Mapped: How ‘proxy’ data reveals the climate of the Earth’s distant past (Carbon Brief)


    At any one moment in time, thousands of measurements are being taken of the world’s weather. Across land, sea and sky, data is being gathered manually and automatically using a range of technologies, from the humble thermometer to the latest multi million-pound satellite.

    By Robert McSweeney and Zeke Hausfather. Design by Tom Prater.

    Put together over many years, these measurements provide a record of the Earth’s climate and how it is changing.

    But even the world’s longest climate archive – the central England temperature record – only goes back to 1659. This is a mere snapshot in time considering the hundreds of thousands of years that humans have roamed the planet.

    Fortunately, the Earth has been keeping its own records. Tucked away in an assortment of unlikely places – from shells and stalactites to pollen and seal pelts – the natural world has recorded the ebb and flow of the climate for millions of years.

    This is known as “proxy data” – indirect records of climate imprinted on different parts of the biosphere.

    In the same way that something “prehistoric” relates to a time before written history, proxy data provides an insight into the climate before dedicated records. It forms a fundamental part of the study of past climates, known as “palaeoclimatology”, while also helping underpin scientists’ understanding of how the climate will change in the future.

    In this in-depth Q&A, Carbon Brief explores what proxy data is, the different types, how scientists draw climate data from them, and what they can tell us about the Earth’s climate in the past, present and future.

    In addition, Carbon Brief has produced an interactive map of the US National Oceanic and Atmospheric Administration (NOAA) archive of more than 10,000 proxy datasets.

    What is proxy data?

    In 1714, German physicist Daniel Gabriel Fahrenheit invented what is considered to be the first example of the modern thermometer. It enclosed mercury in a glass tube and had a standardised scale running up the side. A decade later, he would add the temperature scale that bears his name. (Swedish astronomer Anders Celsius would not devise his alternative scale for another two decades.)

    The thermometer, along with other instruments such as the barometer for measuring air pressure and hygrometer for humidity, went on to become a key part of the formal weather station. These stations – shielded behind the shutters of a Stevenson screen – were first installed in Europe and the US in the 1800s, and spread around the world throughout the century and beyond.

    By the middle of the 19th century, there were sufficient weather stations and enough observations being recorded on land – and sea – to produce a reliable measurement of global temperature. The longest record of global temperatureproduced jointly by the UK Met Office Hadley Centre and the University of East Anglia’s Climatic Research Unit – begins in 1850. Others, such as those produced by NASA and the National Oceanic and Atmospheric Administration (NOAA), start in 1880.

    Annual global average surface temperatures from 1850-2020. Data from NASA GISTEMPNOAA GlobalTempHadley/UEA HadCRUT5Berkeley Earth and Carbon Brief’s raw temperature record. 1979-2000 temperatures from Copernicus ERA5 (as the reanalysis record starts in 1979). Anomalies plotted with respect to a 1880-1899 baseline to show warming since the preindustrial period.

    This means that scientists have a robust account of how global temperatures have changed over the past century and a half, or so. But, of course, the Earth is much older than that.

    To look back even further – and for places that did not have instrumentation until relatively recently – scientists need to cast their eyes beyond direct observations to indirect evidence that is locked up in various forms across the Earth. This is “proxy” data.

    The word “proxy” is typically defined as an intermediary or substitute – often in reference to a person given the authority to vote or speak on behalf of someone else. Proxy data, therefore, is information that is a substitute for direct observations of the Earth’s climate.

    “A climate proxy is something we use to reconstruct variations of climatically relevant factors in the past, such as temperature, precipitation, CO2 levels – or whatever else is of interest,” explains Prof Paul Pearson from the School of Earth and Ocean Sciences at Cardiff University. He tells Carbon Brief:

    “Obviously, these things can’t be measured directly without a time machine, so we need to find something that survives from the past that is dateable and contains something we can measure that would have responded to the variable we are interested in – hence, the name ‘proxy’.”

    Scientists, therefore, look for ways that the climate “has left a mark in the environment”, says Dr Maisa Rojas, associate professor in the Department of Geophysics at the University of Chile and a lead author on the palaeoclimate chapter (pdf) of the Intergovernmental Panel on Climate Change’s (IPCC) fifth assessment report (AR5). She tells Carbon Brief:

    “The living part of our world – the biosphere – responds to the climate and, as such, it leaves marks in a number of environmental indicators that we can then use to reconstruct back the climate.”

    These clues to past climate are scattered across the Earth, from the layers of vast ice sheets and sediments at the bottom of lakes to rings of tree growth and towering stalagmites in caves. (See later section for more on the different sources of proxy data.)

    This information allows scientists to “study the climate over the past centuries and millennia and, thus, to go further back in time than by using instrumental climate data alone”, explains Dr Valerie Trouet, an associate professor in the Laboratory of Tree-Ring Research at the University of Arizona and author of Tree Story, a book about tree-rings. She adds:

    “By studying the climate prior to the 20th century, when meteorological station data become available, we can put current climate change in a longer-term context and study natural, non-anthropogenically driven, climate variability.”

    The way that the climate can leave its mark on the Earth’s surface has long been observed. In the 15th century, for example, Italian artist and inventor Leonardo da Vinci documented that the thickness of tree-rings – the concentric circles found running through a tree’s trunk – varied with rainfall.

    The scientific discipline of tree-ring dating – known as “dendrochronology” – was later pioneered by American astronomer A E Douglass in the early 20th century. His research attempted to connect the pattern of sunspot cycles with fluctuations in climate and tree-ring patterns. From this earliest work, Douglass went on to found the Laboratory of Tree-Ring Research mentioned above.

    Another proxy with a long history is “the oxygen isotope composition of calcite shells” in marine organisms, says Pearson:

    “This method was pioneered by [American chemist] Harold Urey in the immediate post-war years and helped launch the entire field of palaeoclimatology.”

    Urey showed that the chemical composition of these shells (see the next section for more on isotopes) varied depending on the temperature of the water. Extracting this information, thus provided information on the climate when the organisms were alive – going back many millions of years.

    Urey described his discovery as “suddenly [finding] myself with a geologic thermometer in my hands”, explains Pearson.

    Where in the world is proxy data found?

    From the ice sheets of Antarctica and the seabed of the Atlantic, to the boreal forests of Europe and corals of southeast Asia, proxy data is found across the Earth’s land and ocean.

    NOAA holds an archive of more than 10,000 proxy datasets covering more than a dozen categories. With its permission, Carbon Brief has mapped this data.

    Use the categories in the legend on the left to select a particular proxy or archive type, and the buttons in the top-right hand corner to zoom in and out. Clicking on an individual data point will reveal the period covered by the data, the site name and a link to NOAA’s reference webpage for further information.

    What information do proxies capture?

    Proxy data can provide insights on a range of climate-relevant changes. These include sudden events – such as volcanic eruptions or floods – and gradual, long-term trends – such as warming and cooling, drought, changing sea levels, cyclone patterns, monsoon seasons, fluctuating atmospheric CO2 or thinning ice sheets.

    Proxies generally falls into one of three categories – physical, biological or chemical – explains Prof Pearson:

    “Proxies can be something physical like the amount of silt in sea floor mud which can be a proxy for the speed of the current, something biological like the width of a tree-ring or growth band in a marine shell, or something chemical like the elemental or isotopic composition of a substance that we can measure in the lab.”

    These different types of proxies have been captured by an array of palaeoclimate “archives”, such as sediments, ice cores and cave formations. These are the “media that the [proxy] data is being recorded in”, explains Allison Cluett, a PhD candidate at the University at Buffalo.

    For example, Cluett’s research analyses leaf waxes (the proxy) in ocean sediments (the archive) to reconstruct the climate of southern Greenland.

    Each type of proxy is reflecting a change in conditions, but they are not simply capturing temperature or rainfall or some other single variable. Instead, they often reflect a combination of several. Thus, the field of palaeoclimatology involves “disentangling” specific climate information, says Dr Rojas.


    Take tree-rings, which provide a biological record of how a tree has added new layers of wood over time. Each ring comprises a light and dark part – with the pale part signifying the fast growth of spring and early summer and the dark part indicating the slower growth of late summer and autumn. Taken together, each ring indicates a year of a tree’s life.

    Scientists can “core” a tree to extract a cross-section through its trunk. This allows them to analyse its rings without damaging the tree.

    The rate of tree growth can “respond to both precipitation and temperature – and will depend on the tree and where it is located”, explains Rojas. Rings will be wider in warm, wet years where the tree is getting sufficient sunshine and rainfall to support growth, and rings will be narrower during a drought or if the tree is hit by pests, disease or fire.

    It is, therefore, the job of scientists to draw out the climate data from the information the proxy provides. In the case of tree-rings, this will first involve “cross-dating” rings between a number of trees to identify the correct year for each ring.

    Then, using records of local weather data, scientists can calibrate the rings against an observed climate record. Simpler relationships can be calibrated using a straightforward equation, but scientists use models for those that are more complex.

    If the two data sources are well matched, the tree-rings can be used to cast further back – before the observed record began – to analyse the climate during the tree’s full lifetime. (See later section for more on calibrating proxy data.)

    Tree-ring records can go back a very long way, notes Dr Trouet:

    “The longest continuous tree-ring record – that includes a measurement for each year – is the German oak-pine chronology that dates back to 10,461BC…But for palaeoclimate purposes, tree-rings are mostly used to study the past ~500 to 2,000 years.”

    And there are subtleties around which locations are best suited to certain analyses, explains Trouet:

    “Trees grow a lot – and form wide rings – under favorable climate conditions. These can be wet conditions in dry regions, such as the American southwest, or warm conditions in cold regions, such as the European Alps or Scandinavia…To reconstruct past temperature, we use tree-rings from cold regions. To reconstruct past drought conditions, we use tree-rings from dry regions.”

    There are other complications, of course. Rings are most prominent in trees that experience clearly defined seasons throughout the year. This means that “trees in midlatitudes are more responsive to climate than trees in the tropics”, says Rojas:

    “So Europe is good, North America, northern Asia, South America as well – and along the Andes there are lots of trees you can use.”

    In contrast, tropical trees are more of a challenge for dendrochronology, although some species do still form annual rings.

    This can be seen in the map above – the majority of tree-ring data comes from the temperate and boreal forests of the northern hemisphere. Tree-ring data is not impossible in the tropics, adds Trouet – there are hundreds of records available and there is “definitely potential for more”.


    Moving onto chemical proxies, one example is isotopes. And that provides an opportunity to talk about one of the most well-known types of palaeoclimate archive – ice cores.

    These cylinders of ice are drilled out of ice sheets and glaciers, and can run to several kilometres. The Earth’s ice sheets and glaciers are built up from snowfall over thousands of years, with each layer – compacted over time – trapping tiny bubbles of air. Taking a cross-section through the ice thus provides a timeline of these air bubbles over centuries.

    The air bubbles are tiny samples of the atmosphere through the life of the ice sheet or glacier.

    While scientists can analyse the bubbles directly to ascertain the makeup of the atmosphere in the distant past, they also contain a proxy to estimate past temperatures – the oxygen isotope “18O”.

    Isotopes are forms of the same element that are identical except for a different number of neutrons within the nucleus of the atom. The most abundant oxygen isotope is 16O, which has eight neutrons, giving it an overall atomic mass of 16 (eight neutrons plus eight protons).

    18O has an extra two neutrons, giving it an atomic mass of 18. As a result, atoms of 18O are very slightly heavier than 16O. This weight difference has implications when water is evaporated from the oceans and falls as snow at the Earth’s poles, explains Dr Robert Mulvaney, a glaciologist at the British Antarctic Survey, in an article for Scientific American:

    “Simply put, it takes more energy to evaporate the water molecules containing a heavy isotope from the surface of the ocean, and, as the moist air is transported polewards and cools, the water molecules containing heavier isotopes are preferentially lost in precipitation.”

    Both of these processes are temperature dependent, says Mulvaney, which means that measurements of 18O in ice cores can tell scientists how warm the climate was at that time in the past.

    Research into oxygen isotopes – much of it involving marine sediments – “was fundamental to a great discovery of 20th century science that the ice ages are paced by the Earth’s orbital variables of eccentricity, axial obliquity and precession”, says Pearson.

    These orbital variables – known as “Milankovitch cycles” after Serbian scientist Milutin Milankovitch who developed the theory – describe how a collection of variations in the Earth’s position relative to the sun can trigger both the beginning and end of ice ages.


    Finally, for the last of the three categories, an example of a physical proxy is marine and lake sediments.

    Each year, billions of tonnes of sediment are washed into lakes and seas around the world. These sediments build up over time, adding layer upon layer with each year that passes. Drilling a core down through the bed of a sea or lake can, therefore, provide a timeline of how the sediments – and, hence, the climate – have changed.

    The size, shape, structure and colour of these sediments can all provide clues about the climate of the time. For example, explains the US Geological Survey website:

    “Scientists use the size and shape of sediment particles to determine if the sediment was transported, how far it was transported, and how energetic the environment of transportation was (for example, waves crashing on a beach leave behind coarse sand particles, whereas very small grains are deposited in very still conditions).”

    But sediments are also a very important archive for other proxies. Buried along with layers of sediment are all sorts of fossils that scientists can analyse. “Foraminifera” are a classic example, explains Pearson:

    “Foraminifera are – mostly – microscopic shells secreted by single celled organisms that live as plankton or on the sea bed. In the right burial conditions, their shells can survive in virtually perfect condition indefinitely. These shells build up slowly on the seafloor producing a more or less continuous proxy record.”

    Foraminifera build their shells from calcium carbonate extracted from the seawater. Isotope analysis can reveal the conditions in the ocean – and, hence, the climate – when those organisms were alive.

    Diatoms are another microfossil – this time with shells made of silicon dioxide. While foraminifera are restricted to marine and coastal environments, diatoms are also found in inland lakes. Lake sediments are a key natural archive for reconstructing histories of drought and the diatoms they contain have been used, for example, to piece together records of extreme droughts in the US midwest.

    Leaf waxes found in the marine sediments are also a useful climate proxy, adds Cluett:

    “Leaf waxes are a group of simple organic molecules that are widely produced by vegetation, both on the terrestrial landscape and within lakes… [They] are useful biomarkers because plants incorporate hydrogen atoms into the structure of these molecules from the water – generally derived from precipitation – in which they use to grow.”

    Isotope analysis of leaf waxes “provide terrestrial climate records analogous to measurements of stable water isotopes in ice cores”, notes Cluett. This approach has been used, for example, to reconstruct rainfall patterns during the “Green Sahara” period around 11,000-5,000 years ago when the region supported diverse vegetation, permanent lakes and human populations.

    As proxy data is accumulated naturally, its records can extend back as far as that medium exists. So for the isotopes in ice cores, for example, that is as long as the ice sheet or glacier has been in place. Marine sedimentary records can be millions of years long, going all the way back to the Cretaceous period 100m years ago – the time of the dinosaurs. This reflects the fact that the seabed has existed for a lot longer than trees, corals or even ice sheets.

    To probe the oldest parts of Earth history, palaeoclimatologists must use rock formations, explains Dr Jessica Tierney, an associate professor at the University of Arizona and a lead author on the IPCC’s sixth assessment report. She tells Carbon Brief:

    “To study climate changes before about 100m years ago, we must work on rock formations on land, which contain marine or terrestrial sediments that have been lithified.”

    Lithification is the process by which sediments are compacted under pressure to form solid rock. These might naturally “outcrop” on the landscape, says Tierney, or scientists might drill into them to get a core. She adds:

    “In these ancient archives, we find evidence of truly extreme climate changes, like the end-Permian global warming and mass extinction, and ‘Snowball Earth’ – a time when the Earth was totally covered in ice.”

    While the potential record from marine sediments is long, the sampling “interval” that can be derived is more limited. The data might only be able to show climate changes from one century to the next, while data from tree-rings and stalagmites, for example, can show changes from one year to the next. (See the section below for more information on different types of proxies and archives.)

    Proxy data can help provide an insight into the way humans have responded to changes in their environment. For example, a Science Advances paper from earlier this year used pollen and charcoal data from 17 sediment cores – along with archeological surveys – to show how inhabitants of the Isles of Scilly off the south-western tip of England adapted to changing sea levels during the Bronze Age around 4-5,000 years ago.

    Finally, it is worth highlighting another form of palaeoclimate archive – historical documents. These can be diaries, logbooks, photographs and even paintings that carry direct and indirect climate information.

    For example, English manorial accounts from the Middle Ages – financial and farming records kept by rural estates – provide detailed information on harvests and milk production. Scientists have used these records to piece together records of droughts that hit England hundreds of years ago.

    Another example is how historical photographs and sketched maps can be used to reconstruct changes in the lengths of glaciers – and, hence, fluctuations in the climate.

    Other forms include weather descriptions in personal diaries, records of grape harvest dates, and descriptions of wind, weather and sea ice cover in ship logbooks.

    What are the different sources of proxy data?

    The table below summarises the key archives of climate proxies, the data they provide and the typical intervals and time spans that they cover.

    Proxy Archive Types of measurements Typical interval Typical time span Description
    Boreholes Temperature Century Hundreds of years Boreholes are narrow shafts drilled into the Earth, typically to extract substances such as water or oil. As heat at the surface slowly diffuses vertically down into the Earth, temperature readings taken a different depths through the borehold can indicate past temperatures at the surface. While borehole measurements are taken directly, they are classed as a proxy as they are used to indirectly measure past temperatures.
    Corals and sponges Isotopes, chemical properties, growth rate Year Centuries Corals build their hard skeletons from calcium carbonate they extract from seawater. The density of those skeletons changes from season to season and year to year with fluctuations in sea temperature, water clarity and available nutrients. These variations are revealed in annual growth rings similar to those in trees. Scientists take small samples from the corals to analyse these rings, which sometimes need x⁠–rays to identify. Isotope analysis of the oxygen atoms contained in the skeleton can also signal changes in variables such as ocean temperature. Although sponges do not build hard exoskeletons like corals, they grow by putting on layers of calcium carbonate or silicon dioxide, which also produces growth rings.
    Glaciers Glacier extent Year Hundreds of years Mountain glaciers grow and retreat over time in response to the climatic conditions and so records of their length can be used as a climate proxy. Records ⁠– in the form of measurements, photos and paintings ⁠– often go back several hundred years. Carbon dating of plants and other organic material that is uncovered by a retreating glacier can also indicate past glacier extent.
    Historical documents Historical Hour to day Hundreds of years Direct and indirect information about the climate can be gleaned from historical documents. These include accounts of weather in newspapers, ship logs, personal diaries and church records, while documented harvest dates ⁠– for grapes and other crops, for example ⁠– can also indicate climatic conditions of the past. Photos, maps, charts and paintings can all be sources of data too.
    Ice cores Isotopes, dust, accumulation rate, greenhouse gas concentrations Year Hundreds of thousands of years Ice sheets and glaciers form from the accumulation and compaction of snow over thousands of years. Drilling down through the layers of ice to retrieve a «core» provides a cross⁠–section of that accumulation and hence a timeline of snow build⁠–up. The information contained in the ice includes dust from volcanic eruptions, air bubbles that provide sample of past atmospheres, and isotopes that offer evidence of past climates.
    Lake sediments Physical and chemical properties, shells, pollen, insects, molecular fossils, isotopes Decades to centuries Millions of years Sediments are washed into lakes and accumulate through time. Like marine sediments, lake sediments offer a variety of climate proxies. Carbon and hydrogen isotope analysis of molecular fossils like leaf waxes (coming from the protective coat on the leaves of plants) can be used to infer changes in the landscape and the water cycle. The fossilised remains of insects can be used to infer past climate. Grains of pollen from flowering plants are preserved in lake sediments, and can be used to infer changes in both vegetation and climate. The amount of charcoal in lake sediments can be used to infer changes in fire frequency and intensity.
    Loess Dust accumulation, physical and chemical properties, molecular fossils Centuries to millennia Millions of years The formation, transport and deposition of wind-blown silt ⁠– known as «loess» and «Eolian dust» ⁠– is closely linked to changes in climate. Large dust deposits are recorded in dry periods ⁠– for example, during the ice ages when large portions of land were covered with ice sheets and glaciers. Dust deposits on land can reach tens or hundreds of metres thick.
    Marine sediments Physical and chemical properties, shells, pollen, molecular fossils, isotopes Centuries to millennia Tens of millions of years Each year sees billions of tonnes of sediment accumulate on the beds of seas and oceans across the world. These sediments capture tiny clues to the climate of the time ⁠– including microfossils, such as foraminifera shells, and molecular fossils, such as leaf waxes. Isotope analysis of these clues can reveal information about the climate. Properties of the sediment itself ⁠– such as its size, shape, structure and colour ⁠– can also change with the climate.
    Pack rat middens Pollen, insects, plant remnants, bones, teeth, isotopes Decades Tens of thousands of years Pack rats, also known as woodrats, construct debris piles that can be crystallised over time by their urine, preserving the contents for thousands of years. These «middens» contain remnants of plants, bones, teeth, insects, shells and seeds that can be dated and analysed for their isotope content, providing climatic information.
    Rock outcrops Physical and chemical properties, shells, pollen, insects, teeth, plant fossils, molecular fossils, isotopes Millennia Hundreds of millions of years Ancient sedimentary rocks provide the oldest archives of Earth’s climate. As with marine and lake sediments, a number of different proxies can be measured in them. Fossils of plants offer an opportunity to reconstruct CO2 levels, through the analysis of their stomata. The shape and condition of the teeth of fossilised mammals can provide information on past climate conditions. For example, worn surfaces on the teeth of herbivores can indicate the vegetation ⁠– and, hence, the climate ⁠– of the past.
    Seal pelts Isotopes Decades Centuries to millennia The fur/skin of seals have been used for thousands of years to make warm and waterproof clothing and footwear. These skins contain isotopes of elements, such as carbon and nitrogen, that the seals accumulated from the prey they consumed. The concentrations of the different isotopes can, thus, indicate the structure of the food chain in the past and, hence, the environmental conditions of the time.
    Speleothems Isotopes, chemical properties Decades to centuries Tens of thousands of years Speleothems are cave formations, such as stalactites (hanging from the cave ceiling) and stalagmites (rising from the floor). They are formed from the build up of mineral deposits ⁠– primarily, calcium carbonate ⁠– carried by groundwater percolating through the rock. Changes in isotopes and trace elements can be used to determine past climates.
    Tree rings Ring width, wood density, isotopes Year Thousands of years The annual growth of a tree is typically recorded in rings through the trunk. Each ring has a light part (fast growth in spring/early summer) and a dark part (slow growth in late summer/autumn) for each year of growth. The amount of growth ⁠– and, thus, the width of the rings ⁠– reflects the climatic conditions of the time. For example, trees tend to grow faster in warm, wet conditions, and slower in cold and dry. Trees can be «cored» to remove a small cross⁠–section of the trunk in order to access the rings without damaging the tree. Isotopes within the wood can also be analysed to provde climate information.

    Table produced with the help of Dr Jessica Tierney.

    How is proxy data calibrated and used?

    As proxies do not directly measure climate variables, a conversion is needed to turn an oxygen isotope value, tree-ring width or other proxy measurement into a climate variable, such as temperature or rainfall. The conversion is known as “calibration” and usually takes one of two forms.

    Many more recent high-resolution proxies can be “calibrated-in-time”. This is where researchers look to see what the relationship is between the proxy values and direct observations during the recent period in which climate observations exist, and use that relationship to infer values in the more distant past.

    For example, if tree-ring width is closely related to temperature during the 1850-2000 period, scientists can use the tree-ring record before 1850 – say, from 1500 to 1850 – to reconstruct the temperature of that period.

    A second approach is “calibration-in-space”. This involves measuring the proxy across a wide spatial range of modern environments, where the controlling factors – such as temperature – are known. This technique is used when direct comparison with observational records is not possible.

    For example, different kinds of pollen might be measured in modern lake sediments that span a range of temperatures in order to produce a calibration. In some cases, calibrations can also be made in the lab, by culturing organisms – such as foraminifera or algae – under different temperatures.

    In some cases, however, more than one factor can affect proxy estimates. For example, tree-ring width may depend both on rainfall and temperature for a location, and researchers want to ensure that they do not misinterpret a drought as a period of low temperature. Researchers may make use of sophisticated statistical models to distinguish between different factors affecting proxy measurements.

    In addition, the relationship between the proxy value and the climate variable may not hold steady over time. For example, if a changing climate makes tree-ring width – and associated tree growth – more rainfall-dependent than temperature-dependent, the proxy may cease being useful. This has occurred in some Bristlecone pine records.

    An example of this is the “divergence problem” – a tendency of some tree-ring wood density records to “decouple” from observed temperatures after 1950. While wood density and tree-ring width in some regions of the world matched observed temperatures well prior to 1950, some – though not all – tree-ring records failed to capture the rapid observed warming after that point.

    These sorts of events are problematic, however, as researchers have to try and ensure that similar divergences in the relationship between the proxy value and climate variable did not occur during other periods prior to the availability of observational records.

    In some cases, limited historical observations prior to the advent of modern climate monitoring networks can be used for providing a cross-check on reconstructions based on other proxy records and for validating past model simulations.

    How does proxy data inform climate science?

    As discussed above, observed records of global temperature only go back to 1850 and, in many regions, the temperature record is even shorter. Other climate records – such as the composition of the Earth’s atmosphere, streamflow, hurricanes, wildfires or solar output – may have much shorter historical observational records.

    To understand changes across different aspects of the Earth system prior to the start of instrumental records, scientists, therefore, have to rely on proxy measurements.

    Proxies have been instrumental in many ways to the development of modern climate science. For example, proxy records of both greenhouse concentrations and temperatures over the past 800,000 years have helped scientists understand the drivers of ice age cycles – including the critical role of CO2 in the process.

    While many studies have examined individual proxy types and locations – such as a specific coral reef or cave – over the past few decades there has been a focus on using a number of different proxies in conjunction to get a better understanding of regional or global changes.

    This is important, as a single location such as Antarctica may exhibit much larger swings in temperature or other climate variables over time than the globe as a whole. These “multiproxy climate reconstructions” have examined changes in air temperatures, drought, precipitation, sea surface temperatures, sea level, sea ice and vegetation among other factors.

    For example, a recent Nature study pulled together “a large collection of geochemical proxies for sea surface temperature” in an effort to reconstruct global temperatures during the most recent ice age, known as the Last Glacial Maximum. The researchers then validated their results against 18O isotope records derived from ice cores and speleothems.

    One of the largest multiproxy reconstructions is from the Past Global Changes (PAGES) project, a collaboration between thousands of palaeoclimatologists from 125 different countries that began back in 1991.

    In 2019, they published a thorough analysis of global surface temperatures over the past 2,000 years – called the PAGES 2K project. The figure below shows their resulting reconstructions across all the different methods the team examined. The yellow line shows the median across all of the palaeoclimate proxy reconstructions they examined, while the yellow shaded area shows the uncertainty range (2.5th percentile to 97.5th percentile) of the proxy reconstructions. The red line shows observed global surface temperatures after 1850.

    Global mean surface temperature reconstruction (yellow line) and uncertainties (yellow range) for the years 0-2000 period from the PAGES 2k Consortium along with observations from Cowtan and Way from 1850-2017. Data available in the NOAA Paleoclimate Archive.Researchers have also produced palaeoclimate proxy reconstructions over the full Holocene – the modern geologic epoch spanning the past 12,000 years. The figure below shows a range of reconstructions (grey band) as well as the median estimate (yellow line) of global average temperature.

    Global mean surface temperature reconstruction (yellow line) and uncertainties (grey range) for the period from 10,050BC to AD1950 from the Temperature 12k Database. Recent observations are not shown due to the low temporal resolution of the underlying data. Data available in the NOAA Paleoclimate Archive.


    (It is worth noting that there is some disagreement between different methods of Holocene temperature reconstructions, with one recent paper suggesting that the holocene maximum temperatures may have been considerably lower than other proxy reconstructions have estimated.)

    The further back in time reconstructions go, the lower temporal resolution they tend to have. In other words, records from 10,000 years ago might only represent an average temperature value over a 100-year period or more, while more recent proxy data tends to be closer to 20-year averages.

    This somewhat limits the ability of scientists to compare earlier proxy reconstructions to the modern temperature record without applying similar long-term averaging, though researchers have found some ways to get around this problem.

    In addition to understanding how temperatures and other climate variables have changed in the past, proxy data also gives scientists hints at how it might change in the future. Proxy data provides one of three key lines of evidence that scientists have used to better estimate the range of climate sensitivity – which determines how much the Earth will warm in the future if CO2 concentrations double.

    For example, explains Prof Dan Lunt, professor of climate science at the University of Bristol, “50m years ago, CO2 concentrations were greater than today and the planet was substantially warmer. Proxies for CO2 allow us to quantify the former and proxies for temperature allow us to estimate the latter”. Scientists can use this information to estimate climate sensitivity.

    However, he adds, there are considerable uncertainties associated with these estimates. This is because “we do not have complete geographical coverage of the whole planet, and because the correlation between proxy and climate is not perfect”. In addition, the “sensitivity” of the climate (how much the temperature of the Earth changes when CO2 in the atmosphere increases or decreases) when the Earth was in a different state – such as an ice age – may not be “a reliable indicator” for its sensitivity in the future, Lunt says.

    Proxy data can also help determine changes to sea level, ice cover and vegetation during past warm periods. One example is the previous interglacial period around 125,000 years ago – called the Eemian that was likely as warm or warmer than today – which can provide evidence on how all of these factors may change as the world warms.

    Another way that proxies inform climate science is through modelling. Proxy records can be used to help evaluate climate models. Scientists run “hindcast” simulations of models to see how well they reproduce the past climate and proxy data is needed to test against the climate of the distant past.

    The benefits of this are two-fold, explains Lunt:

    “A model that agrees well with the proxy data can then be used with confidence to improve our understanding of that time period…[and] if a model agrees well with proxy evidence of the past, then in some cases this can give increased confidence in its simulation of the future.”

    Lunt leads an international modelling project called DeepMIP (Deep-Time Model Intercomparison Project). The project was conceived in recognition of the fact that climate models are typically evaluated against the Earth’s relatively recent past, while global temperatures might be heading to levels not seen for tens of millions of years (“deep-time”), he explains.

    DeepMIP uses proxy data to evaluate models and learn about the past climate system, Lunt says – “in particular of the super-warm early Eocene Climatic Optimum (EECO, ~50m years ago) and Paleocene-Eocene Thermal Maximum (PETM, ~55m years ago)”.

    What are the limitations of proxy data?

    As the sections above highlight, the process of identifying, extracting, interpreting and calibrating proxy data to produce climate records is anything but straightforward. Such complicated techniques, therefore, have their pitfalls and limitations.

    Proxy data is, by definition, indirect, says Cluett. So while proxies are recording changes in the environment, scientists need to take account of both the local conditions and any wider influences that might be at play.

    On a local scale, for example, “the relationship between proxies and their environments may differ in different systems, so understanding the system you are working in is critical to accurately interpreting proxy data”, she says.

    For example, notes Pearson, the overall amount of the isotopes in the world’s ocean can change over time. This affects the analysis of an individual record, he explains:

    “Changes in the size of the world’s ice sheets changes the whole ocean isotopic ratio and so imprints itself on the [proxy] record.”

    Not all proxy records are made equal, points out Dr Justin Martin, an ecologist at the US Geological Survey. An ideal record is typically “long, accurately dated and of sufficiently high resolution over time to provide useful information”, he tells Carbon Brief. However, this is not always what scientists get, he says:

    “Some proxies may not cover a very long time period, or may only provide a relative estimate of climate variability over time that lacks certain dates. Others may only provide very coarse estimates of variability measured in decades, centuries or longer.”

    And all proxies “are subject to the vagaries of preservation one way or another”, adds Pearson.

    With historical documents, for example, older records are generally less useful. A book on temperature reconstructions by the US National Research Council notes that there are “weather records preserved in Irish and Norse annals back to the middle of the first millennium AD”, but says “their dating is imprecise and descriptions of weather and climate often are exaggerated”.

    All of these complications mean that “there may be quite large error bars that come not just from our analytical precision, but also how well calibrated the proxy is or can be to the target variable of interest”, says Pearson. He adds:

    “For these reasons, there is a lot of technical literature on the subject and, in general, we like to use multiple proxies together, if we can.”

    These obstacles also make palaeoclimate a “somewhat maddening science at times”, Pearson notes, and “certain knowledge about the past is very hard to come by”. Nonetheless, he says:

    “It is also very exciting because the timescales are potentially huge and the proxies we use are limited only by our scientific imagination and ability to measure things of interest.”

    Read at Carbon Brief.