A recent 2026 study in Geophysical Research Letters (Hou et al.) estimates that global mean surface temperature during the Last Glacial Maximum (LGM, ~20,000- 21,000 years ago) was about 5.4°C colder than today, based on tropical land cooling constraints combined with climate model relationships.

Researchers compiled a multiproxy dataset of 61 tropical/subtropical (low- to mid-latitude) land temperature records from pollen assemblages, noble gases in ancient groundwater, glacial evidence (e.g., equilibrium line altitudes), and other proxies.

They estimated tropical land cooling of roughly 4.2°C relative to the present.

Using emergent relationships from 13 state-of-the-art LGM climate model simulations (linking tropical land temperatures, tropical sea-surface temperatures, and global mean temperature), they inferred tropical ocean cooling of about 2.6°C and the global figure of ~5.4°C.

A closely related preprint by the same group reported somewhat stronger cooling (tropical land −4.8°C with 95% CI −5.6 to −4.0°C; tropical SST −2.9°C; global mean −5.9°C with 95% CI −7.3 to −4.7°C), suggesting the published numbers reflect refinements.

Many earlier reconstructions (including those relying heavily on the MARGO sea-surface temperature compilation) suggested milder global cooling, often in the ~4- 5°C range or less for the tropics.

Stronger cooling estimates exist: Wikipedia notes ~6°C global average; a 2020 data-assimilation study (Tierney et al.) found −6.1°C (95% CI −6.5 to −5.7°C); noble-gas groundwater studies indicate low-to-mid-latitude land cooling of ~5.8°C (±0.6°C).

Tropical cooling has long been debated. Older CLIMAP reconstructions suggested limited tropical ocean cooling, but land proxies (pollen shifts, snowlines, noble gases from sites like Brazil) and some ocean proxies have repeatedly indicated larger changes (~5°C or more on tropical land in places).

The new work places the LGM toward the colder end of published ranges.

Because LGM CO₂ was much lower (~180- 200 ppm vs. pre-industrial ~280 ppm), a larger temperature drop implies higher equilibrium climate sensitivity (warming per CO₂ doubling).

The results make very low sensitivity (ECS ≲ ~3°C) less likely and help constrain models that underproduce glacial cooling.

Uncertainties remain: proxy scatter, dating precision, local effects, and the fact that the global number is inferred via model relationships rather than directly observed everywhere.

The approach is presented as adaptable to other periods or variables. Overall, it reinforces that the tropics experienced substantial cooling during the LGM peak and that the global climate system responded strongly to the lower greenhouse- gas levels and expanded ice sheets of that time.

_____________________________________________________________________________________

Tropical land temperature proxies, essentially an “ancient jungle thermometer”, can be used to constrain global cooling during the Last Glacial Maximum (LGM, ~20,000–21,000 years ago), as shown in the 2026 study by Hou et al. published in Geophysical Research Letters.

Tropical regions (roughly 30°S–30°N) preserve a variety of climate proxies that record past land temperatures:

• Pollen and vegetation assemblages (shifts in plant communities that track temperature and moisture).
• Noble gases dissolved in ancient groundwater.
• Glacial evidence such as equilibrium-line altitudes of tropical mountain glaciers.
• Other indicators (e.g., certain geochemical or fossil records).

These provide relatively direct estimates of how much cooler tropical land surfaces were during the LGM compared with today. Because the tropics cover a large fraction of Earth’s surface and respond more uniformly than high-latitude regions dominated by ice sheets, they offer a useful anchor for global temperature reconstructions.

The researchers compiled a multiproxy dataset of tropical land temperature reconstructions and combined it with an ensemble of climate model simulations of the LGM. Within those models they identified stable statistical (“emergent”) relationships connecting:

• Tropical land temperature change,
• Tropical sea-surface temperature change, and
• Global mean surface temperature change.

They then applied the observed tropical land cooling to these relationships. Results from the study (and its preprint) indicate:

• Tropical land cooled by roughly 4.2–4.8 °C,
• Tropical oceans cooled by about 2.6–2.9 °C,
• Global mean surface temperature was approximately 5.4–5.9 °C colder than today.

These figures place the LGM toward the colder end of recent estimates and are colder than many reconstructions that relied heavily on earlier sea-surface temperature compilations such as MARGO.

A larger LGM temperature drop under substantially lower atmospheric CO₂ implies a more sensitive climate system (higher equilibrium climate sensitivity).

The tropical “thermometer” therefore helps tighten constraints on how strongly Earth’s climate responds to greenhouse-gas changes, while also offering a transferable method that can be applied to other time periods once sufficient proxy data exist.

An ancient tropical land- temperature record can serve as a practical thermometer that, when combined with model relationships, reveals the scale of the global chill during the peak of the last Ice Age.

_____________________________________________________________________________________

MARGO stands for Multiproxy Approach for the Reconstruction of the Glacial Ocean Surface. It is a major international collaborative project (supported by IMAGES and PAGES) that produced one of the most influential reconstructions of sea- surface temperatures (SSTs) during the Last Glacial Maximum (LGM).

The primary synthesis was published in 2009 by the MARGO Project Members (led by C. Waelbroeck and colleagues) in Nature Geoscience:
“Constraints on the magnitude and patterns of ocean cooling at the Last Glacial Maximum” (doi:10.1038/ngeo411).

Time interval: Strictly defined as 23–19 thousand years before present (following the EPILOG definition), narrower and more rigorous than the broader window used by the earlier CLIMAP project.

Proxies used (multiproxy philosophy):

• Microfossil assemblage transfer functions (planktonic foraminifera, diatoms, dinoflagellate cysts, radiolarians).
• Geochemical proxies (alkenone unsaturation index $U^k_{37}$U^k_{37} and Mg/Ca ratios in planktonic foraminifera).

Data volume: Approximately 696 individual SST reconstructions from sediment cores worldwide.

• Proxies used (multiproxy philosophy):
  • Microfossil assemblage transfer functions (planktonic foraminifera, diatoms, dinoflagellate cysts, radiolarians).
  • Geochemical proxies (alkenone unsaturation index $U^k_{37}$U^k_{37} and Mg/Ca ratios in planktonic foraminifera).

The project emphasized that no single proxy is superior; combining them with quality assessments produces a more robust picture. Data were calibrated against a common modern temperature dataset and quality- controlled, with reliability flags assigned to individual records.

MARGO revealed stronger and more structured cooling than many earlier views:

• Strongest annual-mean cooling (up to −10°C) occurred in the mid-latitude North Atlantic, with a steep temperature gradient between ~45°N and 60°N.
• Cooling was more pronounced in the eastern than the western North Atlantic (contrasting with some CLIMAP patterns).
• Tropical oceans (30°S–30°N) showed heterogeneous cooling, generally larger and more extensive than in CLIMAP, with greater cooling in the tropical Atlantic than in the Indian or Pacific. Western Pacific Warm Pool cooling was modest (roughly 1–3°C).
• Large longitudinal (east–west) SST gradients existed in all major ocean basins—features not well captured by the climate models available at the time.
• Some limited regions showed relative warmth or minimal cooling (e.g., parts of the northwest Australian margin), attributed to changes in ocean circulation.

Basin- averaged annual-mean cooling estimates (from the multiproxy dataset) were on the order of roughly −1.5 to −2.5°C depending on the ocean, with substantial spatial variability and uncertainty. Global tropical cooling was relatively modest compared with some later land-based or data- assimilation reconstructions.

Comparison with CLIMAP and Later Work

• vs. CLIMAP (1970s–1980s): MARGO improved temporal definition, data density (especially in the North Atlantic and tropics), multiproxy integration, and quality control. It showed more heterogeneous tropical cooling and different longitudinal patterns.
• Influence and limitations: MARGO became a standard benchmark for evaluating paleoclimate models (e.g., PMIP) and for estimating climate sensitivity. However, its relatively modest tropical and global-mean cooling has been criticized or revised by later work. Some studies argue it underestimates cooling in certain regions due to proxy limitations, calibration issues, or sparse coverage (especially in the Pacific and Southern Ocean). Newer compilations (e.g., TEA20) and data-assimilation products that blend proxies with models often indicate stronger overall LGM cooling (global SST anomalies closer to −3°C or more in some reconstructions).

The MARGO dataset (including gridded products on a 5° × 5° grid, individual site data, and reliability assessments) is publicly archived at:

• NOAA/NCEI World Data Service for Paleoclimatology.
• PANGAEA.

It remains widely used as a foundational reference, even as researchers develop updated multiproxy compilations and model- data fusion approaches that incorporate additional proxies and address spatial gaps.

In the broader context of LGM climate research and equilibrium climate sensitivity debates, MARGO’s relatively mild tropical cooling has historically supported lower- to- moderate sensitivity estimates.

More recent reconstructions that find stronger tropical and global cooling (including some land- proxy and emergent- constraint studies) tend to favor higher sensitivity values once pattern effects and other forcings are accounted for.

_____________________________________________________________________________________

“Reassessing Last Glacial Maximum Cooling With Proxy-Constrained Emergent Relationships”

The study addresses the challenge of reconstructing global temperatures from sparse, unevenly distributed paleoclimate proxy data. It applies an emergent constraint framework:

Researchers assembled a new multiproxy dataset of tropical land temperature reconstructions for the Last Glacial Maximum (LGM, roughly 20-21 ka).

They identified robust statistical (“emergent”) relationships in an ensemble of paleoclimate model simulations linking tropical land temperature change, tropical sea- surface temperature (SST) change, and global mean surface temperature change.

Proxy- based tropical land cooling was then used to constrain the harder- to-observe tropical SST and global mean values.

From the preprint abstract:

• Tropical terrestrial (land) temperature change: −4.8°C (95% CI: −5.6 to −4.0°C).
• Constrained tropical SST change: −2.9°C (95% CI: −3.6 to −2.3°C).
• Constrained global mean surface temperature change: −5.9°C (95% CI: −7.3 to −4.7°C).

A contemporaneous news summary of the published paper reported slightly milder central values (tropical land ≈ −4.2°C, tropical oceans ≈ −2.6°C, global ≈ −5.4°C), still pointing to relatively strong LGM cooling.

These estimates indicate a colder LGM climate than reconstructions that rely heavily on the MARGO SST compilation (which generally suggested milder tropical ocean cooling).

Broader Implications

The results support stronger glacial cooling, consistent with some independent lines of evidence (e.g., noble- gas groundwater studies showing substantial low-latitude land cooling).

Stronger LGM cooling (under substantially lower CO₂ concentrations) tends to favor higher equilibrium climate sensitivity (ECS) and makes very low ECS values less plausible.

The authors emphasize that the emergent- constraint approach is flexible and can be adapted to other time periods or climate variables once sufficient proxy data and model ensembles are available.

The paper uses tropical land proxies plus model- derived relationships to argue for relatively large LGM cooling (centered around 5.4- 5.9°C globally), placing the LGM toward the colder end of recent estimates and providing a transferable methodological framework for paleoclimate reconstruction.

Published: Geophysical Research Letters

DOI: 10.1029/2026gl123083

Authors: Alicia Hou and colleagues (Thibaut Caley, Thomas Extier, Manuel Chevalier, Mathilde Brugère, Didier Swingedouw, Masa Kageyama, and Pascale Braconnot)

Abstract

Reconstructing global temperatures from sparse proxy data remains a key uncertainty in paleoclimate science. Here, we apply an emergent constraint framework that combines statistical relationships from paleoclimate model simulations with proxy-based estimates to infer unknown climate variables. We compiled a new multiproxy data set of tropical land temperature reconstructions for the Last Glacial Maximum (LGM), which implies tropical terrestrial cooling of −4.2°C (−5.0 to −3.5°C, 95% Confidence Interval, CI). Applying our emergent constraint framework to this terrestrial temperature reconstruction, we derived updated estimates of tropical sea surface temperature (−2.6°C; −3.3 to −2.0°C, 95% CI) and global mean surface temperature (−5.4°C; −6.5 to −4.4°C, 95% CI). These observation-based estimates support LGM cooling toward the colder end of published ranges, implying that very low equilibrium climate sensitivity (ECS < ∼3°C) is unlikely. Our method is adaptable for different time periods and climate variables, providing a flexible framework for integrating proxy and model information.