The isotopic composition of oxygen in sea water correlates with changes in global mean sea level. Microfossils carrying oxygen-isotope signals have been used to extend sea-level records as far back as 5 million years ago. See Article p.477
On page 477 of this issue, Rohling et al.1 present a convincing approach for calculating sea-level fluctuations over the past 5 million years (Myr). Their method depends on variations in the oxygen-isotope composition of shells produced by unicellular organisms called planktonic foraminifera (Fig. 1). In this way, the authors provide much-needed information that should help to predict future rates of sea-level rise in the event of complete or partial melting of the ice caps over Greenland and Antarctica in response to global warming.
The most recent instances of ice-sheet growth and melting that generated sea-level variation of several metres took longer than hundreds of years, making it impossible to determine their effects directly from historical records. To infer how growing and melting continental ice sheets affect sea level, at least the past 500,000 years must be considered. During this time, there were five periods of sea-level rise of up to 100 m or more, corresponding to the terminations of recent glacial periods (the intervals of time within the current, ongoing ice age that, in general, correspond to colder temperatures and glacier advances).
One could argue, however, that these terminations do not adequately describe what would happen during complete melting of the Greenland ice cap and parts of the West Antarctic Ice Sheet. Hence, it is much more appropriate to consider past conditions when Northern Hemisphere glaciation was still young — that is, when small, juvenile continental ice caps controlled sea-level fluctuations on timescales of tens to hundreds of thousands of years (the timescales associated with variations in Earth's orbit, which dominate the timing of glacial–interglacial cycles and sea-level changes). But how can this be done? The shells of foraminifera offer a potential solution. Made of calcium carbonate, they contain a record of the ambient isotopic composition of seawater oxygen during the organisms' lifetime.
Since the pioneering work2 of the geologist Cesare Emiliani in the 1950s, it has commonly been accepted that periodic variations in the ratios of oxygen-18 to oxygen-16 (18O/16O) in foraminifera preserved in deep-sea sediments follow a global pattern characteristic of orbitally forced climate change during the Late Pleistocene epoch (about 700,000 to 11,700 years ago). Cool temperatures and great ice volumes both resulted in high 18O/16O ratios, whereas high temperatures and low ice volumes had the opposite effect. So, if the temperature effect can be disentangled from this isotopic record, then the remaining signal represents relative changes in continental ice volume. And if this signal can then be scaled to the amplitude of sea-level rise between glacial and interglacial periods (as has been done for the most recent postglacial period by correlating sea-level rises to the oxygen-isotope composition of cores taken from coral terraces3), then fluctuations in global sea level over time can be calculated, as long as ocean sediments provide continuous, undisturbed 18O/16O records.
Going back even further in time, a 5-Myr-long composite record4 of 18O/16O ratios from foraminifera that lived in deep-sea sediments was until now considered the best chronicle of ice-sheet volume as Earth shifted from a hot, 'greenhouse' climate (about 55 Myr ago) to colder, 'icehouse' conditions (approximately 2.6 Myr ago). However, there are two problems with this record. First, the proportions of the temperature and ice-volume effects in it are unclear, because deep-ocean temperatures may have changed substantially over this long period of climate transition. Second, it is difficult to scale deep-sea oxygen-isotope variations at orbital timescales into robust estimates (including error margins) for the amplitude of sea-level rise and fall over the past 5 Myr.
Rohling et al. overcome these problems by converting 18O/16O ratios of fossilized planktonic foraminifera that proliferated in the surface waters of the eastern Mediterranean Sea directly into sea-level variations — an approach previously developed for a study5 of the Red Sea, and which does not require temperature and ice-volume effects to be disentangled first. Their method depends on a hydraulic model of water exchange through the Strait of Gibraltar, which connects the North Atlantic Ocean and the Mediterranean Sea. This exchange mechanism not only controls the balance of evaporation and water renewal in the Mediterranean, but also strongly affects the seawater oxygen-isotope ratios recorded in planktonic foraminifera.
Assuming that there have been no major tectonic movements in the Strait of Gibraltar during the past 5 Myr that affect its depth and width, the oxygen-isotope signal from these foraminifera is simply a function of global sea-level variations relative to the modern hydraulic state of the Mediterranean Sea. The authors find that estimates of ancient sea levels relative to modern levels made using their approach are quantitatively similar to those obtained using other approaches, but have much better-defined error statistics. Accordingly, they confirm that sea level was about 10–30 m higher in the warm period that preceded the current ice age than it was when the ice sheets subsequently expanded.
As Rohling and co-workers acknowlege, their method has several limitations that require further study and improvement. Apparent mismatches exist between their results and those of hydraulic models applied to the Red Sea5, particularly for periods corresponding to the maximum volume of ice, towards the end of three of the four most recent glacial periods. This could be because the researchers' model for converting planktonic oxygen-isotope data to sea levels might not fully account for freshwater input from rivers, or for perturbations in the oxygen-isotope signal of precipitation and rivers over time. Nonetheless, the striking similarity between Rohling and colleagues' results and those from two independent studies that used data from the Red Sea5 and the southwest Pacific Ocean6 to predict sea levels during the past 0.5 and 1.5 Myr, respectively, strongly suggest that the researchers' conclusions for older periods are correct.
As expected for any new approach, the findings leave several open questions. The conclusion that the first major sea-level fall took place more than 0.5 Myr after the onset of significant global cooling 2.75 Myr ago challenges commonly held assumptions (including some of my own7) inferred from the composite record4 of 18O/16O ratios from foraminifera that lived in deep-sea sediments. But it receives a lot of support from other studies of regional palaeoclimatic trends for the period approaching the ice age in the Pleistocene epoch (roughly 2.6 million to 11,700 years ago; see Extended Data Table 2 of the paper1).
Rohling et al. also conclude that the ice volume during the Last Glacial Maximum — the peak of the most recent glacial period — was larger than that during previous glacial periods. This contradicts continental evidence for the extension of ice sheets over North America and Eurasia8, but, if correct, will serve as a crucial benchmark for improved modelling of glacial ice-sheet volumes and heights. Finally, the authors' findings should also lead to improved forecasts for the fate of extant ice caps under global warming.