by Jean Jouzel
The results of these “tedious calculations” were published in April 1896. Arrhenius’s primary motivation was to demonstrate that the cooling that corresponded to the beginning of ice ages could be explained by a decrease on the order of 40% of the ability of CO2 to absorb the “obscure heat” of Joseph Fourier. And thus to refute Croll’s astronomical theory, which proposed a different angle.
We can understand glaciologists’ motivation to attempt to determine if such a decrease in atmospheric CO2 had indeed taken place during the glacial periods. It was nevertheless necessary to wait until the beginning of the 1980s for our quite natural curiosity to be satisfied. Since then specialists of air bubbles have worked extremely hard on the CO2 record during the past and on other greenhouse effect gases—methane, CH4, and nitrogen protoxide, N2O—which are also influenced by human activity. The analysis of carbon 13 in carbon dioxide, but also of deuterium in methane, later proved capable of helping us understand the origin of the variations of the compounds and the way in which their natural cycles function or are disturbed by human activity. These scientists didn’t stop there; they began to explore the isotopic compositions of other molecules present in the air, in particular those of its primary components, nitrogen, oxygen, and argon.
Let’s take the case of nitrogen and argon, whose isotopic compositions we know have remained constant in the atmosphere over millions of years. We note, nevertheless, that they vary in air bubbles. Jeff Severinghaus, an American researcher, has even shown that these changes were abrupt at the time when the temperature of sites with high snow accumulation rates altered rapidly.4 These variations are explained by the combination of isotopic fractionations, one linked to gravity—there are slightly larger quantities of heavy molecules at the bottom of the névé—and the second to thermal origin, which results from the difference in temperature between the surface of the ice sheet and the base of the firn. Consequently, the analysis of nitrogen and argon isotopes offers a method, rather surprising but precise and thus widely used, to estimate changes in temperature when they were relatively rapid due to the effect of thermal fractionation. Temperature measurements carried out in the bore holes of Greenland had shown that the estimates deduced from the composition in oxygen 18 of the ice underestimated by half the cooling of the Last Glacial Maximum there. The analysis of air bubbles has since shown that such an underestimation there, linked to the change in seasonality of the snowfalls, was systematic throughout the last glacial period.
The oxygen 18 in the air trapped in the ice is modified by these fractionations, but the greatest variations observed have a completely different origin.5 They reflect both the changes in isotopic composition of the seawater, which essentially depends on the quantity of ice amassed on the continents, and the modifications of the continental biosphere and the hydrological cycle in the low latitudes strongly affected by variations in insolation. One can be free of the influence of the latter and thus have access to past variations in sea level.
Still looking at the air but now just the amount of it and not its composition, the idea is simple: the air pressure at the moment the bubbles were enclosed, and thus the quantity trapped, decreases with altitude; in principle it would suffice to measure the volume of air trapped to know how the altitude of glacial ice caps has varied over time. The use of this paleoaltimeter over several coring sites, coupled with the modeling of the flow of ice, has revealed that during glacial periods the ice on high plateaus of the ice sheets was thinner due to lesser snowfall, whereas the regions more on the periphery had thicker ice and a lower sea level, which enabled the ice cap to extend farther. Putting this simple idea into play, however, became unexpectedly difficult. Surprisingly, the volume of air also appears to depend on insolation, whose variation may influence the metamorphism of the snow during the formation of the firn and thus the mechanism of trapping, processes that also seem to affect the proportion of oxygen to nitrogen. This is very interesting from the point of view of dating.6
The Headaches of Dating
Compared to the dating of continental and oceanic archives, ice dating has its own specific issues. The first comes from the fact that some of the information we are interested in is inscribed in the ice, some in the air bubbles. Because the air is younger than the ice—since air trapping occurs gradually, these ages are averages—we need two distinct chronologies. Further, ice cores hardly lend themselves to the use of radioactive methods. Carbon 14 dating is generally less precise than other methods for dating ice cores. It can be used in exceptional cases, for example, if there is vegetal debris present—this is the case in a core from the Andes. In addition, carbon 14 dating is inapplicable beyond a few dozen thousand years.
What methods, then, do glaciologists use to establish the chronology of ice core samples, the oldest of which, on the Antarctic site of Dôme C, spans 800,000 years? Here, too, the simplest solution involves counting annual layers, because many of the properties of the snow—concentrations of deuterium and oxygen 18, electrical conductivity, which depends on the content in impurities, composition of many chemical elements—differs depending on whether they accumulated in the summer or the winter. Furthermore, in the summer dust particles are more abundant in the snow because the winds are more favorable at that time to their being carried to the poles: the corresponding layers diffuse more light. The visual observation of the successive layers of snow, on the walls of pits several meters deep dug for the purpose of counting and then dating the recent layers of snow, has been used a great deal, and the stratigraphic variations sometimes remain visible at a great depth. None of the indicators is perfect, but taken together they enable an annual dating as long as the thickness of the layers, which become thinner as we dig into the ice sheet, remains sufficient. Annual dating has been possible in the coastal regions of Antarctica, and more especially in Greenland where these indicators have been applied to date ice as old as tens of thousands of years.
Some events can, if they are dated in other ways, help establish the chronology of ice core samples and validate a dating obtained by counting layers. This is true of the generally unambiguous volcanic eruptions that we can detect by analyzing physical properties, such as the electric or chemical conductivity (marking the presence of sulfates) of the ice. This also true of certain deposits characteristic of ash layers.
Recent large, well-documented eruptions, combined with spikes in radioactive fallout elements produced during nuclear arms testing in the 1950s and 1960s, are useful landmarks. Further back in time, the process is reversed: thanks to precisely dated coring in Greenland, the chronology of volcanic eruptions during the Holocene has been established. Even further back, major eruptions whose dates we know within a few thousand years using radiometric methods, such as that of Toba, which occurred around 75,000 years ago, are used to verify the chronology of deep ice cores. Similarly, we can use a spike in the cosmogenic isotopes fallout, beryllium 107 and chloride 36, from around 40,000 years ago. According to the most widespread hypothesis, this spike would have a connection with the Laschamps event, which corresponds to an abrupt decrease in the geomagnetic field.
The analysis of beryllium 10 offers other possibilities. The production of this isotope depends on the intensity of cosmic rays and on that of the geomagnetic field, but it also varies with solar activity, a factor that seems dominant for the Holocene. These mechanisms are identical to those that rule the production of carbon 14. Even if the processes of deposit differ, there are remarkable similarities between the variations of carbon 14, deduced, for example, from the analysis of tree rings, and those of beryllium 10 in Antarctic ice. Grant Raisbeck and Françoise Yiou, researchers at CNRS in Orsay, have thus established an absolute chronology of Antarctic ice over the last 7,000 years.8
Going back further, glaciologists have developed models that enable them to calculate the thickness of successive annual layers. These layers become gradually thinner as one goes deeper because of the effect of the flow of the ice, and their thickness
is equal to the initial value multiplied by the thinning. The establishment of glaciological dating associates a flow model that provides thinning and a model of the history of surface accumulation that varies depending on the climate. It is weaker in glacial periods, as masses of colder air that bring precipitation over the ice caps then contain less water vapor. The age is calculated by counting the number of years that have elapsed between two successive depth levels, then by adding them starting from the surface. This method has been applied to all deep ice cores in Antarctica. It is also useful to estimate the difference in age between the ice and the trapped air. The depth at which the pores close up and trap the atmospheric air is all the greater when it is cold and the accumulation is great. The model developed by Jean-Marc Barnola in Grenoble and his team calculates this depth and the associated difference in age; in a glacial period it can reach 8,000 years.
These glaciological models have the disadvantage of being less precise as the age increases, and it is then indispensable to have points of reference (often called pinpoints) deduced from comparison with ocean records, themselves dated by orbital adjustment. It might appear more reasonable to achieve this orbital adjustment over variables recorded in the ice cores themselves. Periods linked to the Earth’s orbit are visible, at least in the sampling of Antarctica, when one analyzes the concentration of deuterium in the ice, those of oxygen 18 and methane in the air bubbles, and, as we’ve mentioned, the volume of air trapped and the relationship of oxygen to nitrogen in the air bubbles. Thus, the oxygen 18 in the air, studied by Michael Bender’s team from Princeton University, varies with a periodicity of around 20,000 years, which is directly connected to precession.9 However, these orbital chronologies generally have the disadvantage of being based on the hypothesis of a constant interval—we speak of constant phase relationship—between the considered property and the variations in insolation, a hypothesis whose validity has not been established. Conversely, the mechanisms that we qualify as nonlinear are certainly in play when the climatic system is complex, mechanisms that are translated by the variable phase relationship between the insolation and one of those parameters, which these orbital chronologies do not take into account.
This type of problem is raised in many domains, notably for geophysicists who have developed “inverse” methods that prove well adapted for dating ice core samples. In that case, it is a matter of associating the information provided by several dating methods by taking into account the uncertainty surrounding each of them. This approach, taken by Frédéric Parrenin, consists of adjusting, with the help of numerical processes, the values of the parameters of the glaciological model in such a way that once it is fixed, the dating obtained agrees as much as possible with all the available chronological information.10 This inverse method, which has the advantage of not imposing a constant phase relationship, is no doubt the best adapted and the most flexible in the absence of observed seasonal variations that alone enable a year-to-year dating. Nevertheless, the ages it provides are tainted by a relatively large uncertainty—a few thousand years—as soon as one moves away from the period for which well-dated layers exist. This inverse method has been adopted for large-scale deep ice core records in Antarctica where the accumulation is very small, whereas the counting of layers remains the favored approach in Greenland.
New approaches are emerging that will allow us to achieve an absolute dating of long records of glacial-interglacial cycles: the volume of air contained in the bubbles, a parameter that we have already mentioned, and the oxygen-nitrogen ratio, whose very weak variations appear to have a connection with the variations in summer insolation at the drilling site.11 Although the physical processes involved need to be confirmed, these two parameters are promising markers that record summer variations at the site where ice was formed, variations whose age we can know with great precision using astronomical calculations.
We will now look at the detailed history of these deep ice cores.
CHAPTER 6
The Campaigns
Copenhagen, July 22, 1952. Willi Dansgaard carefully collects the rain produced by a storm occurring over the capital of Denmark. He is interested in the oxygen 18 composition of the successive samples. If it is true, as we learned in school, that the formula for water is written H2O, that which we drink, as we have mentioned, contains different isotopic molecules, H216O, H218O, HDO, and so forth. The different isotopes of water were identified in the period between the two world wars, and the variations in their concentration in a natural environment were quickly proven. But Dansgaard’s study opened the path to a systematic exploration of the distribution of water isotopes in precipitations. The article he published in 1964 on his findings,1 and which is the study’s high point, is still often cited. One of the remarkable aspects it reveals is the relationship between the temperature at the site and the isotopic composition of precipitations. The colder it is, the weaker the number of H218O molecules as compared to those of H216O, and inversely. The variations are minor but largely sufficient to be detected thanks to the development of mass spectrometry, which highlights the fact that the H218O molecules (mass 20) are heavier than those of H216O (mass 18). With the help of a simple model Dansgaard explains this observation, but more important, he was one of the first to measure its potential applications, which logically must give access to temperatures that existed in the past, as long as it is possible to have access to past precipitations.
Camps Century and Byrd: The First Deep Ice Core Drillings
The idea was simple, and it was natural for Dansgaard to propose that it should be applied to Greenland, a Danish land by tradition, but today an autonomous territory. That ice sheet, more than three kilometers thick in the central regions, contains ice that is increasingly old as one digs into it. But digging into the ice—obtaining samples—is no small feat. Fortunately the icebergs that calve around Greenland offer an alternative solution because they are formed by more or less ancient ice depending on whether they come off the center or the edges of the ice cap. A campaign led in 1958 by Willi Dansgaard enabled the sampling of an entire series of icebergs and the dating of them, thanks to the content of carbon 14 in the air that was extracted from them. The oldest sample, approximately 3,000 years old, was formed in the central regions, and its very weak content of oxygen 18 indeed reflects the very cold temperatures that existed there related to high altitude. A French team, led by Étienne Roth at the CEA Saclay, was also very interested in those icebergs. This team was in charge of checking the concentration in deuterium during the enrichment processes leading to the production of heavy water, D2O, and thus developed methods to analyze the content of deuterium, which it proposed to apply to the Greenland icebergs.2 The result they obtained wasn’t surprising. As anticipated from the fact that fractionation processes are similar for HDO and H218O, the content of deuterium and oxygen 18 varied in parallel, but thanks to this study the French geochemists were henceforth involved in the polar adventure.
There was not a lot of hope, however, of going back to the distant past using icebergs. The only means to do so was to dig into the ice sheet, and this objective was made possible thanks to support from across the Atlantic. In Hanover, New Hampshire, a laboratory of the U.S. Army, the Cold Regions Research and Engineering Laboratory (CRREL), was dedicated to the study of cold regions. Polar ice caps, snow-covered surfaces, permafrost, and sea ice were the main areas of interest of this remarkable lab. One of its teams was involved in the development of core drills, instruments that enabled glaciologists to extract cylinders of ice a dozen centimeters in diameter and one to three meters long and to go back through successive “stages” to increasingly deep ice. Further, the CRREL had access to large military transport planes, C130s, which were able to transport from the U.S. East Coast large quantities of materials to Greenland. There, about a hundred miles from Thule, the military base Camp Century (Figure 1.1) was established by the U.S. Army Corps of Engineers in 1959, a true city under the ice with roads, houses, a mov
ie theater, a church, and a network of trenches and tunnels, the longest of which, Main Street, stretches along 350 meters. Everything was powered by nuclear energy.
For obvious logistical reasons, this was the site chosen for the first deep ice core drilling in a polar region. The operation in which the Danish team and that of the CRREL, led by Chet Langway, participated was a success. The rock base was reached at a depth of 1,390 meters in July 1966—not without difficulty, however, as it took six successive seasons to achieve that depth. New methods for analyzing the physical characteristics of the ice, its electrical conductivity, and the impurities that it contained were put into place. But it was the analysis of the oxygen 18 of this ice, undertaken in Copenhagen, that delivered the clearest message. The bottom of the drilling around 1,300 meters had less oxygen 18 than the first 1,100 meters, and these results were beyond doubt. This was ice from the last glacial period, whereas the upper part was formed during the Holocene, in the last 10,000 years or so. A glaciological model enables scientists to estimate the age of the ice accounting for the gradual thinning of the layers as they become deeper. Whereas the first 1,000 meters represent around 6,000 years, the last 200 cover around 100,000. All of the last glacial period is indeed there but increasingly packed down the closer one gets to the rock base. The interpretation of the results was not simple, but that year, 1966, was momentous. With Camp Century, glaciologists had the first deep core drilling to use to reconstitute our climate’s past.3 It was just in time because the following year Camp Century had to be closed due to relatively rapid movement of the ice, which made it impossible to maintain the camp. Fortunately the nuclear plant was moved in time.