The Isotope Archive: How Paleoclimate Science is Narrowing the Gap in Climate Projections
By analysing high-resolution proxies and deep-time isotopes, researchers are uncovering the Earth’s true response to rising greenhouse gas levels.
The world is rapidly moving into an atmospheric state the we humans have never seen, have not evolved for, and have not built our societies to withstand. A big part of our current trepidation and confusion is knowing what to expect, even within 10 years, let alone by 2050, 2100 or beyond. To predict the future, we need to interrogate the past. When and why have temperatures spiked in the past? What can we learn from past interglacials, what was the world like the last time CO2 was more than 425ppm?
Paleoclimatology is the area of science trying to answer these questions using a range of fascinating techniques to shine a light on our current predicament.
The Recent Past: High-Resolution Proxies
Humans have always been fascinated by the weather, and for good reason. We totally rely on it to feed ourselves. From knowing when fruits and berries will ripen, to growing crops, raising livestock, sighting fields and navigating ocean supply routes, weather has always been critical to society. As such we have measured, gauged, interpreted, even prayed to the weather all around the world for thousands of years.
Precise weather data only stretches back 150 years or so though, so how do we know what the pre-industrial was like - the baseline against which we measure our current global warming and climate changes?
One of the key technologies is Dendroclimatology, the study of tree rings. Every year, trees go through a growth cycle, speeding up in the spring and summer and slowing down in the autumn and winter. The growth rate is moderated by the available moisture and the temperature. Tree rings record these annual growth cycles in both width and wood density and so through measuring the rings, assumptions about the moisture and temperature can be extracted.
The subject has collected a vast set of data from all over the world. Modern techniques can take cores from living trees without felling them, so a control set of data can be constructed comparing detailed modern weather records to the last 100 or so tree rings. This provides accurate calibration of the method and confidence in the results. Having calibrated the approach, tree rings from past centuries and millennia can be interrogated for moisture and temperature readings. This proxy data allows temperature records to stretch back over 2,000 years. In fact the coffin wood from a soldier that was buried in China nearly 2,000 years ago was recently analysed and it was discovered that from 270 B.C.E. to 77 C.E. when the wood was growing, average humidity levels were 18% to 34% higher in the region than they are today.1
Using dendroclimatology, a recent study of trees in Spain provided evidence that extreme weather events have increased in the last decade compared to the last 520 years.2
For increased accuracy and reliability new techniques of X-ray densitometry and X-ray fluorescence provide excellent ring boundary sensitivity and therefore high precision temperature comparisons.
Another method of collecting data on rainfall patterns and monsoons is to examine stalagmites and stalactites within cave systems. Oxygen isotope analysis of the deposits formed as water drips through caves, coupled with radio-isotope aging techniques can provide detailed timelines of climate patterns stretching back to the last ice age. The timing of glaciations and permafrost melts can be calculated this way as well.
A study published last year examined deposits from caves in Siberia. No water flows currently as the ground above is frozen permafrost, but by dating the speleothems found in the caves, it was determined that the last time water dripped into the cave was 8.7 million years ago in the Miocene. This was therefore the last time the permafrost above was completely melted.3
The resolution from these studies is very impressive. In a study published in January, a stalagmite from Heshang Cave in the middle Yangtze Valley was analysed to create a precisely dated “rainfall yearbook” for a 1,000 year period centred on 2,600 BCE.4 Their reconstruction showed that the valley experienced three low-rainfall intervals (less than 700 mm of rain per year) which lasted between 40 and 150 years, and two high-rainfall intervals (more than 1,000 mm per year) which lasted 80 and 140 years respectively. Comparing this to archaeological data from the region revealed that these high-rainfall periods were associated with increased flooding, widespread wetland expansion, and a significant decline in population within the valley, linked to the Shijiahe civilisation decline.
This type of “Hydro-Climatology” is the new frontier for high-resolution proxies, moving beyond just “was it hot?” to “was it habitable?”
A key part of gaining confidence in these proxies comes from comparing and calibrating them against each other. It is very useful therefore that scientists have multiple methods to overlay and cross-validate the derived data. International global networks such as PAGES (Past Global Changes) help scientists with this process.5
The Quaternary: The Ice Age Cycles
Creating the temperature records and analysing the constituents of the atmosphere back to 2.6 million years ago requires a different set of samples, analysis and proxies. The first tool is astronomy. Planetary mechanics are now so well understood that the tiny wobbles in Earth’s orbit can be played back over millions of years to calculate the precise angles, distances and therefore incoming solar radiation, for all latitudes of the planet, at any time in the past.
From these Milankovitch Cycles, the timing of glaciations and interglacials can not only be calculated, but their relative effects in each hemisphere can be deduced. This provides a baseline for the examination of physical evidence of the ice age progression.
Two principle methods are used to interrogate the ice age and the how the Earth breathes between glacial and interglacial cycles. Ice and ocean sediment core analysis.
Ice core analysis provides details of atmospheric composition stretching back hundreds of thousands to millions of years. The Vostok core from Antarctica is perhaps the best known. There are actually 7 cores drilled between 1981 and 1996 to a maximum depth of 3,623m, however the section below 3,350 is refrozen lake ice so not useful for a chronology. The ice contains tiny trapped air bubbles which are literally samples of the atmosphere from the time the snow fell on the surface. There have been analysed for their carbon dioxide and methane content, as well as the density of dust particles in the ice. Chemical and isotope analysis provides dating and temperature proxies.
Not surprisingly the temperature profile aligns completely with the Earth’s orbital patterns with interglacials and glacial maxima occurring exactly when we’d expect them to. Obviously the temperature readings are related to the site in Antarctica so there would be regional variations (including polar amplification) but other proxies from around the world align well, so confidence in the timing and relative temperatures are very high.
Ocean sediment cores take a slightly different approach, but provide the best thing we have to a global thermometer for the last 66 million years. Cores are collected by specialised vessels that can extract core samples even several miles below the ocean surface. Advanced Piston Coring for unconsolidated, soft sediments and rotary drilling for older rock are used. Because gaps occur when the ship switches core barrels, they drill multiple “holes” side-by-side and “splice” them together using magnetic susceptibility and density scans to create one continuous, unbroken timeline.
Once the mud and rock cores are retrieved, the search for Foraminifera begins. These are the tiny plankton organisms whose shells, made from calcium carbonate, hold the isotope data. The particular isotopes of interest are those of Oxygen, Carbon and often Boron. Sections of the core are washed and sieved leaving what looks like white sand but is actually thousands of tiny calcium carbonate shells. Specimens of the target species are selected than analysed using gas source mass spectrometry. This separates and measures the relative amounts of isotopes for each of the atoms in question.
The reason for all this effort is that the ratios of isotopes show the surface temperatures of the ocean and the global ice volume levels back through time. The results from hundreds of cores led to the creation of a Benthic δ18O stack (δ referring to the ratio of isotopes). When ice sheets grow they trap the lighter 16O isotope on land, which due to its lightness was more easily evaporated from the ocean surface. This leaves the ocean enriched with the heavier 18O. The tiny plankton don’t care what they use to construct their calcium carbonate shells, so by examining the proportion of isotopes in their shells, scientists can deduce the relative extent of the ice sheets and hence glaciation at the time. Timing is validated through sediment dating and cross validated with magnetic pole switches which are detectable in the core samples and can be matched with the known chronology from seabed spreading on either side of continental rifts, such as the very well studied mid-Atlantic Ridge.
Because sediment cores have been drilled all over the world, a more detailed picture of ocean temperature and glacial extent can be extracted leading to global average condition analysis rather than isolated ice covered regions alone. However, as you might expect, there is very close alignment providing the cross correlation that inspires confidence in the findings. This includes geological evidence of ice sheet extent such as glacial tongue till deposits, even ice rafted debris profiles.
The Benthic δ18O stack is so well defined that it has become the standard for labelling the glacial cycles of the Pleistocene. Marine Isotope Stages (MIS) identify periods of glaciation and deglaciation, and are numbered going back in time. Odd numbers denote interglacials like the Holocene, which is MIS-1, and even numbers denote glaciations. The chart below shows the MIS points over the last 800,000 years. MIS-2 is the Last Glacial Maximum. MIS-3 and MIS-4 are intermediate stages (stadials and interstadials) where the ice sheet shrank then re-grew on shorter timescales. MIS-5 is a succession of warmish periods, with MIS-5e covering the Eemian, the last major interglacial 130,000 years ago (incidentally the last time mean annual temperatures were as warm as 2024). MIS-11 stands out as the longest lasting interglacial.
Combining these proxy records provides a detailed history of global temperature, ice extent and importantly atmospheric composition and CO2 concentrations for the last 2.6 million years. The relationship between temperature and greenhouse gas concentrations can be examined and estimates of climate sensitivity extracted.
Climate Sensitivity is the single most important value in predicting our future climate. It is the amount of global warming experienced when CO2 levels are doubled. Due to the logarithmic relationship, it is the same temperature rise between 200 and 400ppm as 1,000 and 2,000ppm. Ice bubble analysis shows that pre-industrial it was 278ppm. Today it is over 425ppm, so climate sensitivity can tell us what we can expect when the climate settles into a new equilibrium. The IPCC has a range between 2° and 5°C. Whether the actual value is low or high makes a massive difference for the coming years, decades and centuries. Most paleoclimate work suggests a value of 4° to 4.5°suggesting more rapid warming than many expect.
Deeper into the Cenozoic
So, ice core bubbles and seabed sediment cores tell us a great deal about what the Earth was like in previous interglacials, some of which like the Eemian (MIS-5e) and the Hoxnian (MIS-11) provide information on sea level rise and ice sheet state at the temperatures were are now reaching (+1.5°C). However CO2 levels never rose above 300ppm during the whole of the Pleistocene. This shows us that the climate is far from stable and is hunting for a new warmer equilibrium.
To see what the climate was like the last time levels were as high as they are today (425ppm) we need to go all the way back to the Miocene, 5.3m to 23m years ago.
To discover the CO2 levels in the deep past, scientists turn to trees again, not wood this time, but the leaves. Tree leaves have a series of pores or Stomata on their underside that they open and close to regulate the flow of air. These stomata are used to draw in CO2, but also to transpire out water vapour. To maintain efficiency, plants adapt the number of stomata over generations depending on the CO2 concentration. In times of high concentration, they reduce the number to avoid excess water loss. In times of low concentration, they are forced to adapt to having more stomata to collect the CO2 they need for photosynthesis, despite the increased water loss. There is therefore a Stomatal Index where scientists can back-calculate atmospheric CO2 by counting the stomata on fossilised leaves like the one shown below.
Fossil dating follows the traditional approach of geological stratification analysis as well as cross validation with other species and radio-isotope dating where possible.
Another biological proxy method for the Cenozoic era returns to marine sediment cores. Boron Isotopes (δ11B) found in marine shells are a good indicator of the ocean pH at the time the sea creatures were growing. Since ocean acidity is linked to atmospheric CO2 levels, this provides a chemical double-check for the stomatal data of the same age.
Shell carbonate dating is still used for these older cores examining the oxygen isotopes in carbonate based shells. To increase accuracy, scientists are turning to Clumped Isotope Thermometry. This avoids the fact that the ratio of heavy oxygen can be affected by both temperature and the salinity of the water the creature lived in. Instead they look at how often 13C and 18O “clump” together in a carbonate molecule, which is salinity agnostic.
The Miocene Climatic Optimum: Steady state at today’s CO2 concentrations
Using these techniques, scientists can be confident that the last time CO2 levels were in the 400-500ppm range, as they are today, was during the mid Miocene (~15 million years ago). With this knowledge, geology can tell us what conditions were like.
It’s important to note that this period represents the climate in equilibrium at those higher CO2 concentrations. The ice sheets were at the appropriate size for the temperature and the Earth’s energy imbalance would have been neutral. The climate would have been stable and the weather consistent, albeit very different from what we are used to.
The key differences were that temperatures were about 4°C above pre-industrial. Arctic Amplification meant that the poles were considerably warmer than today. Geological records show that forests stretched much further north than they do now, way up into the high Arctic. The temperature difference between the poles and equator were much less with major differences to the engine of global winds and ocean currents.
Because the poles were so much warmer, much of the ice we have today, simply didn’t exist. As described earlier, Siberia was permafrost free with water dripping through the cave systems. As a result, sea levels were between 10m and 40m higher than they are now. Greenland was ice free and much of the land masses of the world were grasslands. We know this since this is the period when large numbers of grazing species evolved such as horses and camels. The West Antarctic ice sheet also didn’t exist while the East Antarctic ice sheet was significantly reduced, probably just 25% of today’s ice volume.
The significance of the Miocene is the fact that it was +4°C warmer with CO2 levels similar to ours. This suggests that the Earth System Sensitivity (ESS) - how much we warm in the long run, might be higher than current short-term models predict and at the higher end of the IPCC estimate range. It also shows that these levels are sufficient to drive warming past a host of tipping points including the melting of both Greenland and the West Antarctic Ice Sheet. This is the long game lesson of the Miocene.
A Hothouse Spike: The Palaeocene-Eocene Thermal Maximum (PETM)
Whilst the Miocene can tell us a lot about what our climate might be like when CO2 levels have stabilised around 500ppm (which they will do unless we rapidly decarbonise AND engage in massive carbon dioxide removal campaigns to clean up the atmosphere), it can’t tell us much about the process of rapid climate change. The Miocene was a period of gradual warming and cooling over millions of years. To see what rapid warming looks like we have to go a little deeper into the past.
56 million years ago, the Earth experienced a warming spike that lasted 200,000 years. It’s only just visible in the chart above. Temperatures spiked by 5° to 8°C higher than previous over a very short geologic period, just 20,000 years. CO2 levels jumped from 800-1,000ppm to 2,000-2,500ppm. This period marks the transition between the Palaeocene and the Eocene and is known as the Palaeocene-Eocene Thermal Maximum (PETM). It was believed to have been caused by huge carbon emissions from the large igneous province created as the North Atlantic ripped apart, together with methane clathrates instability as the climate warmed.
Isotope analysis of the sediment cores reveals a massive dip in the ratio of 13C to 12C found in marine and terrestrial sediments. This is known as a Negative Carbon Isotope Excursion. Because plants and organic matter prefer 12C (the “lighter” carbon), a sudden surge of 12C in the global record means a massive amount of organic carbon was suddenly released into the ocean and atmosphere.
From an atmospheric perspective this is an increase in CO2 of 2.5x. So far, since the pre-industrial, we have only managed 1.53x. However in terms of emission rates, we are currently adding CO2 to the atmosphere 9-10 times faster than during the onset of the PETM.6 We are also warming much faster as a consequence.
There are lots of lessons from this period. In terms of timing, it’s informative that the carbon pulse lasted 5,000 years, but warming carried on for a further 15,000 years. It took the additional time for the equilibrium to be reached. It’s also sobering that it then took a further 200,000 years for the climate to recover to it’s previous stable state.
Ocean acidification has been detected in sediment cores which show a sudden transition from white calcium carbonate shell deposits to a dark, gritty red clay, before eventually turning white again. As CO2 flooded the ocean, it turned the water acidic, raising the Calcium Carbonate Compensation Depth. Essentially, the ocean became so acidic that the shells of marine organisms dissolved before they could settle on the seafloor.
Ocean Acidification is one of the nine safe Planetary Boundaries and was announced to have been breached in 2025. While the PETM tells us that the ocean will acidify, our current rate of change is so fast that marine ecosystems have even less time to adapt than they did 56 million years ago, making the modern ‘Red Clay’ layer a potentially more permanent scar in the geological record.
The PETM is also associated with an extinction event as the rapid changes were too fast for many species to cope with. Those that could migrate to higher latitudes, evaluations or suitable refuges survived but many didn’t. Some movements were extreme, for example crocodile fossils have been found above the Arctic Circle from the peak of the PETM. A recent study of plant ecosystems found that adaptation coped with the initial temperature rise, but after 4°C, they couldn’t adapt and productivity dropped dramatically. As a result they soaked up less carbon dioxide from the atmosphere, the feedback mechanism delaying carbon sequestration and the recovery of temperatures by 70,000 years or more.7
The PETM is the ultimate stress test for our climate models. By calculating the total mass of carbon released (the “forcing”) and measuring the resulting temperature rise (the “response”), researchers can estimate Equilibrium Climate Sensitivity in a warming scenario. The Big Takeaway from this is that the PETM suggests that feedbacks like melting permafrost or changing cloud cover can amplify warming even further than CO2 alone. This is particularly important as we engage feedbacks and approach tipping points in our own climate.
The PETM shows the effect of Equilibrium Climate Sensitivity (ECS) on a fast warming system with timescales of decades and centuries. It is a story of triggered tipping points and rapid feedbacks.
Conclusion: The “Paleo-Sensitivity” Warning
The paleoclimatology work that has been carried out shows the contrast between ECS and ESS very well. The PETM shows us the violent, fast response Equilibrium Climate Sensitivity, driven by feedbacks and triggered tipping points. We are seeing these changes today in the form of changes in cloud properties, both extent and structure. These are significantly reducing Earth’s albedo, or reflectiveness, increasing the amount of energy the Earth absorbs from the Sun and is accelerating warming. We are seeing extinction levels far higher than background and migration of plants and animals chasing suitable climates into higher latitudes.
The Miocene shows us the power of the slow response Earth System Sensitivity and is the long-game running on millennia timescales. It shows us that CO2 levels of 500ppm eventually led to massive sea-level rise and a nearly ice-free world. This is where we are driving the climate long term. By 2100, our CO2 will match the Miocene, but our temperatures likely won’t—yet. The Miocene reminds us that there is committed warming in the system. Even if we stop emissions by 2100, the Earth may continue to drift toward a Miocene-like state of much higher sea levels over the following centuries. Today there is much more warming in the pipeline than has emerged so far.
Paleoclimate research turns the ‘uncertainty’ of climate models into ‘certainty’ from Earth’s history. The Miocene proves that 400-500ppm of CO2 is enough to melt most of the world’s ice; the PETM proves that carbon spikes lead to mass extinctions and oceanic collapse. We aren’t guessing what will happen; we are reading the transcript of what has already occurred.
For a little more context of the deep past, you might also like this article:
Marín-Martín, M.,et al.: A five-century tree-ring record from Spain reveals recent intensification of western Mediterranean precipitation extremes, Clim. Past, 21, 2205–2223, https://doi.org/10.5194/cp-21-2205-2025, 2025.
Vaks, A., Mason, A., Breitenbach, S.F.M. et al. Arctic speleothems reveal nearly permafrost-free Northern Hemisphere in the late Miocene. Nat Commun 16, 5483 (2025). https://doi.org/10.1038/s41467-025-60381-5
Jin Liao, Christopher C Day, Chaoyong Hu, Yuhui Liu, Gideon M Henderson, Precise chronology of hydrological changes at ∼4.2 kyr in Central China to assess the impact of flooding on Neolithic societies, National Science Review, Volume 13, Issue 2, January 2026, nwaf567, https://doi.org/10.1093/nsr/nwaf567
PAGES, Past Global Changes: https://pastglobalchanges.org/
Philip D. Gingerich, Temporal Scaling of Carbon Emission and Accumulation Rates: Modern Anthropogenic Emissions Compared to Estimates of PETM Onset Accumulation, 2019 https://doi.org/10.1029/2018PA003379
Rogger, J., Korasidis, V.A., Bowen, G.J. et al. Loss of vegetation functions during the Paleocene–Eocene Thermal Maximum. Nat Commun 16, 11369 (2025). https://doi.org/10.1038/s41467-025-66390-8
All one needs to do is research "the Great Dying." The worst of the 5 previous mass extinctions. It was CO2 caused.
Hi Tom, do you think that because we have removed some of our ability to soak up our emissions and in part created a higher albedo world through land clearing this may mean that the higher co2 values we have today correlate to lower co2 conditions of the past? I would assume that a greater percentage of the overall equalization would take longer to be felt like Greenland melting and associated albedo changes so although we may be feeling some of the effects the feedback lag will be longer . Many thanks for the article.