Ice Crystals Deep in Tibetan Glaciers Preserve Hidden Records of Past Temperature

Deep beneath the summits of the Tibetan Plateau, ice crystals are quietly keeping a diary of the climate. In a new study published in The Cryosphere, Zhengqiang He and Baiqing Xu of the Institute of Tibetan Plateau Research at the Chinese Academy of Sciences report the first continuous profiles of ice grain size measured from the surface firn all the way to the base of two deep mountain glacier ice cores. Their findings reveal that the microscopic architecture of mountain glacier ice, long assumed to be too chaotic for paleoclimate work, follows patterns strikingly similar to those seen in the polar ice sheets, and that under the right conditions the size of individual ice crystals can still carry the fingerprint of past temperatures. The result opens a new window onto climate history in one of the most climate-sensitive regions on Earth, the so-called Third Pole, where meltwater from glaciers sustains rivers feeding billions of people.

The team worked with two full-depth cores recovered from the accumulation zones of Tibetan glaciers. The first, 169.45 meters long, was drilled in November 2020 from the summit of the Weigeledangxiong Glacier on Mount Animaqing in the eastern Kunlun Mountains, at 5750 meters above sea level. The second, 77.6 meters long, came in November 2022 from the southern slope of Mount Bugyai Kangri in the eastern Tanggula Mountains, at 6180 meters. Both regions are dominated by the Asian monsoon and both glaciers have been retreating in recent decades, but they differ in a crucial way: the Animaqing site is warmer, with borehole temperatures ranging from about -0.22 to -6.84 degrees Celsius, while the Bugyai Kangri core is colder, spanning roughly -1.54 to -8.31 degrees Celsius. That temperature contrast turned out to be the key to interpreting everything the crystals had to say.

Measuring grain size continuously along an ice core is a formidable technical challenge, and the authors solved it with a Microstructure Mapping system adapted from protocols developed for polar cores. Vertical ice slices, cut with a band saw and shaved flat with a Leica microtome, were sublimated at -20 degrees Celsius for about fifteen hours until grain boundaries stood out in sharp relief. Each slice was then photographed under vertical coaxial illumination, in which air bubbles and grain boundaries appear dark against the translucent ice, producing panoramic images at a resolution of roughly eight micrometers per pixel. A U-Net neural network performed the binary segmentation of grain boundaries, and the ImageJ software package with its MorphoLibJ plugin extracted the area, position and shape of every grain. In total the researchers analyzed 1890 slices from Animaqing, covering 91 percent of the core, and 560 slices from Bugyai Kangri, covering 71 percent, a sampling density that captures centimeter-scale variations invisible to traditional discrete sampling.

The resulting profiles divide naturally into four depth zones that mirror, in modified form, the classic three-stage model established for polar ice cores. In the firn layer, extending to about 35 meters in both cores, snowflakes of roughly one square millimeter are slowly compressed as irregular pores close off; average grain areas remain small, below about two square millimeters at the surface, and grow gently with depth. Both cores reached the critical density of 0.830 grams per cubic centimeter for pore closure at approximately 35 meters, sealing the atmospheric archive in bubbles. Below that lies a ten-meter coarsening zone in which the firn becomes true ice and average grain areas rise rapidly, reaching 18.00 square millimeters in the Animaqing core and 8.06 square millimeters at Bugyai Kangri, as larger grains grow by consuming their smaller neighbors.

This early growth is governed by Normal Grain Growth, a textbook process in which grain boundaries migrate to reduce the total surface energy of the ice, and it is strongly temperature-dependent: warmer ice allows boundaries to migrate faster, so crystals grow larger. The data show this vividly. The Animaqing core, several degrees warmer than Bugyai Kangri, maintains substantially larger grains at every comparable depth even where it bears greater overburden stress. In the recrystallization zone that follows, the trend reverses. From about 45 meters down to 140 meters at Animaqing and 73 meters at Bugyai Kangri, average grain area stabilizes and declines below five square millimeters, small grains become more abundant, and micrographs fill with subgrain boundaries, the telltale signature of Rotation Recrystallization. Under accumulating stress, subgrain boundaries rotate and gradually mature into true grain boundaries, splitting large crystals into smaller ones and setting up a dynamic balance between stress-driven refinement and temperature-driven growth.

The deepest zone belongs to Strain-Induced Boundary Migration, a rapid recrystallization mechanism that flourishes when ice is warmer than about -10 degrees Celsius, a condition satisfied at the base of both cores. Here average grain areas surge, the population of large crystals grows roughly ten times faster than in the coarsening layer, and micrographs show interlocked grains with highly curved boundaries. At Animaqing the effect is spectacular: below 150 meters, where temperatures climb above -6.27 degrees Celsius, crystals balloon to more than 400 square millimeters, and near the very base they exceed 3000 square millimeters, their boundaries spilling beyond the edges of the prepared slices. There, the researchers infer, repeated freeze-thaw cycles in excessively warm ice generated a five-meter-thick bubble-free layer of refrozen ice with gigantic, nearly transparent crystals.

Between these hundred-meter trends lie two centimeter-scale troublemakers, and both are ultimately climatic in origin. The first is melt-refreezing. Because surface temperatures at Animaqing approach zero, summer meltwater percolates into the firn and refreezes, producing centimeter-scale ice layers in which liquid-mediated recrystallization drives crystal sizes upward in abrupt jumps. The colder Bugyai Kangri site rarely experiences this, following a classic cold-type ice formation without meltwater involvement. The second troublemaker is impurity. Both cores contain millimeter to centimeter bands rich in dust and soluble impurities, and within these bands crowds of small grains, typically under two square millimeters, pile into walls that drag average grain area sharply down. In the Animaqing core a dense zone of impurity layers between 120 and 140 meters contains up to 90 centimeters of impurity-bearing ice per meter of core, and impurity thickness correlates strongly and negatively with grain size, with a correlation coefficient of -0.715. The mechanism is thought to be pinning: impurity particles obstruct grain boundary migration, holding grains small in much the same way documented in polar cores from Greenland and Antarctica and in the Monte Perdido Glacier of the Pyrenees.

The decisive test of the study is whether grain size survives this barrage of confounding influences and still records temperature. To find out, the authors compared average grain area with the oxygen isotope ratio, expressed as δ18O, the workhorse paleoclimate proxy in Tibetan ice core research. In the interval between 45 and 51 meters, where vertical pressures are low enough that Rotation Recrystallization is negligible and Normal Grain Growth dominates, the Bugyai Kangri core shows a statistically significant correlation between grain area and δ18O, with r equal to -0.286, corresponding to ice dated between 1975 and 1985. The correlation at Animaqing is obscured by heavy refreezing, but in the dense impurity zone the Animaqing core instead delivers a significant positive correlation of 0.205, tying small crystals to cold, dusty periods on an inferred centennial scale. In both settings, the physical texture of the ice is keeping step with the chemical thermometer.

Comparisons with other mountain glaciers reinforce the temperature story. When the authors placed their cores alongside four others from the eastern Tibetan Plateau, the Tianshan Mountains and the European Alps, the ranking of grain sizes at 35 meters depth, where vertical pressures are broadly comparable, broadly tracked the ranking of ice temperatures. The Guliya core tops the grain size list and the Tsanfleuron core tops the temperature list, with Animaqing, Bugyai Kangri and the KCC core following in the expected order. The consistency, the authors argue, marks temperature as the dominant control on grain growth in mountain glaciers, outweighing even the greater stresses borne by deeper, warmer ice.

The implications reach well beyond crystallography. Ice cores from the polar sheets have long used grain size as a supplementary climate archive, with abrupt changes in crystal dimensions marking cold-warm transitions, but mountain glaciers were widely considered too warm, too fast-flowing and too contaminated to preserve such signals. This study demonstrates that a mountain glacier ice core can, in specific layers where stress-driven grain refinement is suppressed, hold onto temperature information despite melt layers, impurities and rapid flow. That matters for the Third Pole, where observational records are short and paleoclimate archives are scarce, and it suggests that physical properties of ice, from grain size to bubble number density, could eventually complement isotope chemistry in reconstructing how High Mountain Asia responded to past warming. As glaciers across the Tibetan Plateau retreat under modern climate change, their crystals may prove to be among the last witnesses of the climates that built them, recording in millimeter-scale geometry the temperatures of decades, centuries and perhaps millennia gone by.

Subject of Research: Temperature-dependent grain size variation and climate signal preservation in Tibetan Plateau ice cores

Article Title: Temperature dependence of grain size in Tibetan ice core

Article References: He, Z., & Xu, B. (2026). Temperature dependence of grain size in Tibetan ice core. The Cryosphere, 20(10), 5745-5759. https://doi.org/10.5194/tc-20-5745-2026

Image Credits: AI Generated

DOI: 10.5194/tc-20-5745-2026

Keywords: ice cores, Tibetan Plateau, grain size, recrystallization, paleoclimate, glaciology, oxygen isotopes, Third Pole, The Cryosphere, ice microstructure, mountain glaciers, climate change