Sorry, you need to enable JavaScript to visit this website.

PAGES Magazine articles

Publications
Author
John Slattery and Louise C. Sime
PAGES Magazine articles
2023
Past Global Changes Magazine
John Slattery1,2 and  Louise C. Sime1

Greenland ice cores provide high-resolution records of Dansgaard–Oeschger events – abrupt climate transitions, which happened repeatedly during the last glacial. These records can allow us to understand past climate tipping behaviour and to predict possible future tipping points.


Climate tipping points

The ability for society to reliably forecast future anthropogenic climate change is challenged by the likely presence of climate tipping points. These are hypothesized thresholds of global warming at which certain subsystems within the Earth’s climate, known as tipping elements, may undergo sudden and irreversible change in a process called a critical transition. Potential tipping elements that have been identified include the West Antarctic Ice Sheet, the Amazon Rainforest, and the Atlantic Meridional Overturning Circulation (AMOC) (Armstrong McKay et al. 2022, Wunderling et al. 2021). Were any of these elements to tip in future, it would lead to dramatic regional or global climate change, making this an extremely policy-relevant concern (Armstrong McKay et al. 2022). Furthermore, interactions between tipping elements raise the possibility of a disastrous cascade, with a domino-like effect allowing one abrupt change to lead to several more (Wunderling et al. 2021). Thus, quantifying the likelihood of passing a tipping point in the future has become a major focus of climate research.

Dansgaard–Oeschger Events

Dansgaard–Oeschger, or DO, warming events are a series of abrupt transitions in the climate between colder Greenland stadials and warmer Greenland interstadials that occurred repeatedly during the last glacial. These events were first observed in stable water-isotope records from Greenland ice cores, as shown in Figure 1. These stable isotope records show abrupt changes that suggest warming of up to 15°C in just a few decades (Kindler et al. 2014). This extreme rate of warming makes DO events the fastest instances of regional temperature change seen in the paleoclimate record, and so they are seen as the epitome of abrupt climate transitions. As well as the rapid warming inferred by stable water isotopes, Greenland ice cores show transitions in the annual layer thickness, as well as the concentrations of sea salt and mineral aerosols (Capron et al. 2021), respectively suggesting abrupt changes in precipitation, sea-ice extent, and atmospheric circulation in the North Atlantic region. Therefore, these ice cores contain a wealth of information about the climate changes that occur during DO events.

Beyond pure scientific curiosity, there are very practical and pressing reasons to be interested in DO events. There is compelling evidence that DO events are themselves the consequences of a tipping point being crossed (Boers 2018), in this case not because of human actions, but instead due to gradual internal changes within the climate system. A consensus has developed that at the heart of DO events lie transitions between strong and weak states of the AMOC (Li and Born 2019; Malmierca-Vallet et al. 2023; Vettoretti et al. 2022), a crucial system of ocean currents that transports large amounts of heat towards higher latitudes in the Northern Hemisphere and is possibly approaching a tipping point (Armstrong McKay et al. 2022). This hypothesized tipping point would be driven by different mechanisms to those involved in DO events, thus we cannot draw direct comparisons. Nevertheless, this possibility means that it is imperative that we develop a deeper understanding of past tipping behavior, as seen in DO events, such that we stand the best possible chance of understanding potential future critical transitions.

Greenland is, undoubtedly, where the best records of DO events are found, but their signatures can be seen further afield too. Approximately simultaneous with the local changes in surface temperature, precipitation, and sea-ice extent recorded in Greenland ice cores, speleothem records from East Asia and Atlantic marine-sediment cores show abrupt transitions in the large-scale atmospheric (Wang et al. 2008) and oceanic (Lynch-Stieglitz 2017) circulations. These changes are all intertwined due to a complex set of feedbacks between the three key components of ocean, atmosphere, and sea ice (Li and Born 2019; Malmierca-Vallet et al. 2023). The picture that emerges is a cascade of transitions, with an initial transition in one climate element leading to a transition in the next, and so on. What remains very unclear is the order of this cascade. Particularly important is the question of which element begins this cascade. Based on our current knowledge, we cannot rule out any of the three key climate components previously mentioned, or even narrow down the location. Every imaginable answer to this question has been suggested, with different researchers often arriving at different conclusions using simulations from the exact same model (Kleppin et al. 2015; Vettoretti et al. 2022)! At the same time, locating the start of this chain of causality is needed before we can be certain of the mechanism responsible for triggering DO events – which is ultimately the aim of this field of research.

Figure 1: The famous Greenland δ18O record, in this case from the NGRIP ice core (Andersen et al. 2004). Less negative values of δ18O qualitatively correspond to higher temperatures. A 100-year smoothing has been applied. DO warming events are shown by gray vertical lines.

Systematic timing differences

It is clear that a new approach is needed to better understand the cascade of transitions that occurs during a DO warming event. One option is to search for systematic timing differences between the different climate elements which show rapid change; a line of research for which Greenland ice cores are uniquely suited. A single ice core can provide independent measurements of multiple species associated with different components of the climate system at sub-decadal time resolution within a single archive, getting around the problem of dating uncertainties, which makes timing comparisons between different paleoclimate archives so difficult (Capron et al. 2021). The measurements that have been used to investigate the temporal phasing of DO events are the mineral-dust aerosol (Ca2+) and sea-salt aerosol (Na+) content from both the North Greenland Ice Core Project (NGRIP) and North Greenland Eemian (NEEM) ice cores, as well as the annual layer thickness (λ) and the water-isotope ratio (δ18O). These four measurements are respectively interpreted as reflecting the Northern Hemisphere atmospheric circulation, North Atlantic sea-ice extent, local precipitation rates, and surface-air temperature at the ice-core site (Erhardt et al. 2019). This gives potentially independent data on four climate elements. It is important to note, however, that individual measurements are not always uniquely associated with a single climate element. For example, the δ18O of Greenland ice cores is both reflective of sea-ice change around Greenland, and temperature change at the ice-core site (Sime et al. 2019). For these reasons, whilst ice cores are fantastic archives of DO events, care is required when using these measurements to draw conclusions about the timing of changes across different elements.

Even with the excellent precision that ice-core measurements afford, pinpointing the timing of a tipping point in a noisy climate record is statistically challenging. The acceptance of this uncertainty, and resultant use of Bayesian inference to produce probabilistic distributions for the start and end times of DO warming events, has been a major step forward (Capron et al. 2021). An example of the application of this type of approach to a DO event in the NGRIP δ18O record is shown in Figure 2. Stacking all the DO events covered by the ice cores, we can then assess the mean differences in timing between the transitions in different core measurements. The first pioneering study to utilize this method indicated that the transitions in Ca2+ and λ systematically led those in Na+ and δ18O (Erhardt et al. 2019), suggesting that atmospheric changes were in fact the first element of the DO warming event cascade, and apparently representing a breakthrough in our understanding.

Figure 2: An example application of the Bayesian method to characterize a DO transition. The blue line shows the raw measurements of  δ18O from the NGRIP ice core (Andersen et al. 2004). The black line shows the median fit, with the 90% uncertain range shaded in orange. The blue and orange histograms are, respectively, the probability distributions for the start and end of the transition.

Subsequent work showed that attempting to estimate the population mean time difference from a relatively small sample of DO events, of which each is itself uncertain, meant that the two-fold uncertainty inherent to this method had been underestimated. Rigorously propagating this uncertainty revealed that the previously reported systematic timing differences were not statistically robust (Riechers and Boers 2021). Thus, the current picture is that the variation between individual DO events is too great, and the rapidity of the element cascade too rapid, for any certainty regarding the sequence of changes to be possible using the measurements currently available from ice cores (Capron et al. 2021). In future, new ice cores, higher resolution measurements, or improved analytical techniques may allow us to make further progress on this crucial problem. But for now, despite giving us tantalizing teases, the ice cores withhold their secrets.



Publications
Author
Rodrigo L. Soteres, E.A. Sagredo, M.R. Kaplan, M.A. Martini, F.M. Riquelme and J.M. Schaefer
PAGES Magazine articles
2023
Past Global Changes Magazine
Rodrigo L. Soteres1,2, E.A. Sagredo3,4, M.R. Kaplan5, M.A. Martini6, F.M. Riquelme3 and  J.M. Schaefer5,7

Surface-exposure dating of moraines reveals that Patagonian glaciers fluctuated at the pulsebeat mimicked in polar ice cores from both hemispheres. These findings favor hypotheses that invoke coupled oceanic–atmospheric drivers to generate and propagate millennial-scale climate shifts during Termination 1.


Moraines as past climate archives

Mountain glaciers are very sensitive to climate oscillations, especially to atmospheric temperature and precipitation. When climate conditions favor glacier growth, ice mobilizes massive quantities of sediments, including large erratic boulders, which are piled up alongside the ice margins, forming sharp and elongated hilly landforms, called moraines. We assume they form mostly when glaciers are in close equilibrium with the local prevailing climate, culminating at the transition between cold and warm periods. Therefore, well-preserved sequences of moraines are valuable archives of the timing of terrestrial paleoclimate.

Cosmogenic nuclides (or isotopes) are mostly generated by the interaction between cosmic radiation and target minerals within rock surfaces exposed to the atmosphere. Given that these nuclides are naturally near-absent in most lithologies, their concentration will start building up when the rock begins to intercept cosmic radiation products. So, if the production rate of these nuclides is known, we can measure them and then calculate the “exposure age” (Schaefer et al. 2022). Continual refinement of surface-exposure dating based on in situ terrestrial cosmogenic nuclides, particularly 10Be in moraine boulders, has allowed us to assign precise ages to former glacial oscillations and, therefore, to constrain the chronology of past climate variability.

The conceptual interpretation of 10Be glacier chronologies (Fig. 1a) assumes that boulders in resting position at the highest part of the moraine ridges represent the cessation of its construction, so the surface exposure age of these boulders will pinpoint the onset of glacier recession (Fig. 1b). We prefer to sample tall and large boulders well embedded in the top of the moraine to prevent potential post-depositional displacements and/or exhumation that could provide anomalously young ages. We also aim to collect samples from glacially polished surfaces to try reducing inherited cosmogenic nuclides generated during prior exposure episodes, which could yield anomalously old ages (Fig. 1c). This sample approach permits us to diminish geological uncertainties affecting surface-cosmogenic-nuclide concentrations and, thus, to obtain accurate ages for culminations of glacier pulses.

Figure 1: Rationale behind surface-exposure dating of moraines considering (A) glacial; and (B) post-glacial stages. Red lines on the clocks in panel (B) depict the exposure time, while the size of the molecule scheme represents the concentration of cosmogenic nuclide. (C) Example of a moraine erratic boulder sampled for 10Be dating. Photo credit: Gonzalo Amigo.

Patagonia: the southernmost mid-latitude continental landmass in the world

Patagonia, including Tierra del Fuego, spans between ~40° and ~56°S in South America, comprising both Chilean and Argentine territory (Fig. 2a). It is the only continental landmass that covers the entire latitudinal range of the Southern westerly winds (SWW), whose interaction with the Southern Ocean play a key role in global climate. Seasonal temperatures and precipitation in Patagonia are strongly correlated with low-level zonal winds indicating that regional climate is mostly driven by the migrating locus of this hemispheric wind belt. Additionally, the Austral Andes generates a strong orographic effect that promotes a sharp transition between a western hyper-humid to an eastern semi-arid flank (Garreaud et al. 2013). Regional climate permits the existence of numerous mountain glaciers, including the Patagonian icefields, which constitute the most extensive ice masses of the Southern Hemisphere, outside Antarctica.

During the last glaciation, western Patagonia was covered by an ice sheet, which featured numerous outlet glacier lobes flowing towards both flanks of the Andes (Davies et al. 2020). These large glaciers formed a wide variety of ice-marginal landforms, including outstanding sets of moraines. This glacially shaped landscape has enabled Patagonia to play a major role in moraine-based paleoclimate reconstructions in the Southern Hemisphere.

Moraine-based T1 paleoclimate reconstructions in Patagonia

The end of the last glaciation, known as Termination 1 (T1), comprises a sequence of millennial-scale climate shifts that ultimately led global climate from cold-glacial to warm-interglacial conditions, between ~18 and ~11.7 kyr. δ18O from polar ice cores (Fig. 2b) exhibit a stepwise net warming trend interrupted by cold snaps between ~14.5–12.8 kyr in Antarctica, the so-called Antarctic Cold Reversal (ACR), and between ~12.7–11.7 kyr in Greenland, known as the Younger Dryas (YD). Millennial-scale climate shifts during T1 have been inferred as occurring synchronously, but in an out-of-phase style between hemispheres, which led to a proposed “thermal bipolar seesaw” as the potential mechanism responsible for this climate mode (Pedro et al. 2018). Therefore, deciphering the timing and geographical extent of both the ACR and the YD at planetary scale is imperative to properly test the influence of this hypothetical mechanism on global climate. To solve this fundamental paleoclimate conundrum, surface exposure chronologies of mountain glaciers offer crucial insights, including those from Patagonia.

Following pioneering cosmogenic dating of T1 moraines in Patagonia, recent highly resolved 10Be glacial chronologies are yielding unequivocal evidence of glacier pulses during the ACR and the YD from north to south. At Cerro Riñón valley in Lago Palena/General Vintter (44°S), Soteres et al. (2022) found moraine crests constructed between ~13.5–13.1 kyr and at ~12.4 kyr. In central Patagonia, Sagredo et al. (2018) documented moraine ridges formed between ~13.7–13.4 kyr and at ~12.0 kyr in the Río Tranquilo valleys around Cerro Cochrane/San Lorenzo (47.5°S). Furthermore, recent modeling analysis conducted by Muir et al. (2023) indicate that glaciers in Cerro Riñón and Río Tranquilo fluctuated in unison to a ~3°C cooling coeval with the ACR, followed by a ~0.5°C temperature increase coinciding with the YD. They concluded that glaciers in the northern half of Patagonia responded simultaneously to deglacial climate forcings. Nearby, in Lago Belgrano (47.8°S), Mendelová et al. (2020) reported moraines built between ~13.2–13.0 kyr and at ~12.4 kyr. In southern Patagonia, Ackert et al. (2008) and Kaplan et al. (2011) analyzed 10Be concentrations in moraine boulders while Strelin et al. (2011) used radiocarbon to date moraines at Lago Argentino (50°S), both constraining glacier advances culminating between ~13.2–13.0 kyr and at ~12.2 kyr. In the latter location, a Patagonian production rate for 10Be on rocks was also established, with implications for the accuracy of all Patagonian 10Be surface exposure ages. Further south, García et al. (2012) and Menounos et al. (2013) discussed moraines with ACR cosmogenic ages ranging between ~14.1–13.8 kyr in Torres del Paine (51°S) and ~13.4 kyr in Ushuaia (54°S), respectively (Fig. 2c).

Figure 2: Selected glacial chronologies around Patagonia. (A) Circles show sites with name and label. White and gray shaded areas outline the former Patagonian Ice Sheet and the present-day icefields, respectively. (B) δ18O records from WAIS Divide (blue line) and NGRIP (purple line) ice cores. (C) Individual recalculated 10Be ages (Kaplan et al. 2011) per moraine by label and numeric morphostratigraphic order. *In Lago Argentino, although used to establish the Patagonian 10Be production rate, Herminita moraines 10Be concentrations are also consistent with glacier activity during the YD. (D) Bayesian relative probability plots of the selected Patagonian 10Be moraine chronologies.

Altogether, Bayesian 10Be ages probability correction of the Patagonian moraines reveal ubiquitous glacial advances/standstills in close morphostratigraphic order at least at ~14.0, ~13.4 kyr and ~13.0 kyr during the ACR, followed by pervasive ice retreat interrupted by a minor glacier pulse at ~12.2 kyr during YD times, spanning ~10° of austral latitude (Fig. 2d). Overall, these moraine chronologies show an emerging scheme that mimics millennial-scale climate shifts detected in both Antarctic and Greenland ice cores during T1 (Fig. 2b). This is supported by similar mountain glacier records obtained in the Northern Hemisphere (e.g. Bromley et al. 2023), challenging the widely accepted interhemispheric “thermal bipolar seesaw” mechanism. Denton et al. (2022) offered an alternative hypothesis that reconciles the apparent global synchronicity in mountain glacier fluctuations by invoking extreme seasonality episodes in the North Atlantic. Accordingly, the isotope signature in Greenland ice cores might be reflecting regional hyper-cold winters rather than a hemispheric climate signal, particularly during the YD. Their hypothesis, dubbed Heinrich Summers, implies that the Antarctic-like fashion of millennial-scale climate oscillations would have prevailed in both hemispheres during T1. If correct, this would favor the SWW-Southern Ocean coupled system as the main driver to generate and globally propagate climate signatures at millennial timescales during the end of the last glaciation.

Since early naturalists began to investigate glaciations in South America over a century ago, Patagonia has united several generations of scientists from all around the planet to establish the region as a key site for past glacier and paleoclimate studies. However, further research is needed to better understand the present-to-future evolution of global cryosphere and climate.

Publications
Author
Ling Fang, T.M. Jenk and M. Schwikowski
PAGES Magazine articles
2023
Past Global Changes Magazine
Ling Fang1,2,3, T.M. Jenk1,2 and  M. Schwikowski1,2,4

An accurate chronology of alpine ice cores is essential to interpret the climate-signal and atmospheric-pollution history archived in glaciers. The radiocarbon in water-insoluble organic carbon (WIOC) has emerged as a valuable tool for dating alpine ice cores.


The challenge of alpine ice-core dating

The most common ice-core dating approach is annual layer counting, which relies on seasonal variations in chemical and physical signals, such as ammonium, stable isotope ratios (δ18O, δ2H) and solid electrical conductivity. However, pronounced thinning beyond a certain depth limits this approach in its application for ice cores.

To establish a complete age–depth scale down to the bedrock, simplified ice-flow models can be employed (e.g. Dansgaard and Johnsen 1969), but they are unable to resolve small-scale variations in ice flow (i.e. thinning/strain), particularly closer to the bedrock, where the often complex glacier geometries of high-mountain glaciers become increasingly important. Also, they rely on fundamental assumptions, such as constant accumulation, that likely do not reflect the actual conditions over time. Even complex 3D models cannot convincingly simulate the age of the deepest sections, if no additional age constraints are available (e.g. Licciulli et al. 2020).

Absolute time horizons can pin down the age of the ice. A valuable marker for alpine ice cores is the signal from atmospheric nuclear-weapon testing, showing up as a clear peak in 1963 CE in multiple proxies, such as tritium or cesium-137. Known volcanic eruptions, such as Katmai in 1912 CE and Tambora in 1815 CE, indicated by peaks in sulfate and conductivity, are also commonly used as time markers (Herren et al. 2013). Moreover, Saharan dust events during the 20th century (e.g. 1977, 1947 and 1901 CE) are well documented and can easily be identified in the European Alps (often visually, coinciding with peaks in calcium concentration in the ice cores). However, these events were only documented for the last two centuries, and these horizons cannot be established in deeper sections where no conventional dating techniques are applicable.

The development of radiocarbon analysis of WIOC

Radioactive nuclides entrapped in the ice offer an opportunity to obtain absolute dates. The environmental radionuclide 210Pb, with a short half-life of 22.3 years, enables the dating of ice over roughly one to two centuries (Gäggeler et al. 2020). The noble gas 39Ar and 32Si with a half-life of 268 ± 8 years and of 144 ± 11 years, respectively, have been demonstrated as ideal dating isotopes for ice samples from the last thousand years (Morgenstern et al. 2010; Ritterbusch et al. 2022).

The recent developments in atom-trap trace analysis (ATTA) have allowed scientists to substantially reduce the required amount of ice to 1–3 kg for 39Ar dating (Ritterbusch et al. 2022). The long-lived 81Kr, with a half-life of 229,000 years, can date ice up to 1.5 million years old (Tian et al. 2019). However, due to the low abundance of 81Kr, ~10 kg of Antarctic ice or 20–40 kg of ice from the Tibetan Plateau is recommended for sampling (Tian et al. 2019). Given the half-life of 5370 years, radiocarbon (14C) is considered the most suitable radionuclide for dating ice samples up to around 20,000 years old, covering most of the time range typically accessible by alpine ice cores (Uglietti et al. 2016).

Figure 1: Map showing (A) the sites from which ice samples were 14C dated with WIOC; and (B) the averaged WIOC concentrations (μg/kg) for different regions. The size of the gray circles on the map corresponds to the WIOC concentrations.

Previously, 14C dating of ice was only possible where sufficient organic matter such as plant, wood or insect fragments was found (Thompson et al. 1998). However, the occurrence of such findings is rare in glacier ice, and even when they are present, do not allow for continuous dating. To overcome this challenge, Jenk et al. (2009) introduced the use of water insoluble organic carbon (WIOC) for 14C dating of alpine ice cores and Uglietti et al. (2016) later validated the method.

Carbonaceous particles are a major component of the atmospheric aerosol and deposit onto the glacier by precipitation. They are composed of two main bulk fractions: organic carbon (OC) and elemental carbon (EC). OC can be split into WIOC and dissolved organic carbon (DOC; see below) by solubility. WIOC and EC are separated based on their specific thermal properties (combustion temperatures). The micro-carbon 14C dating method relies on the finding that WIOC originated solely from biosphere emissions prior to the use of fossil fuels (~1850 CE; Jenk et al. 2006). Once emitted, the 14C decays according to its half-life time, starting the radiometric clock. A detailed method description for WIOC 14C-dating can be found in Uglietti et al. (2016).

Various ice samples have been dated using micro-14C WIOC since the introduction of this technique (Figs. 1–2). WIOC concentrations ranged from 2–15 μg/kg in samples from the Polar Regions to 155 μg/kg in ice from the Tibetan Plateau (Fang et al. 2021; Uglietti et al. 2016), while Alpine ice samples used for dating contained 44 ± 14 μg/kg on average (Fang et al. 2021; Uglietti et al. 2016; Fig. 1). To date, the oldest sample was determined with an age of ~22 kyr BP at the bottom of an ice core retrieved from Belukha (Russian Altai; Fig. 2). On Colle Gnifetti (CG03 core) in the European Alps, the oldest ice retrieved from the Alps was >15 kyr BP, and on Illimani in the Andes ~12.6 kyr BP (Fang et al. 2021; Jenk et al. 2009; Sigl et al. 2009; Uglietti et al. 2016; Fig. 2).

The majority of dated alpine ice cores, however, was younger than 10 kyr BP (e.g. Uglietti et al. 2016). The dating precision strongly depends on the 14C content of the sample, defined by the carbon mass and the age of the sample, i.e. its 14C/12C ratio. The uncertainty decreases sharply with increased carbon mass due to the blank correction. Therefore, a total carbon amount of 10 μg, typically equivalent to around 300–500 g of ice, is recommended for reliable dating.

Figure 2: Ice-core chronology based on 14C dating (modified from Figure 6 in Uglietti et al. 2016). Horizons and 14C results are shown as black triangles and green circles, respectively. Gray shaded areas represent the 1σ range of the respective fit for retrieving a continuous age–depth relationship. To enhance visibility, the curve for the Mt Ortles glacier was shifted downward by 20 m water equivalent (m w.e.) and is referenced in the right-hand y axis, denoted with an asterisk (*).

The most recent developments and outlook

In glacier ice, the higher concentration of DOC compared to WIOC (by a factor ranging from 2 to 5; Fang et al. 2021) provides a motivation to investigate the possibility of using DOC for 14C dating. The required mass of ice could potentially be further reduced, if WIOC and DOC were to be extracted from the same piece of ice. Two studies found that the DOC fraction can be biased in its 14C/12C ratio due to in situ 14C production by cosmogenic radiation (Fang et al. 2021; May 2009). This limits the theoretically achievable gain in dating precision given by the higher carbon mass.

Nevertheless, Fang et al. (2021) showed the great potential of the DOC fraction to date ice from sites where in situ 14C production is relatively low. This is the case for sites at altitudes of 4000–5000 masl and below (low radiation), and/or characterized by snow accumulation rates greater than 0.5–1 m water equivalent (less exposure) such as in the European Alps. Method details can be found in Fang et al. (2019), also describing the most recent setup allowing simultaneous extraction of DOC and WIOC samples for 14C dating of ice, which was built at the Laboratory for Environmental Chemistry (PSI, Switzerland).

In conclusion, due to lower demands of ice mass and the time coverage of the lowest and oldest ice-core sections, 14C analysis has become a crucial and widely applied tool for the dating of ice cores from mountain glaciers. Although the possibility of using the DOC fraction for ice-core dating has been demonstrated, further studies are needed to explore the full dating potential in terms of ice requirement and analytical precision, and accuracy for using both OC fractions, extracted from the same sample.

Publications
Author
Marie Bouchet, A. Landais and F. Parrenin
PAGES Magazine articles
2023
Past Global Changes Magazine
Marie Bouchet1, A. Landais1 and  F. Parrenin2

We review some of the possible methods for building optimized and coherent timescales of deep polar ice cores. We focus on drilling sites characterized by a low temporal resolution due to minimal accumulation of snow at the ice-sheet surface.


Deep polar ice cores are unique archives of past climate. Their investigation is valuable to study mechanisms governing the Earth’s climate variations during the glacial–interglacial cycles of the late Quaternary. Precise ice-core chronologies are essential to determine the sequences and durations of climatic events, as well as questioning phase relationships between the external forcings and the climatic responses. One example of climate forcings are the orbital parameters governing the amount of solar energy received at the Earth’s surface.

Three challenges are associated with the dating of deep ice cores:

  1. A coat of unpacked snow (50–120 m), the firn, covers the ice sheet. The atmospheric air circulates freely within the firn. At the firn-ice transition, the air is enclosed in bubbles and no longer diffuses. Hence, the construction of two separate chronologies is required: one for the ice and one for the younger air.

  2. Most of the paleoclimatic information is recorded within the deepest part of the ice core, due to the thinning of ice layers from their deposition at the surface to the bottom of the ice sheet. Improving the timescales of deep ice cores is therefore of great concern for the ice-core community, along with extending them further back in time.

  3. Ice cores drilled at sites characterized by high accumulation rates of snow at the surface (10–30 cm/year) are dated by counting annual layers via identification of a seasonal cycle in some records (Sigl et al. 2016). Conversely, some East Antarctic sites show comparatively low accumulation rates (1–5 cm/year), which prevent annual layers from being identified and counted. Chronologies of deep ice cores therefore involve other strategies, summarized below.

Glaciological modeling

Glaciological models simulate the flow and thinning of annual layers over time, from surface deposition down to the bedrock, thus providing the ice age–depth relationship (Parrenin et al. 2004). The model inputs are past scenarios of snow accumulation and temperature at the surface, estimated from water–isotope measurements, together with a calculated temperature–depth profile in the ice sheet. This strategy is highly dependent on poorly known boundary conditions and physical constants. Glaciological modeling is thus combined with dating constraints, which are depths with a known ice or gas age.

Dating constraints

Absolute dating constraints in ice cores can be determined using radioactive isotope records. The 10Be production rate in the atmosphere relates to the geomagnetic field and solar activities. The Laschamp Excursion, a rapid drop in the Earth’s geomagnetic field intensity, is visible as a peak in the ice-core 10Be records, and is independently dated with different series at 41 kyr BP (thousand years before 1950; Raisbeck et al. 2017). Ice-core 40Ar records reflect past atmospheric concentration modulated by the radioactive decay of 40K in the Earth’s crust (Yan et al. 2019). Recently, 81Kr measurements on ice samples of a few kilograms provided age estimates between 1300 and 300 kyr BP (Buizert et al. 2014).

Another approach, called “orbital dating”, consists in synchronizing ice-core proxies to the Earth orbital parameters (or targets), whose variations are precisely modeled in time. The alignment of the proxy with its target gives ice- or gas-age constraints (Fig. 1). Three orbital proxies are used: δ18Oatm, δO2/N2, and total air content. The oxygen in air bubbles (δ18Oatm) is sensitive to ocean water δ18O (and, therefore, to the global ice volume), as well as to the biosphere productivity and the low latitude water cycle. Conversely, the oxygen in precipitation (δ18 Oice) depends on local temperature changes, and, thus, not used for orbital dating. δ18Oatm was synchronized to the Earth’s axial precession, delayed by 5000 years, because such a delay was observed during the last deglaciation. However, the lag of δ18Oatm behind precession fluctuates. Rapid climatic instabilities linked to breakdowns of the Northern Canadian Ice Sheet (Heinrich-like events) occur during deglaciations, which could be responsible for occasionally delaying the response of δ18Oatm to orbital forcing via changes in the water cycle (Extier et al. 2018). The variability of this delay induces a lack of confidence in the δ18Oatm–precession synchronization, associated with an uncertainty of 6000 years, which corresponds to the quarter period of a precession cycle. Further, the ice core δ18Oatm and Chinese speleothems δ18Ocalcite signals display identical features. The two series show orbital-scale (induced by the precession forcing) and millennial-scale oscillations, both types of variations being associated with changes in the low latitude water cycle imprinted in δ18Oatm and δ18Ocalcite (Fig. 1a).

To improve the precision of the gas chronology, it is preferable to synchronize the δ18Oatm variations with the δ18Ocalcite record from uranium-series-dated Asian speleothems (Cheng et al. 2016). In addition, Bender (2002) and Lipenkov et al. (2011) observed that the δO2/N2 and total air content records simultaneously oscillate with the local summer insolation (Fig. 1b). They formulated the subsequent hypothesis: insolation modulates near-surface snow properties (grain size and shape). This imprint is preserved as snow densifies in the firn and, later, affects the ratio δO2/N2 and the total air content in deep ice. The total air-content variations share more similarities with Earth’s axial obliquity than δO2/N2, hence its insolation target is integrated over an extended summer interval. Wiggle-matching between δO2/N2 and total air content, and their insolation targets gives dating constraints with a relative uncertainty varying between 1000 and 7000 years (Bazin et al. 2013). The orbital dating accuracy is liable to:

  1. The choice of the well-suited orbital target;

  2. its synchronization with the orbital proxy, which can be ambiguous when Earth’s orbit is nearly circular; and

  3. the poor quality of measurements in the deepest sections of the cores (gray areas in Fig. 1).

Other tracers supplying relative dating constraints, or stratigraphic links, improve the consistency between timescales of different ice cores over the last glacial–interglacial cycle. The synchronization of globally well-mixed atmospheric-methane records from Greenland and Antarctic ice cores brings in stratigraphic links with an accuracy of 60 to 500 years (Epifanio et al. 2020). Climate independent constraints, such as large volcanic eruptions, leave singular sulfate patterns in ice cores from both hemispheres. The detection of these deposits results in highly precise (within 5 to 150 years) stratigraphic tie points between cores (Svensson et al. 2020).

Figure 1: Synchronization of ice-core records with well dated series. Alignment of EPICA Dome C records of (A) δ18Oatm and (B) δO2/N2 (Extier et al. 2018) to δ18Ocalcite from East Asian speleothems and local summer insolation, respectively. δO2/N2 is filtered in the insolation frequency band. Gray areas indicate time intervals of large dating uncertainty.

Connecting ice and gas timescales

The lock-in depth, indicating the depth threshold where the air is trapped in enclosed bubbles and no longer diffuses (Fig. 2), determines the age difference between the ice and gas phases at each depth. Through the diffusion column (the interval between the surface and the lock-in depth), the preferential downward diffusion of heavy isotopes increases the δ15N fraction of N2. Measurements of δ15N in air trapped in ice cores yields a first estimate of the diffusion column thickness, and, therefore, of the lock-in depth. This depth can also be calculated with firn densification modeling (Bréant et al. 2017).

Figure 2: Deep East Antarctic ice cores and age scale. The numbers at the bottom of each ice core indicates the maximum depth drilled. Dome F (720 kyr BP), South Pole (54 kyr BP), Vostok (420 kyr BP), Dome C (800 kyr BP) and Talos Dome (340 kyr BP). The Little Dome C drilling aims to reach 2750 m, and 1.5 Myr BP.

Bayesian dating tools

To build consistent ice-core chronologies combining independent synchronization, absolute and relative dating constraints, as well as glaciological modeling, Bayesian dating tools have been developed. Now, they have gained improved mathematical, numerical, and programming capacities (Parrenin et al. 2021). Prior estimates of gas and ice timescales built by glaciological models are statistically adjusted by the Bayesian tools to comply with the dating constraints. These probabilistic tools use an inverse method, integrating all dating information and associated relative uncertainty, to produce a coherent timescale for distinct ice cores.

Perspectives

Deep ice-core chronologies strongly rely on gas measurements. To improve the chronological precision, it is crucial to collect highly resolved data from ice samples stored at cold temperatures (-50°C) to avoid gas diffusion and loss of signal for δ18Oatm and δO2/N2. The dating accuracy is soon to be challenged by the upcoming Beyond EPICA ice core, expected to provide much more paleoclimatic information within a shallower depth range than present ice-core drillings (1.5 Myr BP zipped in ~2750 m, Fig. 2) (Fischer et al. 2013).



Publications
Author
Lindsey Davidge
PAGES Magazine articles
2023
Past Global Changes Magazine
Lindsey Davidge

Stable water-isotope measurements from ice cores reflect paleo-atmospheric thermodynamics along upstream moisture pathways. New analytical methods simplify the measurement of triple oxygen isotopes, which will improve measurement resolution and provide more complete information about the past atmosphere.


Water isotopes in ice cores (δ17O, δ18O, δD, deuterium excess, and 17O excess) reflect past climate conditions

Stable water-isotope measurements (e.g. δ18O and δD) from ice cores reflect temperature and other thermodynamic conditions of the past atmosphere. As atmospheric moisture moves poleward from evaporative source regions to eventual precipitation sites, mass-dependent fractionation processes progressively distill its isotopic composition. First-order, temperature-dependent equilibrium fractionation during precipitation is the dominant control on the observed ratio of heavy-to-light isotope abundance (i.e. δ17O, δ18O, or δD) at the ice-core site, and the water-isotope paleothermometer has consequently been a cornerstone of ice-core paleoclimate science for decades (e.g. Dansgaard 1964).

However, even these conventional applications of the water-isotope temperature proxy rely on quantitative models of upstream fractionation pathways. In other words, even though the condensation temperature exhibits large control on the water-isotope signal of the precipitation, the isotopic composition of the air parcel that condenses is predetermined by all upstream thermodynamic processes (e.g. Merlivat and Jouzel 1979). Modeling many unknown upstream fractionation processes (e.g. evaporation, atmospheric transport, and precipitation) is improved by the inclusion of additional water isotope observations that reflect those processes.

Second-order water-isotope quantities like deuterium excess (d) or 17O excess (∆17O) – which are defined by the relationships between δD and δ18O (d) or δ17O and δ18O (∆17O) as indicated in Figure 1a–b –, are dominated by these upstream kinetic fractionation events, and they can therefore provide information about the integrated history of an air parcel that has reached an ice-core site. While d and ∆17O both depend on temperature and humidity variations in the atmosphere, the sensitivities of d and ∆17O during fractionation are different, e.g. the relative effect of evaporation temperature is more important for d, and the relative effect of evaporation humidity is more important for ∆17O (e.g. Uemura 2010). Therefore, measuring d and ∆17O together should provide the most complete information about the past hydrosphere. The differences between d and ∆17O are highlighted in Figures 1c–f, which show seasonally resolved measurements of d and ∆17O from three ice-core sections from Greenland.

Despite the theoretical potential for ∆17O to be a complementary tracer to d, traditional ice-core work has not included δ17O or ∆17O due to measurement limitations. Most commonly, δ18O and d have been used to reconstruct past condensation-site and evaporation-source temperatures, but this method is imperfect because, in addition to the evaporation temperature, d is also influenced by other thermodynamic conditions during evaporation, atmospheric transport, and precipitation (Merlivat and Jouzel 1979). Including corresponding measurements of ∆17O would provide an additional constraint for reconstructing the water-isotope fractionation pathways, and it is therefore desirable to produce records of ∆17O and to develop climate models that account for ∆17O (e.g. Markle and Steig 2022; Schoenemann and Steig 2016). The following sections describe recent improvements to ∆17O measurement methodology and present existing ice-core records of ∆17O.

Figure 1: The relationships between δD or δ17O and δ18O – which define d and ∆17O – are provided in (A) and (B), respectively. Formal definitions of d and ∆17O are given by the equations in (C) and (D). (C) and (D) provide seasonally resolved records of d and ∆17O from NEEM (Landais et al. 2012a) and Summit (Davidge et al. 2022), both ice-core sites in Greenland. The decade of the measured ice-core layers is provided in the legend – see the original publications for dating methodologies. All data were aligned to the seasonal cycle of corresponding δ18O measurements to highlight seasonal patterns in d and ∆17O. Corresponding histograms of these same data are binned by the typical analytical uncertainty and highlight the differences in d (E) and ∆17O (F) distribution observed at these sites.

Recent advances in instrumentation improve temporal resolution of ∆17O from ice cores

Although d has routinely been measured on ice cores for decades, measuring ∆17O has only been possible for about 20 years (e.g. Barkan and Luz 2005), and observations of ∆17O are limited in spatial and temporal resolution. However, new analytical methods have the potential to simplify the measurement of ∆17O – which, when measured at all, is typically determined separately from other water-isotope quantities by discrete isotope-ratio mass spectrometry (e.g. Barkan and Luz 2005). Unlike d or δ18O, which vary in meteoric water by several or tens of "per mil" (‰, or parts per thousand), respectively, the natural variability of ∆17O in precipitation is only tens of "per meg" (or parts per million), which exacerbates measurement difficulties. However, recent advances in cavity ring-down laser spectroscopy (CRDS) enable the simultaneous measurement of all stable water isotopes (i.e. δ17O, δ18O, δD, ∆17O, and d) with precision that meets or exceeds that of traditional methods (see Steig et al. 2014). CRDS is an appealing method not only because it can measure all water isotopes at once, but also because it can be combined with continuous sample melting strategies that are already in use for other ice-core analyses. Over the last 10 years, continuous-flow analysis (CFA) has been widely adopted by ice-core laboratories, and measurements of δ18O, δD, and d by CFA-CRDS are already routine for ice-core measurement campaigns (e.g. Emanuelsson et al. 2015). Recent work (Davidge et al. 2022; Steig et al. 2021) demonstrates that CFA-CRDS for all stable water isotopes can greatly reduce the analysis time for ∆17O and it can therefore improve the time resolution of ∆17O measurements. CFA-CRDS methods will be useful for improving the temporal and spatial resolution of ∆17O to characterize the natural variability of meteoric ∆17O.

Recent work demonstrates that CFA-CRDS for ∆17O can indeed improve the resolution of ∆17O observations with high precision (<10 per meg), especially when CFA for ∆17O is developed with specific attention to calibration strategies (Davidge et al. 2022; Steig et al. 2021). Steig et al. (2021) measured the lower 1200 m of the South Pole ice core by CFA-CRDS for all stable water isotopes, revealing significant millennial-scale variability in ∆17O that is not observed in coarser records of ∆17O from other ice-core sites, but that is coincident with other climatic events recorded by d and δ18O (Figs. 2a–c). However, they also identify the importance of frequent (i.e. daily) data calibration against multiple reference waters, adopting a new calibration method that utilizes more reference water measurements than typical CRDS strategies. Davidge et al. (2022) demonstrated that CFA-CRDS for ∆17O performs as well as discrete methods by measuring replicate sections of an ice core from Greenland; annually resolved data from that study are provided in Figure 1c–d. They also found that, though small, the greatest source of uncertainty for ∆17O by CFA-CRDS is the calibration technique. Both studies suggest that the measurement resolution depends on the desired precision for ∆17O and the rate of the continuous melter. Continuing to develop and implement CFA-CRDS methods so that more existing ice cores can be measured for ∆17O will improve the spatial and temporal resolution of ∆17O, which is a critical step for studying atmospheric controls on second-order water-isotope quantities and refining interpretations of the paleoclimate record.

Figure 2: Corresponding measurements of δ18O (A) , ∆17O (B), and d (C) for available Antarctic ice-core sites, whose locations are mapped in (D). Continuously measured data from the South Pole is averaged to show 50 cm resolution for all isotope values (from Steig et al. 2021). All discrete data are shown at their measured resolutions. Details about data normalization protocols for ∆17O, and source information for all discretely measured data, are available from Schoenemann et al. (2013) and Schoenemann et al. (2014), respectively.

17O data varies on seasonal, millennial, and glacial timescales

Utilizing CFA-CRDS to characterize the full range of variability in both d and ∆17O – and the differences between them – should allow the decoupling of equilibrium and kinetic fractionation during evaporation, which will improve reconstructions made from water-isotope measurements. Because the signal-to-noise ratios for both d and ∆17O are generally quite small, quantifying the relationship between them will be most straightforward on timescales and at locations where variability is greatest. Existing ice-core records of ∆17O from both Greenland and Antarctica are provided in Figures 1 and 2. Figure 1c–d shows the seasonality of ∆17O and d in Greenland glacial ice and indicates a seasonal magnitude of 40–50 per meg for ∆17O. Similar seasonal magnitudes have been observed over Antarctica (e.g. Landais et al. 2012b; Schoenemann and Steig 2016). Figure 2 presents deep ice-core records of ∆17O, d, and δ18O from Antarctica, where the observed increase in ∆17O between the last glacial period and the Holocene range from four to five per meg at some coastal sites (e.g. Taylor Dome or Siple Dome) to more than 20 per meg at inland locations like Vostok. The amplitude of millennial-scale variations observed in the CFA-CRDS record from the South Pole is more than 30 per meg (Steig et al. 2021). Efforts to pinpoint the specific controls on d and ∆17O by comparing measurements with climate reanalysis products (e.g. Landais et al. 2012a) or isotope-enabled climate simulations (e.g. Dütsch et al. 2019; Schoenemann et al. 2014) will be facilitated by corresponding, high-resolution measurements of all first- and second-order water-isotope quantities, and CFA-CRDS techniques provide a method for developing those data.

Publications
Author
Mathieu Casado and Anaïs J. Orsi
PAGES Magazine articles
2023
Past Global Changes Magazine
Mathieu Casado1 and Anaïs J. Orsi1,2

Water isotopes in ice-core records are often used as a proxy of past temperature variations. Their use is based on an empirical relationship which requires care to limit the impact of the multiple contributions to the isotopic signal.


Water isotopes in ice-core records are a favored proxy of past temperature variations (Dansgaard 1964). The isotopic signal is formed by the distillation of the heavier isotopes during the advection of moist air masses from the oceanic areas, where evaporation takes place, to the precipitation sites in polar regions (Fig. 1a). As such, the temperature signal of precipitation integrates all the changes of temperatures along the pathway, following the geophysical fluid dynamic (Bailey et al. 2019). In addition, the isotopic signal is modulated by the precipitation intermittency at the ice-core drilling site (Casado et al. 2020; Münch et al. 2021), the local exchange between the snow and the atmosphere (Steen-Larsen et al. 2014; Wahl et al. 2021), the redistribution by the wind and its interactions with the local stratigraphy (Fisher et al. 1985), and snow metamorphism and diffusion inside the snow (Casado et al. 2021).

Isotopic composition of precipitation

Historically, the link between isotopic composition and temperature was assessed by spatially correlating the concurrent change of temperature and isotopic composition across Greenland and Antarctica (Dansgaard 1964). This spatial correlation was then used to convert isotopic records from ice core to past temperature records (Stenni et al. 2004) and supported by models ranging from simple Rayleigh models (Ciais and Jouzel 1994) to isotope-enabled global coupled models (iso-GCM) (Werner et al. 2018).

In a pure Rayleigh distillation model, the isotope-temperature relationship is dictated by the temperature control of the rainout fraction, under a moist-adiabatic framework (i.e. following the Clausius-Clapeyron law) (red dots in Fig. 1b). Using the spatial relationship between isotopic composition and temperature to predict the temporal relationship (space for time analogy) would work if the moisture pathways always remained the same, and the spatial gradients of isotope and temperature were evaluated directly over these moisture pathways. In reality, each precipitation event is associated with the advection of moisture air masses with different origins in terms of distance, temperature and humidity conditions (blue dots in Fig. 1b). This leads to the isotope-to-temperature relationships varying with space, time, and timescales, which is not reproduced by Rayleigh models. Yet, Rayleigh models are still heavily used for their simplicity compared to more complex models, such as iso-GCM which have shown the limits of the space-for-time analogy at timescales ranging from seasonal to multi-millennial (Werner et al. 2018). Newer distillation models using a moist-isentropic framework (i.e. not following Clausius-Clapeyron, but instead keeping the potential temperature constant) remain easy to use, and they can explain more features, such as the difference between spatial and temporal slopes (Bailey et al. 2019).

Figure 1: (A) Schematics of the acquisition of the isotopic signal in ice cores: Rayleigh distillation and diffusion; and (B) variations of isotopic composition as a function of temperature, across spatial gradients (red), and across temporal changes (blue).

Archiving of the signal in the snow

As the signal is only recorded when snowfalls occur, the precipitation intermittency creates a significant modulation of the recorded signal (Casado et al. 2020). Overall, the aliasing of the seasonal cycle by precipitation intermittency creates a white noise contribution which can be more than 10 times stronger than the climatic signal at interannual and decadal scales. This can be easily visualized using the power spectral density, i.e. the amount of energy that is included in the signal at a given frequency, which shows that the peak associated to the seasonal cycle (Fig. 2a) is redistributed across frequencies (dashed blue arrows in Fig. 2b). In glacial times, winter precipitation is suppressed, which causes an under-estimation of temperature change. This effect is particularly clear during stadial–interstadial cycles in Greenland (Guillevic et al. 2013). In modern times, interannual variability in precipitation is driven by the presence of extreme events in winter (Servettaz et al. 2020). The diagnostic of precipitation intermittency is important for the analysis of each ice-core record. Isotope-enabled global climate models are the tool of choice to quantify the relationship between atmospheric circulation patterns and the precipitation water-isotope composition. The relatively small spatial footprint (100–200 km) of precipitation events can also be used to design an optimal array of cores to average out the precipitation noise (Münch et al. 2021), and mitigate the impact of precipitation intermittency.

After the snow has been deposited, it can remain exposed near the surface for a long period of time, especially in low accumulation areas. This leads to a wide range of further alterations of the isotopic signal. Most interactions between the snow and the atmosphere, such as wind redistribution (Fisher et al. 1985), and sublimation and condensation (Casado et al. 2021; Wahl et al. 2021), also induce an aliasing of the signal, and create more white noise with a common structure of a few meters only (Münch et al. 2016). Inside the firn, snow metamorphism (Casado et al. 2021) and isotopic diffusion tend to lead to a low pass filtering of the signal, removing the high frequency variability (solid blue arrow in Fig. 2b). The impact of stratigraphic noise can be mitigated by stacking cores from a few meters apart, thanks to the very short decorrelation length of the noisy component. The diffusion effects can be numerically removed when the measurement noise is sufficiently low (Casado et al. 2020).

Limits of the isotopic paleothermometer

The variable isotope-temperature relationship, as well as these archival processes, limit the possibility for an absolute isotopic paleothermometer. Indeed, the temperature signal which can be simulated by a red noise signal (Fig. 2a), undergoes the conversion into isotopic units (green curve in Fig. 2b), and then is heavily affected by diffusion and archival noises (blue solid curves in Fig. 2b). All these effects make it arduous to obtain a calibration by matching the isotopic signal with times series obtained from weather stations (Fig. 2c; Osborn and Briffa 2004). In addition, as weather stations at ice-core study sites (Antarctic, the Arctic, high mountain regions) usually have very short record length, and under the influence of climate change, the variability against which the isotopic signal is matched does not correspond to natural variability. Overall, even with a “perfect” independent calibration, if the noisy contributions are not removed, the variability is only well estimated at low frequency (Fig. 2d).

Figure 2: Description of the isotopic paleothermometer calibration through the power spectral density (PSD) at different timescales (centennial, decadal and annual) showing: (A) climatic signal spectrum; (B) alteration of the isotopic signal (solid blue line) compared to the climatic signal during archival processes; (C) calibration against short, recent temperature records at annual and decadal timescales; and (D) independent calibration that does not match the variability at a given scale, but does not remove the noise from the signal before the climatic reconstruction (inspired from Osborn and Briffa 2004 and Casado et al. 2020).

Conclusion

Although the interpretation of water isotopes remains challenging, there are exciting new developments in the interpretation of water isotopes in ice-core records beyond the linear isotopic paleothermometer. Comparing the different noisy components amongst several ice cores can provide information on past precipitation patterns, as well as on the wind conditions and the surface roughness. Combining several isotopic compositions (d-excess, 17O-excess) also expands the scope of reconstructions from water isotopes, including the latent heat fluxes at the surface and within the firn (Casado et al. 2021). Infrared spectrometry offers new possibilities for high-resolution measurements in ice cores to study in situ post-depositional processes in the snow and in the water vapor. Better measurements will support updated trajectory models, isotope-enabled global climate models, and proxy system models, making water isotope science an expanding field of research.

Publications
Author
Yuzhen Yan
PAGES Magazine articles
2023
Past Global Changes Magazine
Yuzhen Yan

Researchers have found ice as old as 2.7 million years in East Antarctica, but such ice has a disturbed stratigraphy and a complicated age–depth relationship. Nevertheless, innovative data analyses and sampling plans can still reveal valuable paleoclimate information.


Oldest ice, but in stratigraphically disturbed form

Ice cores contain a wealth of information about the Earth’s climate history. The oldest continuous ice-core record comes from Dome C on the East Antarctic Plateau, with ice near the bottom dating back to approximately 800 kyr (EPICA community members 2004). A “holy grail” of the ice-core community is to obtain a record that extends to 1.5 Myr (Dahl-Jensen 2018), a time when the Earth’s glacial cycles showed a shorter periodicty of 41 kyr and a smaller amplitude of changes than their more recent counterparts (Raymo and Huybers 2008).

While efforts to recover a 1.5-Myr-old ice core are currently underway in East Antarctica, researchers have obtained shallow ice cores near the ice-sheet margin that date as far back as 2.7 Myr (Yan et al. 2019). The ice came from the Allan Hills blue ice areas, where sublimation and a subglacial mountain range contribute to the movement of old ice toward the surface (Spaulding et al. 2012). Yet, interpreting the data from the "oldest ice" is challenging due to disrupted age–depth relationships (Higgins et al. 2015), uncertainties related to absolute dating methods (Bender et al. 2008), and alterations to the chemical composition of the trapped gases near the bottom (Yan et al. 2019).

If classic ice-core records are like a movie, disturbed ice resembles random, separate snapshots of the film. However, innovative data treatments and sampling plans can still extract valuable paleoclimate information from these “climate snapshots.”

Asking different questions

The key to interpreting stratigraphically ordered deep ice-core records is the ability to convert depth into age continuously. This accounts for the biases due to the different accumulation rates between glacial and interglacial intervals. However, when ice stratigraphy is no longer continuous, such biases can greatly affect our observations. For example, from the Allan Hills ice dating back to 1.5 Myr, there are 33 discrete CO2 measurements yielding an average CO2 concentration of 232 ppm (Yan et al. 2019). This value does not necessarily represent the true average CO2 concentration of the 41-kyr glacial cycles due to potentially incomplete or biased sampling from the disturbed stratigraphy.

Consequently, we need to reevaluate the data through a different lens. Instead of means and standard deviations, what other types of information can we learn from the individual climate snapshots?

One answer is the range of data, defined as the difference between the maximum and minimum values. While the problem of incomplete sampling remains, now we can quantitatively estimate what fraction of the true data range a certain number of points can capture. Critical to this question are the shape of the true CO2 time-series and the relative abundance of interglacial- versus glacial-CO2 samples preserved in the ice.

In the case of Allan Hills samples, Yan et al. (2019) constructed a synthetic CO2 timeseries based on the climate records from ocean sediments. Assuming that ice from interglacial periods is over-represented by a factor of 10 in the Allan Hills ice (Yan et al. 2021), we estimated that 33 discrete samples capture 81% of the true range (95% confidence interval: 56–99%; Fig. 1). Since the observed range of the 33 CO2 samples from the 1.5 Myr ice is 65 ppm (from 214 to 279 ppm), the true glacial–interglacial range of CO2 is estimated to be between 206 and 287 ppm, assuming that glacial and interglacial extremes are equally absent.

Figure 1: (A) Synthetic CO2 time-series between 1.2 and 1.8 Myr, modified from Yan et al (2019). (B) Counts of the fraction of the recovered CO2 variability (observed range/true range) of the synthetic CO2 record by 33 discrete samples in Monte Carlo simulations (105 iterations). In each iteration, a 10-fold over-representation of interglacial periods in the ice was assumed.

By comparison, atmospheric CO2 varied between 300 and 180 ppm over the past 800 kyr. The result suggests a smaller CO2 variability 1.5 million years ago, resembling the smaller temperature variability of glacial cycles at that time. The last 800 kyr’s greater glacial–interglacial CO2 range primarily resulted from a lower glacial CO2 level (Yan et al. 2019).

Scatter plots instead of time-series

Another way to examine stratigraphically disturbed ice that is out of chronological order is to make property-versus-property plots. Measurements of different properties taken from the same depth should have a similar age regardless of the ice stratigraphy, making this approach particularly useful in evaluating the relationship between two properties of interest, such as CO2 and Antarctic temperature proxies (Yan et al. 2019).

In addition to greenhouse gases, the composition of major gases trapped in polar ice reveals useful paleoclimate information. An empirical relationship between the oxygen-to-nitrogen ratio, expressed as δO2/N2, and local summer insolation has been established for over two decades (Bender 2002). Higher insolation corresponds to lower δO2/N2 values. By using δO2/N2 as an insolation proxy, we can evaluate the relationship between Antarctic temperature and local solar forcing. The isotopic composition of the hydrogen molecules in the ice, expressed as δDice, indicates local temperature.

Over the past 800 kyr, Northern Hemisphere insolation has paced Antarctic temperature on orbital timescales (Bazin et al. 2016; Kawamura et al. 2007). In this time period, a scatter plot between δO2/N2 and δDice measured on Antarctic ice shows a positive correlation (Figs. 2a-b), which means lower Antarctic temperature was associated with higher local insolation.

On the contrary, correlation between the δO2/N2 and δDice data obtained from the 1.5- and 2-Myr ice shows a statistically significant negative correlation (Fig. 2c), meaning that a warm Antarctica occurred during intervals with high local summer insolation. The most reasonable explanation is that Southern Hemisphere insolation paced Antarctic temperature in those 41-kyr glacial cycles (Yan et al. 2023). The result here supports the hypothesis that there were no precession-related ice-age cycles 1.5 Myr ago globally, because precession forcings are out-of-phase between the two individual hemispheres (Raymo et al. 2006).

Figure 2: Relationship betewen the isotopic composition of the hydrogen in Allan Hills ice molecules (δDice) and δO2/N2 measured on ice dating back to (A) 400 ± 70 kyr (95% CI), (B) 810 ± 100 kyr, and (C) 1.5 ± 0.1 and 2.0 ± 0.1 Myr, modified from Yan et al (2023). The uncertainty associated with age arises from the fact that 40Ar dating could not fully resolve the chronology of the blue ice, so the data are binned together to form a series of “climate snapshots”.

More robust approach with large diameter samples

The examples above have a catch, however: limited ice availability means that the depths of measurements that determined gas composition differ slightly from those that constrained the age of the ice. As a result, certain assumptions had to be made about the age of the gas data, such as linking it to the nearest absolute age-control point. However, this assumption might not always hold true, since stratigraphic disturbance could occur between two age-control points.

Fortunately, engineers have developed a solution in the form of a large-diameter ice drill designed specifically for blue ice (Kuhl et al. 2014). This drill produces ice-core samples with an inner diameter of 241 mm, allowing for multiple measurements at the exact same depth. Although we still need to make the assumption that stratigraphic layers are perpendicular to the vertical axis of the cores, having an age-control point from the same depth is more accurate than attaching the data to the nearest age control points. In two austral summer field seasons (2019–20 and 2022–23), one such large-diameter drill was deployed in Allan Hills. A number of ice cores were succesfully retrieved, and analyses are currently underway.

Conclusions

Ice older than 1 Myr has been found in blue ice areas in East Antarctica, but the disturbed stratigraphy prevents the establishment of a conventional timeseries. Rather, the blue ice data represent a series of “climate snapshots”. Consequently, we must ask different questions about the data, or take advantage of the fact that properties measured at the same depth have similar age. Lessons learned from these climate snapshots preserved in the blue ice could prove useful in case stratigraphic disturbance is observed near the bottom of the future deep Antarctic ice cores dating back to 1.5 Myr.

Publications
Author
Sarah Shackleton
PAGES Magazine articles
2023
Past Global Changes Magazine
Sarah Shackleton

Marine-based reconstructions of ocean temperature have provided fundamental insight into past climate. The novel ice-core based proxy for mean ocean temperature from atmospheric noble gas ratios has demonstrated promise in furthering these insights.


Modern ocean warming

The oceans represent the largest source of thermal inertia in the climate system and play a key role in modulating the pacing of climate change. Measuring how much heat the oceans have taken up since the onset of the industrial era is important for understanding how the planet has responded to changes in our atmospheric composition driven by human activity. However, measuring total ocean warming is not an easy task, as heat uptake is spatially heterogeneous. Until recently, measurements of much of the ocean’s temperature have been sparse, and few measurements exist of ocean temperature below 2000 meters, where almost half of the ocean’s volume resides.

Probing ocean-temperature change before the instrumental era is an even greater challenge. Most of our information about past ocean temperature comes from marine-sediment records. However, as in the case of the instrumental era, ocean-temperature change at one location does not necessarily give you information about the global trend. In addition, most of our marine-sediment proxies for temperature provide information about sea-surface temperature, which represents a tiny fraction of the total ocean volume.

Changes in ocean temperature can also change its composition, including the quantity of gases dissolved in seawater. As the ocean warms, it can hold less gas, which leads to a net degassing of seawater as it surfaces and warms. This has important consequences for our atmosphere and climate; as seawater warms it can take up less CO2, leading to an increase in atmospheric CO2 and further warming.

Inert atmospheric gases trace ocean heat

This temperature dependence of gas solubility in seawater also has consequences for the inert (or non-reactive) noble gases, including krypton (Kr) and xenon (Xe). The larger the noble gas, the higher its solubility in seawater, and the stronger the temperature dependence of that solubility (Fig. 1).

Figure 1: (A) Temperature dependence of inert gas solubilities in seawater (35 practical salinity units) and (B) schematic of the noble gas mean ocean temperature (MOT) proxy.

Between the ocean and atmosphere, about 5% of xenon is dissolved in the ocean and 95% resides in the atmosphere. At the average ocean temperature (3.5°C), the solubility of xenon changes by about 4% per °C of warming. Therefore, if the whole ocean warmed by 1°C, the concentration of xenon in the atmosphere would increase by roughly 0.2%. This may sound like a tiny change, especially given that xenon has an atmospheric concentration of only 87 parts per billion. However, these small changes may be measured.

In ice cores, we can measure the ratios of xenon and krypton in air bubbles with respect to one another (Xe/Kr) or relative to N2 (Xe/N2 and Kr/N2) to reconstruct global mean ocean temperature (Headly and Severinghaus 2007). While N2 is not entirely inert and may be converted to bio-available forms via nitrogen fixation, this bio-available nitrogen represents a miniscule portion (<0.01%) of the total nitrogen in the ocean and atmosphere; even a dramatic change to the nitrogen cycle would lead to a negligible change in the total N2 inventory. We may therefore treat N2 as an inert tracer, as we do Xe and Kr.

Why is this a whole ocean thermometer?

As implied above, seawater will only degas or ingas when it is at the surface and may equilibrate with the atmosphere, so it is not necessarily intuitive why the atmospheric noble gas ratios reflect mean ocean temperature change rather than that of the surface. To gain insight, first we must understand that all of the ocean’s temperature (including water at depth) is set at the sea surface. Geothermal heating may warm the ocean from below, but this term is miniscule in comparison to the ocean–atmosphere heat flux. When a parcel of water is cooled at the surface it will take up more noble gases before it sinks into the deep. Once at depth, it conserves these properties (heat and inert gases), which are only reset once the parcel resurfaces. This means that at any given time, the total sum of dissolved xenon, krypton, and nitrogen in all these parcels of water (which controls atmospheric Kr/N2, Xe/N2, and Xe/Kr) should reflect total heat content. Because the atmosphere is mixed on annual timescales, the atmospheric noble gas ratios track ocean temperature change with no lag.

What can we learn?

Only a small number of studies have been published using this noble-gas-based technique in ice cores. However, they have shown the great potential of this proxy in probing the dynamic relationships between ocean-heat content and other components of our climate system.

  • Ocean circulation and ocean-heat content

The first timeseries of mean ocean temperature came from the West Antarctic Ice Sheet (WAIS) Divide ice core and covered the most recent transition between glacial and interglacial climate (i.e. the last 25,000 years; Bereiter et al. 2018). This transition involved mean ocean warming in two pronounced steps, both of which occurred during millennial-scale disruptions in ocean overturning. This was later confirmed by two other reconstructions of mean ocean temperature over this interval (Baggenstos et al. 2019; Shackleton et al. 2019; Fig. 2) and suggests an important link between ocean circulation and heat content.

Figure 2: Published mean ocean–temperature reconstructions for the last 25,000 years from the West Antarctic Ice Sheet (WAIS) Divide (WDC, yellow; Bereiter et al. 2018), Taylor Glacier (TG, green; Shackleton et al. 2019; 2020), and EPICA Dome C (EDC, blue; Baggenstos et al. 2019) ice-core records. Values are shown as an anomaly relative to today. Errorbars indicate 1σ uncertainty.

  • CO2 and ocean temperature

The relationship between CO2 and ocean temperature is bi-directional; CO2 warms our planet (thus warming the oceans), and due to reduced solubility, warmer oceans can hold less CO2, thereby increasing atmospheric CO2. The noble gas proxy for mean ocean temperature is well suited to probe this relationship, because the noble gases and CO2 are measured on the same archive, which makes it possible to reconstruct changes in ocean temperature and CO2 without uncertainty in the relative timing of these changes.

The links between ocean temperature and atmospheric CO2 have been probed in several recent studies. Shackleton et al. (2021) examined the role of ocean cooling in lowering atmospheric CO2 during an interval of abrupt CO2 drawdown in the last glacial cycle. Haeberli et al. (2021) evaluated the relationship between atmospheric CO2 and climate on orbital timescales by assessing mean ocean temperature and CO2 records during the interglacials and glacial maxima of the last 700,000 years.

  • Ocean temperature, sea level, and ice volume

Ocean warming contributes directly to sea-level rise through the thermal expansion of seawater. Mean ocean–temperature reconstructions may be used to quantify this direct role of ocean warming in contributing elevated sea levels during part warm intervals, such as the last interglacial (Shackleton et al. 2020).

Ice-core noble gases may provide additional constraints on past sea-level and ice volume through insights into records of the oxygen isotopic composition (δ18O) of benthic foraminifera, which is set by ocean temperature and seawater δ18O. As mean seawater δ18O is controlled by the growth and decay of 18O-depleted ice sheets, global composites of benthic-foraminiferal δ18O record changes in mean ocean temperature and global ice volume. Insight into the relative influences of ocean temperature and ice volume on δ18O may therefore be gained by comparing contemporaneous records of global δ18O and mean ocean temperature.

Conclusions

While mean ocean–temperature reconstructions in ice cores have shown promising results and offered new paleoclimatic insight, there is plenty to learn about noble-gas-based tools. Questions remain about the potential complexities of the proxy and ways in which the noble gases measured in ice cores may become decoupled from ocean-heat content. In the future, as much attention should be put into understanding the potential pitfalls of the proxy as in producing new records.

Publications
Author
François Burgay
PAGES Magazine articles
2023
Past Global Changes Magazine
François Burgay

Ice cores are unique environmental archives for reconstructing the Earth’s past climate. Each sample can contain thousands of different molecules, which, thanks to technological advances, can now be identified to gain a broader understanding of the Earth’s system.


The organic challenge

Over the last decades, analytical chemistry applied to ice cores has developed rapidly. Among the most significant innovations, there are those related to the study of elements, which until the beginning of the 21st century was extremely challenging due to their very low concentration in ice, well below the detection limit of the analytical instrumentation. Until the mid-1990s, only a few elements could be analyzed simultaneously, and the analysis of a single sample could take several hours (Barbante et al. 1997). Today, an entire ice core can be continuously analyzed for virtually the entire periodic table of elements (Erhardt et al. 2019). The same applies for organic compounds, such as pollutants or wildfire tracers that are now routinely analyzed at extremely low concentrations (Vecchiato et al. 2020; Zennaro et al. 2014). The possibility to detect and quantify organics in ice samples has opened up new opportunities, such as investigating the anthropogenic perturbation of the environment, or allowing a deeper understanding of specific environmental processes. For example, whereas common inorganic proxies for wildfires, such as ammonium, only give us information about the occurrence of a biomass-burning event (Legrand et al. 2016), organic compounds, like methoxyphenols, can also tell us the type of vegetation that burned (e.g. grasses, conifers; Müller-Tautges et al. 2016).

However, the number of organic compounds that are usually measured only represents a small fraction of the overall organic burden, meaning that the identity of the large majority of the molecules remains unknown. Indeed, most of the analytical methodologies applied so far to ice cores are defined as targeted, meaning that only specific compounds are investigated. To put this into perspective, there are more than 60 million molecules recorded in the Chemical Abstracts Service (cas.org), while less than a hundred are routinely analyzed. These compounds are typically anthropogenic markers, terrestrial and marine biomarkers and biomass burning tracers (Giorio et al. 2018). If we step outside ice-core science for a moment, we can say that the approaches adopted so far are similar to those of a person walking through a meadow with a metal detector. What that person will find are pieces of metal, but they will not see, for example, fragments of plastic or the living beings that inhabit the meadow itself. Not seeing them does not mean that they do not exist, but they are simply invisible to the eye. Coming back to analytical chemistry: target methods, although essential, give us only a partial view of the chemical space. How, then, to proceed?

Non-target screening analysis

The development of high-resolution mass spectrometers (HRMS) unlocks the possibility of exploring what was previously invisible, by simultaneously detecting up to thousands of different molecules from a single sample by providing their exact mass (Fig. 1). The application of HRMS to environmental samples is a fast-growing research field. Methods have been successfully applied to freshwater, aerosol, soil and sediment samples (e.g. Ma et al. 2022). Generally, two approaches are followed: non-target screening (NTS) and suspect screening (SUS). The former refers to the identification in a mass spectrum of those masses that are particularly relevant according to specifically defined criteria (e.g. intensity, occurrence), followed by a characterization and confirmation using reference standards. The latter refers to a slightly different approach that involves the screening of a mass spectrum for specific masses related to a defined list of molecules, followed by a characterization through ion fragmentation and comparison with reference standards.

Figure 1: Example of the application of non-target screening (NTS, dashed orange rectangle) and suspect screening (SUS, purple rectangle) approaches when coupled to liquid chromatography (LC). These approaches enable the possibility to investigate a wider chemical space than standard target methodologies (yellow rectangle).

Despite being widely used in several different environmental matrices, NTS applied to ice cores is still in its infancy, meaning that, to date, the information we can get from ice cores is still under investigation. However, results obtained from the analysis of a single sample from the Belukha ice core (Siberian Altai, 4072 masl) highlight the great potential of the application of NTS (Burgay et al. 2023). Indeed, up to 313 different compounds have been detected, the majority of which (80%) consist of carbon, hydrogen and oxygen. In addition, 7% of the molecules also contain nitrogen in their structure, while the remaining 13% contain other heteroatoms (Fig. 2). Focusing only on the most intense peaks, several carboxylic acids (e.g. succinic acid, glutaric acid, levulinic acid) and biomass burning tracers (e.g. p-hydroxybenzoic acid) were characterized.

Potentialities, challenges and future work

When applied to an entire ice-core record, the developed NTS workflow will allow scientists to understand how anthropogenic pollution has altered the aerosol molecular composition, and how the oxidative capacity of the atmosphere has changed between the pre-industrial and industrial periods. Additionally, NTS methods can be exploited for the identification of novel molecular proxies that can overcome limitations of existing proxies. For example, commonly used marine productivity proxies, such as methanesulphonic acid, may suffer from migration within the ice column, thus potentially compromising the reliability of paleoclimate reconstructions (Osman et al. 2017). However, phytoplankton also emits isoprene compounds, which can be oxidized in the atmosphere to secondary organic aerosol species (Hu et al. 2022) that are, in turn, deposited on the snow. Identifying these products using an NTS approach could provide unprecedented opportunities to test their suitability as reliable proxies for marine productivity.

Unfortunately, NTS methods are not universal. In other words, they allow the identification of hundreds or even thousands of new molecules, but many more may still be present in the ice-core samples. Coming back to the previous metaphor: NTS allows us to observe other materials as well, not just metal. However, others still remain invisible to the eye. This is due to their different chemical (polar/non-polar) and physical (volatile/non-volatile) properties. For this reason, there are no methods or instruments that can cover this wide range of molecular heterogeneity. Continuous methodological development based on well-constrained scientific questions is, therefore, essential to fully exploit the potential of NTS for comprehensive ice-core reconstructions. To date, the few available NTS ice-core methods are optimized for the detection of polar substances, i.e. those that are easily ionizable by electrospray ionization and compatible with liquid chromatography (Burgay et al. 2023; Vogel et al. 2019). However, future developments should also focus on the identification of non-polar and/or volatile substances relying on other ionization techniques and instruments.

Figure 2: Mass-to-charge (m/z) ratios plotted against the retention time. The size of the circles is proportional to the area of the molecular ions. Green circles refer to compounds consisting of carbon, hydrogen and oxygen atoms. Purple circles refer to compounds that also have nitrogen in their structure. Light blue circles refer to compounds defined as “other”, which contain other heteroatoms. Figure from Burgay et al. (2023) under CC-BY 4.0.

A further challenge is the characterization of the identified molecules. We have seen that the application of HRMS provide the exact mass of the compounds, making it possible to unambiguously define their molecular formula. However, there may be many compounds with the same molecular formula, but different structure, known as isomers. This highlights the need for additional efforts to uniquely characterize a compound, for example by comparing the fragmentation spectra and retention time of the unknowns with those of reference standards. However, standards do not exist for all molecules, especially for those formed after reactions in the atmosphere, which are of particular interest to ice-core scientists. To fill this knowledge gap, a novel approach known as aerosolomics has been recently developed (Thoma et al. 2022). In brief, chamber experiments have been carried out to determine the oxidation products of specific precursors, such as terpenes, i.e. biogenic-derived molecules. When applied to environmental samples, this strategy would link the detected oxidation products to their respective precursors, assessing their oxidative pathways. For ice-core studies, this approach can shed light on past changes of the different oxidative pathways of terpenes, as well as on the temporal evolution of the source’s strength.

As with any new scientific adventure, the difficulties and unknowns are many. However, the rewards of exploring a mysterious universe of molecules can be great. We are only at the beginning of this exciting journey, and there is much to discover.

Publications
Author
Llorenç Cremonesi and Claudia Ravasio
PAGES Magazine articles
2023
Past Global Changes Magazine
Llorenç Cremonesi and  Claudia Ravasio*

In addition to air bubbles and ions, ice cores contain insoluble particles, mainly mineral dust. These particles provide a temporal record of the atmospheric aerosol content of the past, which is key to understanding Earth’s energy balance.


In an era of rapid, global, and significant climatic changes, the scientific community is devoting great efforts to the study of the current and past major drivers of these changes. Fully understanding these changes also means studying their trends, determining their periodicity, and possibly assessing the extent to which they can be dealt with or mitigated. In this context, atmospheric aerosols have attracted much attention in the literature, as they have a significant impact on the climate system (Kok et al. 2023). Cloud cover depends primarily on the ability of water molecules to condense or crystallize around condensation nuclei, i.e. aerosols. Cloud cover is a determining factor in planetary albedo and is highly dependent on the aerosol species that populate the atmosphere. In addition, aerosols themselves contribute to the Earth&apos;s energy balance by scattering and absorbing solar and terrestrial radiation, thus cooling or warming the climate (Forster et al. 2021). While direct measurements in the atmosphere and remote sensing can be used for information about the present time, measurements of impurities in ice cores can be relied upon to obtain valuable information about the past.

Overview

In the framework of studying the Earth’s energy balance and climate, the optical properties of impurities in ice cores give us insight into the radiative properties of aerosol particles, and provide data that can be used for radiative transfer models. Light scattering, absorption, and albedo depend on many characteristics of the particles, including their size, shape, and composition, and are challenging to retrieve without direct measurements. Nonetheless, radiative transfer models still use the spherical shape approximation, which contributes to the uncertainties in the estimate of the impact of aerosols on the Earth’s energy balance. This requires some further steps toward the integrated measurement of as many parameters as possible, on an experimental basis. A continuous flow analysis (CFA) system is currently being developed in the EuroCold laboratory in Milan, oriented toward the light-scattering characterization of particles in polar and Alpine cores using single-particle extinction and scattering, an optical particle sizer, and digital holography microscopy. The aim is to characterize dust records and contribute to the monitoring of the fast evolution of climate, which is having a detrimental impact on glaciers, among other consequences.

Figure 1: Data from an Alpine ice core. (A) Two-dimensional histogram of the extinction cross-section and the optical thickness. The histogram at the bottom corresponds to the cumulative Cext distribution (arb. unit). The expected data for spheres with a refractive index of 1.55 is shown with a dashed line. Absorption is particularly high for particles giving signals above this line. (B) Vertical profiles of Cext and particle concentration are reported as a function of the core length.

What to look for with optical-based instruments, in a nutshell

A powerful system for studying ice cores is the CFA of different chemical and microphysical compounds. Significant temporal resolution can be achieved by slowly melting cores from one end to pump meltwater into in-line instruments (Bigler et al. 2011; Erhardt et al. 2019). Here we provide a snapshot of what can be observed with optical-based instruments, i.e. light scattering within the optical spectrum. The radiative properties of a particle can be predominantly traced back to its extinction cross-section (Cext), which has the units of a surface area. This is the area that effectively interacts with solar light, or any other incoming radiation. Similar definitions exist for the fraction of light that is scattered and absorbed and give the scattering and the absorption cross-sections, respectively. These optical cross-sections are determined by the conservation of radiative energy and may have little to do with the geometric cross-section of the particle, which is another parameter that may be of interest in its own right. From the scattering and the extinction cross-sections, we can assess the single-scattering albedo of an aerosol particle and make an educated guess about its refractive index. Other radiative parameters of single particles include the optical thickness and effective polarizability, related to their refractive index, shape, and size (Cremonesi 2020).

Light extinction and forward scattering

In a recent CFA campaign on an alpine ice core, we used the single particle extinction and scattering (SPES) method. With this in-line instrument, in addition to particle concentration, two optical parameters can be measured without calibration on a particle-by-particle basis (Potenza et al. 2015): the extinction cross-section and the optical thickness (Cext, ρ). These parameters tell us how much power the particle removes from the incident light and the phase lag of the wave scattered by the particle (Potenza et al. 2016). As a general rule, a larger extinction cross-section corresponds to larger particles and vice versa; similarly, optically dense particles exhibit correspondingly high optical thickness. Some spikes in the particle concentration, related to advection events, show a peculiar trend of the combination of Cext, and ρ. Figure 1 shows a snapshot of the optical parameters of a ~15 m deep Alpine ice core from the Adamello glacier (Eastern Alps, Italy). This location is affected by local natural and anthropogenic sources, in addition to the long-range transport of aerosols. Figure 1a shows the cumulative two-dimensional histogram of the extinction cross-section and the optical thickness for all the particles in the ice core. A variety of particle shapes, compositions, and sizes gives a widespread distribution along the two axes (Simonsen et al. 2018). The dashed line corresponds to the expected data for spherical particles with a refractive index of 1.55 (ranging from 0.3 to 2 μm in diameter), which is a threshold for identifying highly absorbing particles. The vertical profiles of Cext and particle concentration are reported as a function of the core length in Figure 2b. Both parameteres depend on the characteristics of the particles and the transport pathways, therefore, Cext and particle concentration do not always covariate.

Figure 2: Age–depth relationship at Roosevelt Island ice core (Lee et al. 2020; Winstrup et al. 2019). The samples of Holocene age analyzed with the holographic technique are shown as dots. (A) Hologram fringes of a mineral dust particle and (B) corresponding reconstructed image. The white bar on the top-right corner is set to 2 μm. (C) (csa, Cext) plane reported as a two-dimensional histogram and normalized on its maximum (dark blue), from a 313 m-deep sample.

Characterization by digital holography

Another technique that we integrated into our CFA system is digital holography microscopy, by which we investigate the optical and geometric properties of larger dust particles in the micrometer size range (Berg et al. 2022). As an example, we report an ice-core record from the eastern Ross Sea, as part of the Roosevelt Island Climate Evolution (RICE project, see Bertler et al. 2018; Lee et al. 2020; and Winstrup et al. 2019). In Figure 2, we show the age–depth  relationship (gray continuous line), while samples of Holocene age are identified on the timeline by different dots.

With the holographic technique we acquire the so-called hologram patterns (Fig. 2a), i.e. interferometric images where information about the size and optical properties of the particles are encoded. The holographic pattern is then processed numerically in real-time or post–measurement, without the need to check when particles are in the field of view. Moreover, multiple objects can be imaged at different focal planes simultaneously. The result of the reconstructed algorithm is an image of the silhouette of the particle, as shown in Figure 2b. We characterized insoluble particles suspended in meltwater as described in Ravasio et al. (2021; 2022). We obtained the value of Cext, the cross-sectional area (csa), and the thickness over diameter ratio (tdr) of each particle, as well as the particle count.

The importance of performing both optical and size characterization is shown in Figure 2c. Here, we show the (csa, Cext) data from one of the samples (313 m of depth), represented as a two-dimensional histogram and normalized on its maximum, and selecting only isometric-shape particles. We show with a black solid line the expected result for spherical particles from Lorenz–Mie theory (1.4–2.8 μm in diameter, refractive index of 1.55), as reference. We note that most of the data falls below this line and spans a considerable range of both csa and Cext. Indeed, many particles are plate-like with tdr values between 0.1 and 0.25, which lowers the actual Cext of the particles compared to spheres with the same geometrical cross-section.