Short episodes of warming and cooling occurred throughout the last glaciation. An innovative modelling study indicates that ocean-circulation changes produced much of the causative variation in greenhouse gases.
Much of what we know about abrupt climate change and tipping points in the climate system comes from polar ice cores1. But these unique data archives provide only a narrow view of the richness of climate dynamics and impacts. Moreover, the origin of the variations in the greenhouse gases associated with the pronounced climate swings during the last ice age, an interval between about 110,000 and 10,000 years ago, remains largely unknown. Hence the significance of computer models in providing a wider perspective.
On page 373 of this issue, Schmittner and Galbraith2 present climate-model simulations for an episode of abrupt climate change during the last ice age. Their results show agreement with the palaeoclimatic record3, not only in terms of physical climate variables, but also, remarkably, in changes in the greenhouse gases carbon dioxide (CO2) and nitrous oxide (N2O). The researchers conclude that the interaction of physical and biogeochemical processes in the ocean is largely responsible for the observed variations.
Schmittner and Galbraith2 use a coupled climate model of intermediate complexity suitable for palaeoclimatic studies. The physical part of the simulation features a comprehensive ocean-circulation model coupled to an energy-balance model of the atmosphere. A marine-ecosystem module, which includes two classes of phytoplankton, simulates the distribution of nitrate, phosphate, oxygen, inorganic carbon and alkalinity in the ocean. Although the model accounts for carbon cycling in the ocean, atmosphere and terrestrial vegetation, it deals with the global nitrogen cycle only in a simplified fashion. Production of marine N2O is diagnosed from oxygen concentrations, and the stratospheric sink and the soil source of N2O are assumed to remain constant.
During the last ice age, the climate system exhibited a series of rapid changes known as Dansgaard–Oeschger events. These involved temperature changes of up to 15 °C in Greenland within a few decades4. There is good evidence for the view that a major part in these events was played by the Atlantic meridional overturning circulation (AMOC), the flow of warm surface water to the far north that is balanced by the southward flow of cold water at depth. But as long as the smoking gun is missing, scientists must continue to collect circumstantial evidence from palaeoclimatic data and from model simulations to test this hypothesis. Modellers still do not have the all-encompassing climate model that would simulate a series of Dansgaard–Oeschger events in a self-contained way. They therefore need to resort to provoking such abrupt change in their models by adding and extracting large amounts of fresh water to and from the North Atlantic. The manipulation has the effect of switching the AMOC off and on, and this is what Schmittner and Galbraith do with their climate model.
The physical changes simulated by the model are in reasonable agreement with the palaeoclimatic record, although the amplitude of the temperature change in Greenland is significantly underestimated. This is probably due to insufficient sea-ice response, which is known to amplify temperature changes in the northern North Atlantic5. At the time when a reduced AMOC is responsible for cold temperatures in Greenland, temperatures in Antarctica start rising slowly, a phenomenon referred to as the thermal bipolar seesaw6. The Antarctic warming correlates strongly with the increase in the concentration of atmospheric CO2. In this model, the increase is primarily caused by a reduction in the efficiency of nutrient use by phytoplankton in surface waters of the Southern Ocean — and hence, when they die, of the transport of carbon from the surface to depth. The response is physically driven by more deep-water formation in the Southern Ocean when deep water in the North Atlantic recedes during a shutdown of the AMOC.
A notable difference between the model results and the palaeoclimatic data appears during the end of the Antarctic warming. In the temperature records, this change extends over about 2,000 years. The model's north–south coupling seems too strong, in that the abrupt resumption of the AMOC causes a similarly abrupt cooling in Antarctica. This may be due to the absence of a dynamical response in the atmosphere, and Schmittner and Galbraith's decision to keep wind patterns constant7.
The model does not capture the full complexity of the marine nitrogen cycle8; rather, a simple empirical relationship between oxygen content and changes in marine N2O production is used. So a combination of local mixing, water-mass distribution and variability in the carbon cycle determines the concentrations of atmospheric N2O. Also, the N2O contribution from soils, which is known to be important, is held constant. In spite of these pragmatic simplifications, the simulations exhibit good agreement with the palaeoclimatic record, not only in amplitude but also in temporal behaviour — at least in the first phase of the abrupt climate change.
Shutdown of the model AMOC causes a cooling in the Northern Hemisphere, which results in better ventilation of the surface waters in the eastern parts of the equatorial Pacific and Indian oceans. Better ventilation increases oxygen content, thereby reducing N2O emissions from the ocean and hence N2O concentrations in the atmosphere. The model also seems to capture a characteristic fingerprint of N2O changes during Dansgaard–Oeschger events first found in ice-core data3: longer coolings produce larger reductions in N2O. This good agreement leaves little room for an effect from N2O emissions on land, surprisingly so given that about two-thirds of the N2O emissions today come from the terrestrial biosphere.
Overall then, Schmittner and Galbraith's model2 performs remarkably well. But there is obvious room for improvement. For example, the simulation of atmospheric N2O is poor after the abrupt cooling event has ended. The ice-core record clearly shows a peak of N2O followed by a slow decrease that evolves over several centuries in the sequence of Dansgaard–Oeschger events (Fig. 1). Neither feature is captured by the model. As the major areas of N2O production are located in the eastern ocean basins, where upwelling of nutrient-rich water directly responds to the wind, one wonders whether changes in wind patterns and strength, which must have occurred during Dansgaard–Oeschger events9, might have been responsible for the rapid N2O increase at the time of abrupt warming. Also, it is inferred from the ice-core methane (CH4) record that the water cycle changed significantly during Dansgaard–Oeschger events10 because the main source of CH4 is wetlands. Therefore, a better account of soils and their changes will not only permit the simulation of CH4, but may also contribute to an improved understanding of the centennial variations in N2O that have not yet been captured in simulations.
Further progress will come with the generation of more data from isotopic studies of palaeoclimatic archives. Changes in the stable-isotope concentrations of greenhouse gases such as 13C (CH4), 2H (CH4), 15N (N2O) and 18O (N2O) are powerful fingerprints of different factors affecting the climate system and show their response to climate change. Because such measurements are so challenging, only a fraction of the information that is still locked up in polar ice cores has yet been revealed11,12. There is thus a unique opportunity for those working with physical–biogeochemical models such as that of Schmittner and Galbraith. Before more — and more detailed — isotopic data from polar ice cores become available, modellers must venture predictions of the climate changes that these data will reveal. This will be the least prejudiced approach to the problem — and so the best test-bed for our understanding of how oceanic, atmospheric and biogeochemical processes operate and interact in the climate system.
EPICA Community Members. Nature 444, 195–198 (2006).
Schmittner, A. & Galbraith, E. D. Nature 456, 373–376 (2008).
Flückiger, J. et al. Glob. Biogeochem. Cycles doi:10.1029/2003GB002122 (2004).
Huber, C. et al. Earth Planet. Sci. Lett. 243, 504–519 (2006).
Flückiger, J. et al. Clim. Dyn. 31, 633–645 (2008).
Stocker, T. F. & Johnsen, S. J. Paleoceanography doi:10.1029/2003PA000920 (2003).
Levermann, A., Schewe, J. & Montoya, M. Geophys. Res. Lett. doi:10.1029/2007GL030255 (2007).
Gruber, N. & Galloway, J. N. Nature 451, 293–296 (2008).
Timmermann, A. et al. J. Clim. 20, 4899–4919 (2007).
Dällenbach, A. et al. Geophys. Res. Lett. 27, 1005–1008 (2000).
Sowers, T., Alley, R. B. & Jubenville, J. Science 301, 945–948 (2003).
Fischer, H. et al. Nature 452, 864–867 (2008).
Flückiger, J. et al. Science 285, 227–230 (1999).
North GRIP Members. Nature 431, 147–151 (2004).
About this article
Climate of the Past (2012)