In 1947, Harold Urey pointed out that the oxygen-isotope composition of fossil seashells could serve as a paleothermometer. The more 18O a shell incorporated, he showed, the colder was the water in which it was formed. For decades now, the concentration of the heavy isotope 18O in the microshells of foraminifera—single-celled marine animals—has been widely used to reconstruct the temperature profiles of ancient seas.

Since the mid-1980s, however, models of CO2 greenhouse warming have confronted 18O data from fossil planktonic (floating) foraminifera with the so-called “cool-tropics paradox.” In stark defiance of the global climate models, 1 the planktonic 18O data seemed to suggest that 50 million years ago, a time when the CO2 level was almost certainly much higher than it is today and the Arctic was balmy enough for crocodiles and giant monitor lizards, tropical ocean surfaces were about 10°C cooler than they are now.

A new analysis of planktonic foraminifera from the late Cretaceous to the late Eocene (67-35 million years ago), by Paul Pearson (University of Bristol) and coworkers, does much to lay the troubling cool-tropics paradox to rest. 2 Because planktonic foraminifera, while they live, float at or near the surface, researchers had assumed that their 18O concentration reflects the temperature of the sea surface. But Pearson and company, doing isotopic analyses of unusually well-preserved samples of pristine shells selected with the help of electron microscopy, conclude that the surprisingly high 18O level of traditional samples is a misleading consequence of extensive recrystallization of the fossil shells in the much colder waters at the bottom of the sea. (See figure 1 and the cover of this issue.)

Figure 1. Electron micrographs contrast Eocene fossil shells of planktonic foraminifera in different states of preservation.2 The many pores of the complex structure (see the cover of this issue) belie the fact that each shell is the work of one single-celled creature. The top specimen is from a nearly pristine sample preserved in impermeable clay. The more typical bottom specimen is a related species whose pores exhibit extensive calcite recrystallization through long-term contact with seawater.

Figure 1. Electron micrographs contrast Eocene fossil shells of planktonic foraminifera in different states of preservation.2 The many pores of the complex structure (see the cover of this issue) belie the fact that each shell is the work of one single-celled creature. The top specimen is from a nearly pristine sample preserved in impermeable clay. The more typical bottom specimen is a related species whose pores exhibit extensive calcite recrystallization through long-term contact with seawater.

Close modal

Daniel Schrag (Harvard University) reached much the same conclusion a few years ago by means of a mathematical model of the recrystallization of these very porous microshells buried on the seafloor. 3 At higher latitudes, where there’s less temperature contrast between the surface and the bottom of the sea, recrystallization is less of a problem for 18O paleothermometry.

The surface temperature of the tropical oceans 50 million years ago is an issue of more than just academic concern. Before the end of this century, atmospheric CO2 levels are expected to reach, at the very least, twice the preindustrial level of 280 parts per million. To learn what this portends for our climate, one wants to look back to a time when CO2 levels generally exceeded 1000 ppm.

The fossil record makes it quite clear that, 50 million years ago, middle and high latitudes were much warmer than they are now. At the latitude of London, there were mangrove swamps, and mean annual temperatures were as much as 15°C warmer. Independent of the climatic evidence, we know from carbon-cycle models and from a variety of proxy indicators that atmospheric CO2 levels were very high during that phase of Earth’s history. One has to make do with proxy indicators like leaf stomata or boron isotopic ratios 4 because direct measurement of CO2 levels in ice-core bubbles takes us back only a few hundred thousand years.

The essential problem is that CO2 greenhouse climate models have been unable to reconcile the indisputably moderate polar and midlatitude conditions of those distant epochs with the planktonic 18O evidence that the tropical oceans were then barely warmer than the oceans at mid-latitudes. So small a latitude gradient would require an extraordinarily efficient heat-transport system that the global modelers cannot simulate. 5 At stake is the whole question of CO2 global warming prediction and its attendant policy issues.

The stable heavy isotope 18O accounts for about 0.2% of the oxygen in seawater. This slightly heavier nucleus affects evaporation, precipitation, and crystal formation. Rainwater and polar ice caps, for example, have less of the heavy isotope than does sea water. Urey argued, on thermodynamic grounds, that the formation of crystalline CaCO3 (calcite) in sea water with dissolved calcium should involve some isotopic fractionation: The fraction of 18O incorporated into the calcite should decrease with increasing water temperature, essentially because the heavier nucleus slows down the vibrational modes of the H2O molecule.

To exploit this isotopic effect for paleothermometry, one has to correct for the extent of the polar ice caps and glaciers at the time in question. From the late Cretaceous to the late Eocene, ice caps were negligible. Therefore seawater was correspondingly lighter in oxygen than it is now, when so much of the lighter water is locked up in polar ice.

There’s also the complication that different planktonic species, though they all float in the top few hundred meters, live at different depths below the surface, often changing habitat with the seasons. Some species, for example, keep to the topmost precincts because they live symbiotically with photosynthesizing algae. Not knowing the detailed life cycles of all the long-extinct creatures in these paleoclimate studies, researchers customarily deduce surface temperature for a given multi-species sample from the species that yields the lowest 18O fraction, presumably corresponding to the habitat closest to the surface.

Another heavy isotope, carbon-13, helps to differentiate habitats. “It’s also what makes our study convincing,” Pearson told us. Dissolved carbonate in the ocean becomes steadily poorer in13 C with increasing depth. That’s because organic processes, which send a steady rain of biological debris to the bottom, preferentially sequester the lighter12 C. Thus the shells of different planktonic species exhibit quite a range of13 C concentrations, correlated with the depths at which they were formed. In fact, the considerable range of13 C fractions found in traditional planktonic foraminifera samples has often been invoked as an argument against the notion that recrystallization might invalidate the 18O data. If recrystallization at the seafloor seriously corrupts the 18O results—so the argument goes—it would also largely wipe out the13 C differences between species.

To acquire samples of pristine fossil microshells, uncorrupted by posthumous recrystallization, Pearson and company availed themselves primarily of impermeable clay deposits, laid down in shallow seas off East Africa during the epochs in question and now embedded in surface rock formations on the coast of Tanzania. Electron microscopy verifies the expectation that the impermeability of the clay has left the shells essentially free of recrystallized calcite.

In figure 2, the isotopic composition of one of these pristine Tanzanian samples, 45 million years old, is compared with that of a traditionally selected deep-sea sample of about the same age from the Angola basin in the South Atlantic. By convention, the isotopic fractions of 18O and13 C are given as percentages in excess of the isotopic fractions in a convenient laboratory calcite standard—the so-called Vienna Pee Dee Belemnite standard. (The Pee Dee is a river in South Carolina.)

Figure 2. Isotopic compositions of two different samples of fossil planktonic foraminifera shells suggest conflicting tropical sea surface temperatures 45 million years ago. The oxygen-18 excess, δ18O, relative to a laboratory calcite standard, is plotted for various species against the carbon-13 excess, δ13C. Common species are connected by lines. δ18O serves as a paleothermometer. The well-preserved Tanzanian sample (red) suggests a sea about 15°C warmer than does the Angola basin sample (blue), which is compromised by posthumous recrystallization on the ocean floor. Variations within each sample reflect the different depths at which species floated. The recrystallization metamorphosis appears to converge on a point corresponding to inorganic calcite made entirely on the ocean floor.

Figure 2. Isotopic compositions of two different samples of fossil planktonic foraminifera shells suggest conflicting tropical sea surface temperatures 45 million years ago. The oxygen-18 excess, δ18O, relative to a laboratory calcite standard, is plotted for various species against the carbon-13 excess, δ13C. Common species are connected by lines. δ18O serves as a paleothermometer. The well-preserved Tanzanian sample (red) suggests a sea about 15°C warmer than does the Angola basin sample (blue), which is compromised by posthumous recrystallization on the ocean floor. Variations within each sample reflect the different depths at which species floated. The recrystallization metamorphosis appears to converge on a point corresponding to inorganic calcite made entirely on the ocean floor.

Close modal

The 18O excess, corrected for the absence of polar ice caps at the time, translates into the temperature at which the shells were formed. The more 18O, the colder the water. Assuming that the range of temperatures in the pristine Tanzanian sample reflects the different depths at which the various floating species resided, the authors take the highest temperature—about 30°C—to have been the temperature at the surface. The traditional Angola basin sample in figure 2, presumably corrupted by evident recrystallization, suggests a much chillier surface temperature of about 16°C.

The13 C range of the Tanzanian sample is another manifestation of the vertical range of the habitats of different species. The lines connecting the same species in the two samples shows the13 C range shrunk to half its pristine width. The argument that both the oxygen and carbon isotopic changes are due to recrystallization is strikingly bolstered by the observation that the changes appear to be converging on the point in the figure that represents the expected isotopic composition of purely inorganic calcite precipitated at the seafloor.

Because the traditional Angola basin sample is about halfway between the pristine sample and the convergence point on this plot of carbon and oxygen isotopic fractions, the Pearson group concludes that roughly half the calcite in the fossil microshells of the traditional sample has been replaced, over the eons, by recrystallized material.

The paper concedes that the two samples may have been somewhat different from the start. Unlike the planktonic foraminifera of the Angola basin, those in the Tanzanian sample lived in shallower coastal waters. Nonetheless, the authors argue, the new results should help to free the discussion of global warming from the cool-tropics paradox that has bedeviled paleoclimatology. “For those parts of the late Cretaceous and Eocene epochs that we have sampled, tropical temperatures were at least as warm as they are today, and probably several degrees warmer.”2 If it were otherwise, we would have trouble believing any of the CO2 greenhouse models.

1.
E.
Barron
,
Paleoceanography
2
,
729
(
1987
) .
L. C.
Sloan
,
D. K.
Rea
,
Paleogeogr. Paleoclimatol. Paleoecol.
119
,
275
(
1995
) .
2.
P.
Pearson
 et al,
Nature
413
,
481
(
2001
) .
4.
P.
Pearson
,
M.
Palmer
,
Nature
406
,
695
(
2000
) .
5.
L. C.
Sloan
,
J. C. G.
Walker
,
T. C.
Moore
 Jr
,
Paleoceanography
10
,
347
(
1995
) .