completely melted.

How to read the next "oldest ice" headline

New numbers will keep arriving, because COLDEX plans more drilling at Allan Hills and the Beyond EPICA team is still processing its core. A few questions help sort a real advance from a garbled one.

1. Is the ice continuous or fragmentary? A continuous core and a set of dated fragments are different kinds of record, and a headline that does not say which one is leaving out the most important fact.
2. How was it dated? Direct dating of trapped air, such as the argon method, is a different claim from an age inferred by counting layers or matching to a model.
3. Is the age measured or expected? The Beyond EPICA release words the 1.2-million-year span as an expectation for the upper section, and later laboratory work is what firms it up.
4. Is it ice age or gas age? Air in bubbles is younger than the ice around it, so a quoted age should say which one it refers to.
5. Which peer-reviewed paper is behind it? The Allan Hills result has a title, a volume, and an issue number. A claim that cannot be traced to a paper or an institutional release deserves caution.

Those five questions would have caught most of the confusion between the 1.2-million-year and 6-million-year numbers when both stories ran in 2025.

The takeaway

There are two separate, well-documented records buried in Antarctic ice. The longest continuous climate archive is about 1.2 million years old, drilled by Beyond EPICA at Little Dome C and expected to hold an unbroken record in its upper 2,480 meters, which crosses the Mid-Pleistocene Transition. The oldest individual ice is about 6 million years old, pulled from shallow blue-ice cores at Allan Hills and reported in PNAS in October 2025, reaching back to the Miocene in scattered pieces rather than a single sequence.

The lesson reaches beyond the trivia. Different questions need different kinds of evidence, and the best archives are the ones whose limits are stated as clearly as their dates. Antarctica has been quietly recording the atmosphere for longer than any other readable archive on Earth, and scientists are only now learning to read both the long unbroken diary and the loose pages.

---

Read next:

• The oldest things on Earth
• How scientists measure the age of Earth
• Climate change science: how it actually works

FAQ

Q: What is the oldest ice ever found on Earth? A: About 6 million years old, from shallow cores at the Allan Hills blue ice area in East Antarctica. A team led by Sarah Shackleton and John Higgins reported it in PNAS in October 2025 (published online October 28). It is a set of separately dated samples, not one continuous sequence.

Q: What is the oldest continuous ice core ever recovered? A: The Beyond EPICA core, drilled to bedrock at 2,800 meters at Little Dome C in Antarctica, announced in January 2025. Its upper 2,480 meters are expected to hold a continuous record reaching back about 1.2 million years. The bottom 210 meters is deformed ice of unknown origin.

Q: How do scientists date ice that is millions of years old? A: For the Allan Hills fragments, the team measured an isotope of the noble gas argon in air trapped in the ice, which dates the sample directly. For continuous cores they combine layer counting near the surface with ice-flow models, orbital matching using methane and the oxygen-to-nitrogen ratio, and volcanic markers.

Q: Why do scientists care about ice older than 1 million years? A: Ice from that interval lets them measure carbon dioxide and methane directly across the Mid-Pleistocene Transition, when glacial cycles shifted from a roughly 41,000-year rhythm to a roughly 100,000-year rhythm. The cause is still debated, and ocean sediment cores, the main earlier source, only record the change indirectly.

Q: What is the difference between the Beyond EPICA core and the Allan Hills ice? A: Beyond EPICA is a single deep core drilled straight down to bedrock, designed to keep layers in order. The Allan Hills ice comes from shallow cores in a blue ice area where old ice has been pushed near the surface, so it reaches much further back but only in disconnected pieces.

Q: How much has Antarctica cooled since the oldest Allan Hills ice formed? A: Oxygen-isotope measurements on the ice indicate a gradual cooling of about 12 °C (roughly 22 °F) over about 6 million years, according to Woods Hole Oceanographic Institution. It is described as the first direct measure of Antarctic cooling over that span.

Sources

• Shackleton et al. (2025), PNAS, "Miocene and Pliocene ice and air from the Allan Hills blue ice area, East Antarctica"
• Woods Hole Oceanographic Institution press release, October 29, 2025
• Phys.org summary of the Allan Hills study
• CORDIS, Beyond EPICA project record
• CNR press release on the 2,800-meter Beyond EPICA core, January 2025
• Eos, "New lessons from old ice"
• Wikipedia: Ice core, EPICA, and Mid-Pleistocene Transition