In a study in Science Advances, a team around Lonnie Thompson of Ohio State University reports that the record from cores of Tibet's Guliya ice cap reaches back more than 100,000 years. Hou et al., however, in a study published in September 2025, used radiometric methods to date a layer of a newer core from the same ice cap to a much lower age than the 1992 core's timescale had given it.
01Drilling
The young date comes from a core that did not reach bedrock
Guliya is an ice cap in the north-west of the Tibetan Plateau, surrounded by vertical ice walls 30 to 40 metres high. In 1992 a joint American–Chinese expedition pulled a core 308.6 metres long from it.
The Guliya plateau was drilled a second time in 2015, at an altitude of about 6,200 metres, and the 309.79-metre GP2015 core reached the bedrock. The third core, GP2021, was drilled by Hou et al. at an altitude of 6,061 metres. It is 175.1 metres long and ends about 55 metres above the bedrock, so it did not capture the lowest ice.
The 2026 study works with the GP1992 and GP2015 cores. Thompson says that outside the polar regions, no other mountain ice core reaches as far back in time as the record from these cores.
02Estimates
Age estimates differ by hundreds of thousands of years
The timescale of the upper part of the GP1992 core was long in common use. It reaches back as far as 110 thousand years and rests on oxygen isotopes in the ice. Later age estimates differ depending on which ice was dated and by what method.
| Year | Who | What was dated | Method | Result |
|---|---|---|---|---|
| 1997 | Who: Thompson et al. | What was dated: ice at the base of the GP1992 core | Method: chlorine-36 | Result: more than 500,000 years, and the timescale of the core reached about 760,000 years |
| 2019 | Who: Tian et al. | What was dated: ice from the base at the southern margins of the cap | Method: krypton-81 | Result: upper age limits only, 15 to 74 thousand years |
| 2025 | Who: Hou et al. | What was dated: one layer of the GP2021 core at 175.1 m | Method: lead-210, argon-39, carbon-14 | Result: about 2.7 thousand years |
| 2026 | Who: Thompson et al. | What was dated: the whole record of the GP1992 and GP2015 cores | Method: beryllium-10, chlorine-36 and comparison with caves | Result: more than 100,000 years, and according to the press release possibly more than 128,000 |
03Methods
Every method of dating ice has its weak point
Reading an ice core requires an accurate timescale. In the upper part of a core, annual layers can be counted. Deeper down the ice is compressed and its age is determined indirectly, which is why the authors combine several methods.
| Method | What is measured | What follows | Objection |
|---|---|---|---|
| Oxygen isotopes in ice | What is measured: the pattern in the GP1992 and GP2015 cores | What follows: the two records resemble each other, so they are repeatable | Objection: the similarity of two cores does not by itself determine the age |
| Beryllium-10 and chlorine-36 | What is measured: the trace of the Laschamp excursion, a reversal of Earth's magnetic field more than 41,000 years ago | What follows: a fixed point on the timeline; in the GP2015 core, earlier work by Thompson et al. from 2024 places it at a depth of 187.4 m, based on the chlorine-36 peak | Objection: Hou et al. consider the chlorine-36 in the ice near the bottom of the GP1992 core unreliable. They also question the Laschamp peak because samples were taken more densely only around the expected depth. |
| Comparison with Tibetan caves | What is measured: the shape of the oxygen isotope curve in ice and in cave deposits | What follows: an estimate that the record may reach back over 128,000 years | Objection: it is a comparison of curves, not direct dating of the ice |
04Contradiction
Two cores from one ice cap give different ages
Hou et al. dated a different core from Thompson's and published the result in September 2025 in Science Advances. The layer at 175.1 metres of the GP2021 core corresponds stratigraphically to about 253 metres of the GP1992 core, so it is the same layer at a different depth. The 1992 core's timescale gives it roughly 100 thousand years, while lead-210, argon-39 and carbon-14 give roughly 2.7 thousand years, with an upper uncertainty limit of about 4.5 thousand.
Hou and colleagues thereby question the timescale of the 1992 core, by which the layers are ordered. Their core, however, did not reach the deepest ice, whereas the 2026 study assesses the whole record of two longer cores.
05Resolution
Only dating of deeper ice will settle the dispute
Thompson says that if Tibet held only ice from the Holocene, it would mean the large ice sheets of the last glacial period did not survive the early Holocene. In his view the evidence shows that they did. He relies on a timescale that Hou and colleagues question.
Hou et al. propose radiometrically dating the deeper parts of the cores too, for example with krypton-81, argon-39 and carbon-14. Thompson's team wants to use further samples to find out how other climate and environmental indicators changed.
Frost and temperatures where you liveIn the local forecast you will see what frost and temperatures to expect where you live.Open your local forecast