PerspectiveClimate Change

A Hotter Greenhouse?

See allHide authors and affiliations

Science  18 Jul 2008:
Vol. 321, Issue 5887, pp. 353-354
DOI: 10.1126/science.1161170

Scientists have long been puzzled by the fact that mid-to-high latitude continental interiors and the poles in the Eocene (55 to 34 Million years ago) were much warmer than today, without freezing winters (1), while tropical sea surface temperatures (SSTs) were apparently near-modern (2, 3). Mechanisms proposed by climate modelers to maintain high-latitude warmth require substantial tropical temperature increases (4, 5). The implication is that fundamental gaps remain in our understanding of climate dynamics. Many hypotheses have been advanced to resolve this paradox: Paleoclimate proxies require reinterpretation, boundary conditions need improvement, or a major mechanism is missing from climate models. However, no proposal has led to a simple, general solution. A resolution might be in sight based on efforts to develop better climate proxies and multiproxy Eocene records. But this resolution may present new challenges.

Most of what we know about Eocene SSTs comes from records of the oxygen isotopic composition (δ18O) of planktonic foraminiferal calcite shells that reflect the temperature of shell formation and other factors (3, 5). Reconstructing temperatures using (δ18O of foraminifera requires knowing or assuming the δ18O of the seawater in which they grew. This composition changes globally as terrestrial ice volume varies, and regionally with changes in net evaporation. Estimated evaporation is, in turn, affected by SST. Interpretation is complicated by ambiguity about the depth range at which foraminifera were calcifying and by seasonal biases.

Even more troubling, the δ18O of planktonic foraminiferal shells may be diagenetically altered once they reach the seafloor (that is, in the early phases of burial, secondary calcite can precipitate from the much colder sediment pore-water fluids onto or into the shell itself). Alteration can push SST estimates toward cooler (bottom water) values (6). Alteration is enhanced by increased rates of pore-water flow and increased burial temperature and pressure (6). Thus, the best preservation is expected from relatively impermeable sediments that have not been buried deeply or exposed to high temperatures.

Pearson et al. (7) sought and found records from clay-rich regions in the hope of identifying whether previous records were biased. In a series of pioneering studies (710), they recovered foraminifera from shallowly buried impermeable clays in several regions, most notably in Tanzania. When viewed under a light microscope, the recovered foraminifera were translucent or “glassy,” as modern samples are, unlike the “frosty” or “chalky” appearance of typical foraminifera used in most previous analyses. Under a scanning electron microscope, the frosty shells were revealed as recrystallized. Analysis of δ18O in the glassy foraminifera indicated temperatures warmer by 5° to 10°C than previous reconstructions (8). The implications are that the shells in many reconstructions from open-ocean sediments were altered and that much of the original isotopic SST signal had been overprinted by cold deep water trends. Acceptance of this viewpoint is, however, not universal (11, 12). Focus has shifted to using other proxies, such as Mg/Ca and TEX86, to develop independent SST records.

The Mg/Ca ratio of foraminiferal shells is a function of temperature and offers a paleoclimate record independent of seawater δ18O. However, the exact relationship is species-specific; depth, seasonality, and alkalinity have influences, the global value of seawater Mg/Ca must be measured or modeled as it varied in time, and Mg/Ca values may also be diagenetically altered, although perhaps to a lesser degree than δ18O (8).

Hot tropics.

Recent studies suggest that during the Eocene, tropical temperatures were much higher than they are today. The resulting heat stress may have led to tropical vegetation die-offs, with profound implications for climate and the carbon cycle.


By careful site, sample, and calibration selection, these uncertainties can be reduced, but the global seawater Mg/Ca ratio is relatively unconstrained, imposing temperature uncertainties of ±4°C during the Eocene. For reasonable choices of seawater Mg/Ca composition, the temperature agreement between δ18O and Mg/Ca is highest in glassy foraminifera (8), indicating that they are more likely to represent the true SST. Furthermore, Mg/Ca-derived SSTs are much warmer than those derived from the δ18O of frosty or chalky (presumably altered) foraminifera in open ocean sites (8, 13). Unfortunately, comparison at individual sites of δ18O and Mg/Ca still reveals discrepancies of 2° to 9 °C in glassy foraminifera, giving a plausible range of SST from 25° to 34°C at one site (8). Clearly, another independent proxy is needed to further improve estimates.

TEX86 is a promising new proxy for mean annual SST, based on organic molecules derived from crenarcheota (14). The technique is independent of surface freshwater balance and seawater composition. Pearson et al. used this approach in their Tanzanian record, and their three proxy records showed impressive congruence (8, 9), providing a more credible interpretation than any single proxy. But TEX86 is imperfect (15) and novel enough that important kinks have not been worked out (16). The method has been difficult to calibrate at the hot (>30°C) temperatures central to Early Eocene climate paradoxes. Various studies, including that of Pearson et al. (9), used a previously developed ad hoc calibration using different slopes at high and low temperatures, clipping maximum temperatures.

Recent incubation (14) and core-top (15) studies resulted in a new calibration for TEX86 that is linear up to 40°C, which raises interpreted peak tropical SST by ∼5°C from those originally published using TEX86 (9). Thus, a newer interpretation (see supporting online material) for the warmest Eocene suggests tropical SSTs in the 35° to 40°C range, not the 33° to 28°C range published in 2007 (9), or the 25° to 30°C range as thought a decade ago (3), or the 20° to 25°C range accepted two decades ago (2).

Half of Earth's surface area is in the tropics, so changes and uncertainties in tropical temperatures dominate any climate sensitivity estimate. If SSTs were truly ∼35°C at times in Tanzania (19°S) or New Jersey (∼30°N), some tropical regions must have been much hotter. This has thought-provoking implications for paleoclimate, vegetation, and carbon cycle evolution.

First, tropical temperatures above 31°C offer no evidence for a climate thermostat, that is, a strict mechanism that maintains tropical SSTs in the modern range; climate dynamicists trying for decades to explain thermostats may have been chasing a chimera. Second, climate models might be able to reproduce warm poles and warm extratropical continental winters, given that these new tropical SSTs imply closer to modern temperature gradients (5).

Third, during the warmest parts of the past 65 million years—that is, the Paleocene-Eocene Thermal Maximum (PETM) and subsequent brief, sudden “hyperthermal” phases of the Early Eocene Climate Optimum (17)—tropical vegetation may have been above the upper limits of its thermal tolerance (18). Most plants, especially the C3 plants that comprised Eocene floras, have physiological mechanisms that break down in the 35° to 40°C range (18, 19); in particular, they can die because photorespiration dominates over photosynthesis (1820). Annual mean temperatures greater than 35°C can be plausibly reconstructed to have been widespread equatorward of 35° latitude (8, 9, 21, 22), so floras may have been thermally stressed, and perhaps undergoing water stress in the warmest intervals. There is some evidence of tropical floral extinctions during the warmest periods (23, 24), while forests thrived at higher latitudes.

This scenario may be a missing link in the hypothesis (25) that carbon cycle and climate changes during the PETM were caused by oxidation of the terrestrial biosphere. It is well established that a major tropical vegetation die-off in a global warming world has profound temperature, precipitation, and carbon feedbacks (20). Carbon cycle modeling (26) suggests that the terrestrial carbon pool could have been much larger than modern, ∼6000 gigatons of carbon. The gradual warming preceding the PETM may have loaded a terrestrial carbon storage gun, and crossing the 35°C threshold may have triggered it. Tropical die-back after an initial warming (22) might have added thousands of gigatons of carbon into the atmosphere and further increased temperatures by radically reducing evapotranspirative fluxes that normally cool tropical landmasses. Tropical heat death helps resolve two mysteries: the magnitude of the carbon and climate excursion at the PETM, and the fact that these abrupt warmings occur during broader intervals of extreme warmth, rather than in cold intervals as expected from methane degassing (27).

The recent results suggest that, rather than being a stable cradle for tropical life, the tropics may have been a crucible; during warming, many taxa may have been forced to flee poleward, innovate, or face extinction (28). These far-ranging implications are a lot to place on the narrow shoulders of the few published proxy records, but they highlight the importance of the next challenge: collecting more tropical multiproxy records and establishing the accuracy of existing ones.

Supporting Online Material

Fig. S1



  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
View Abstract

Stay Connected to Science

Navigate This Article