Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. 2002 Oct 1; 99(20): 12567–12571.
Published online 2002 Sep 16. doi:  10.1073/pnas.202304999
PMCID: PMC130500

Low atmospheric CO2 levels during the Permo- Carboniferous glaciation inferred from fossil lycopsids


Earth history was punctuated during the Permo-Carboniferous [300–250 million years (Myr) ago] by the longest and most severe glaciation of the entire Phanerozoic Eon. But significant uncertainty surrounds the concentration of CO2 in the atmosphere through this time interval and therefore its role in the evolution of this major prePleistocene glaciation. Here, I derive 24 Late Paleozoic CO2 estimates from the fossil cuticle record of arborsecent lycopsids of the equatorial Carboniferous and Permian swamp communities. Quantitative calibration of Late Carboniferous (330–300 Myr ago) and Permian (270–260 Myr ago) lycopsid stomatal indices yield average atmospheric CO2 concentrations of 344 ppm and 313 ppm, respectively. The reconstructions show a high degree of self-consistency and a degree of precision an order of magnitude greater than other approaches. Low CO2 levels during the Permo-Carboniferous glaciation are in agreement with glaciological evidence for the presence of continental ice and coupled models of climate and ice-sheet growth on Pangea. Moreover, the Permian data indicate atmospheric CO2 levels were low 260 Myr ago, by which time continental deglaciation was already underway. Positive biotic feedbacks on climate, and geotectonic events, therefore are implicated as mechanisms underlying deglaciation.

Keywords: fossil plants‖paleoclimates‖stomata‖carbon isotopes

Paleobotanical approaches for quantitatively estimating the pre-Quaternary CO2 content of the Earth's atmosphere exploit the inverse relationship between leaf stomatal index (SI, fraction of epidermal cells that are stomata) and CO2 shown by modern trees (1). However, their application has focused on the Cretaceous and Early Tertiary when extant species are represented in the fossil record (2). Attempts to extend this approach further back into the Paleozoic era with the genus Ginkgo are much less secure because of the need to switch to related taxa, such as pteridosperm genera, whose SIs lie outside the range of the calibration dataset (3). These considerations severely limit the utility of the plant fossil record for evaluating the role of CO2 as a greenhouse gas during the Late Paleozoic glaciation (4), a time when quite large discrepancies are evident between geochemical modeling studies of the long-term carbon cycle (5), and paleobiological proxy CO2 measurements (3, 6).

Here, I report 24 quantitative Carboniferous and Permian atmospheric CO2 estimates derived from the stomatal characteristics of fossil arborscent lycopsids, the spore-bearing vascular plant group dominating equatorial Euramerican and Cathaysian coal swamp floras during the Late Palaeozoic (7). The fossil lycopsid SIs were calibrated against the response to atmospheric CO2 variations of three geographically distinct contemporary tropical Lycopodium cernuum populations, taken to be a reasonable nearest living equivalent (NLE) taxon among extant lycopsid species (8). Phylogenetically, Isoetes represents the nearest living relative to members of the Lepidodendron, Bothrodendron, and Sigillaria etc. However, modern Isoetes species have either submergent, or emergent growth habits, which are unlikely to sensitively record global changes in atmospheric CO2 levels. L. cernuum fulfils those criteria of an NLE set out previously (8) because it occupies a comparable ecological setting to its Late Paleozoic arborescent counterparts, with both groups of organisms being characterized by a subtropical/tropical distribution, a preference for open, disturbed habitats, an erect growth habit, and possession of the same microphyll leaf structure. In addition, stable carbon isotope analyses of the organic carbon sources were used to more directly test the similarity of leaf gas exchange metabolism between the extinct and extant lycopsids before estimating paleo-CO2 levels with the ancient lycopsids.

Materials and Methods

Stable Carbon Isotope Analyses.

Stable carbon isotope analyses were performed on herbarium shoots of L. cernuum and fossil lycopsids shoots assigned to the Lepidodendraceae, Sigillariaceae, and Bothrodendraceae. Fossil materials were handled with forceps, etched from the rock surface, finely powdered, and reacted with 38% HCl for 24 h, to remove possible carbonate contamination. Samples were then resuspended in deionized water, and centrifuged, and the residual pellet was dried for 72 h at 80°C. Stable carbon isotope ratios were measured by introducing 1 mg of powdered samples with a PDZ Europa ANCA-GSL preparation module into a PDZ Europa (Cheshire, U.K.) 20–20 stable isotope ratio mass spectrometer. Replicate analyses with the same instrument on the same specimens had a reproducibility of 0.5‰. All values are expressed as [Rsample − (Rsample/Rstandard) × 1,000], where Rsample is the 13C/12C ratio of each sample and Rstandard is the Pee Dee Belemnite standard.

Fossilized bulk organic carbon is characteristically isotopically less negative by 1–2‰ compared with modern materials, due to the preferential loss of 13C-rich compounds during diagenesis (9). Therefore all fossil plant δp measurements were corrected by a conservative 1‰ to account for this effect and facilitate comparison with isotope ratios of herbarium samples.

To isolate isotopic variations due to the biological characteristics of the organisms (10, 11), isotopic discrimination (Δ) was calculated from (δa − δp)/(1 + δp/1,000), where δa is the isotopic composition of atmospheric CO2, estimated from marine calcium carbonate fossils (12). This approach assumes δa is regulated on a long-term basis by equilibrium with marine inorganic carbon, with a 7‰ negative offset. Carbon isotope stratigraphies for the interval 320–300 million years (Myr) ago differ somewhat between Panthalassan and Paleotethyan regions, probably due to changes in oceanic circulation associated with the closure of the seaway between Laurussia and Gondwana (12). Here, a mean marine inorganic carbon value from both regions (+5‰) was used, giving δa = −2.0‰ for the Westphalian. The ratio of atmospheric CO2 in the leaf intercellular air spaces (ci) relative to the air surrounding the leaf (ca), was then calculated from Δ by rearrangement of the well-validated model linking Δ to leaf gas exchange (10) by a + (ba) × ci/ca, where a and b are the fractionations associated with diffusion (4.4‰) and carboxylation by Rubisco (27‰), respectively.

Paleozoic Atmospheric CO2 Reconstructions.

A training set was developed with historical time series of leaves of L. cernuum collected since the early 19th century, as atmospheric CO2 levels increased (13), and from multiple altitudinal transects encompassing spatial gradients of atmospheric CO2 partial pressure. Leaves of L. cernuum were obtained from populations originating in Central America, tropical South America, and the Malaysian Peninsula of southern Asia to assess the response to a global CO2 signal and account for possible genotypic heterogeneity in the response. The inverse regression transfer function describing the collective SI response (SI,%) of L. cernuum populations (Fig. (Fig.1)1) is given by: CO2 (Pa) = [(SI − 52.259)/ −1.2305]. It is conceivable that leaves of L. cernuum respond to CO2 levels above the global mean, because of the potential for soil-respired CO2 to accumulate beneath tropical forest canopies. However, CO2 measurements in tropical forests (14) indicate that at 1–2 m above the soil surface, the height attained by L. cernuum, this effect is small (<5%). The occurrence of L. cernuum on disturbed, exposed sites suggests these plants grew under irradiance levels similar to those experienced by arborescent lycopods, which with their “pole habit” and unusual canopy architecture (7), may have experienced irradiance levels characteristic of an open habitat.

Figure 1
Relationship between SI (fraction of epidermal cells that are stomata) and atmospheric CO2 partial pressure of different tropical populations of L. cernuum. Solid line denotes the linear regression, dotted lines indicate the ±95% confidence ...

To quantitatively reconstruct atmospheric CO2 levels in the Late Paleozoic, SI determinations were made from stomatal and epidermal cell measurements on fossil cuticles of a wide range of tropical lowland arborescent lycopsids from European mid-to-late Carboniferous (1519) deposits. Estimates of CO2 in the Permian were based on published SI values of equatorial arborescent lycopsids from deposits in North Chinese upper Permian sections (20) (Table (Table1).1). The modern training set captures 70% of the SI values of fossil cuticles (Table (Table1)1) despite differences in stomatal placement (leaf cushions of arborescent lycopsids rather than leaves of L. cernuum). Observations for Ulodendron majus and Ulodendron landsbergii suggest the SI of leaves is 20–30% lower than that of leaf cushions (Table (Table1).1). If this effect is systematic across all taxa, it translates into an underestimation of atmospheric CO2 levels of 7–10%. Paleo-CO2 values have been expressed in this article as concentration, since at sea level, where the Carboniferous and Permian tropical swamp forests were located, CO2 concentration is equivalent to partial pressure (1 Pa ≈10 ppm).

Table 1
Stomatal characteristics of fossil Carboniferous (C) and Permian (P) lycopsid microphyll leaves and leaf bases

Results and Discussion

Physiological Comparisons.

Comparative stable carbon isotope analyses allow a first-order test of the assumed physiological similarities between extinct and extant lycopsid taxa by providing an indicator of the set point of photosynthetic metabolism (11) between the two groups of organisms. It emerges that the range of carbon isotope discrimination (Δ) values of fossils (Table (Table2)2) is within the variation displayed by the three geographically separated tropical populations of L. cernuum (Fig. (Fig.2).2). With Δ values of ≈20–22‰, the arborescent lycopsids display operational set points of leaf gas exchange activity similar to those of L. cernuum, which indicates a rather low (<27%) diffusional limitation on photosynthesis, a likely consequence of their relatively high SIs (Table (Table2).2). The isotopic data suggest that photosynthetic activity and stomatal control of water loss in contemporary L. cernuum populations is comparable to that of the ancient Carboniferous lycopsids, despite obvious differences in scale between the two groups of organisms. It should be noted, however, that in one respect the comparison is imperfect because L. cernuum in this study experienced a drop in CO2 partial pressure as a function of altitude (i.e., falling total pressure) whereas the fossils grew at a time when total atmospheric pressure was similar or slight higher than now. If the comparison is taken to be reasonably robust, a further implication is that the physiology of lycopsids has apparently been highly conserved, despite the potential for over 300 Myr of evolutionary change, in a group that also shows a considerable degree of phylogenetic conservatism (21).

Table 2
Carbon isotope characteristics of Carboniferous compression fossil lycopsid shoots
Figure 2
Comparison of the carbon isotope discrimination ([open triangle]) of modern (□) and extinct (■) lycopsids. Each box delineates the 25th and 75th percentiles, errors bars show the fifth and 95th percentiles, the horizontal line represents the ...

Paleozoic CO2 Reconstructions.

Quantitative paleo-CO2 reconstructions obtained by calibration of the fossil SI values show consistent and tightly constrained values for the Carboniferous and Permian (Table (Table1).1). The assumed CO2 signal in the fossil lycopsid cuticles is stable within the Namurian (330 and 320 Myr ago) and the Westphalian (320–300 Myr ago) stages of the Carboniferous and gives mean global values of 379 ppm and 336 ppm, respectively. These results are supported by two qualitative estimates of CO2 obtained by analysis of fossil conifers dating to the Carboniferous and early Permian (22). Within the Westphalian, the record exhibits a strong degree of self-consistency, with leaves of four extinct families of arborescent lycopods all giving similar CO2 estimates (Table (Table1).1). Low atmospheric CO2 levels prevailing during the Late Carboniferous are reconstructed for the Permian, with lycopsids from North China yielding a mean value of 313 ppm between ≈270 and 260 Myr ago (Table (Table11).

The paleoatmospheric CO2 values for the Carboniferous span 30 Myr and are consistent with low levels of net radiative forcing (23, 24) and the occurrence of major ice sheets on Gondwana (25, 26). Further, the CO2 reconstructions are tightly constrained and have an order of magnitude greater precision (95% confidence limits ± ≈35 ppm) than carbon cycle models and paleosols, which are subject to larger (±200 ppm and ±500–1,000 ppm, respectively) error terms for this interval (27). They therefore narrow a significant uncertainty in the role played by atmospheric CO2 during the Permo-Carboniferous glaciation (4, 23, 25, 26). My CO2 estimates are in agreement with the lower set of values predicted by geochemical carbon cycle modeling studies (5) and reconstructed from isotopic analyses of Late Paleozoic North American paleosols (28) (Fig. (Fig.3).3).

Figure 3
Late Palaeozoic evidence for ice (summarized from refs. 23, 25, and 27) and atmospheric CO2 concentrations reconstructed from paleosols and predicted by a geochemical model of the long-term global carbon cycle (solid line). Red boxes indicate CO2 estimates ...

Implications for the Mechanisms of Deglaciation.

The Permian CO2 data shed some light on the role of atmospheric CO2, as a greenhouse gas, in the closing phase of the Late Paleozoic glaciation, a subject of considerable debate (36). Paleosol studies suggest low CO2 concentrations (150–200 ppm) 280 Myr ago (28), and all five Permian lycopsid SIs indicate that atmospheric CO2 levels were also low between 268 and 260 Myr ago with values reconstructed to be ≈313 ppm (Table (Table1).1). These results are generally supported by geochemical modeling (5), but are lower than CO2 estimates from Permian paleosols in Central India dating to 260 Myr ago (540–890 ppm) (29). They fail to support the contention (3, 6) that atmospheric CO2 levels were high around 260 Myr ago. However, the interpretation of high Late Permian CO2 levels (1,000–2,000 ppm) reconstructed from paleosols (6) depends on the isotopic composition of marine carbonates (30). Calculated with the paleosol CO2 barometer by using an alternative marine isotopic record (12) to that of ref. 31 results in a considerable damping of this CO2 rise, with values lower by some 700–2,500 ppm (27). High Permian atmospheric CO2 levels inferred from the SI of peltasperm seed ferns (3) appear to be compromised by dating imprecision, taxonomic uncertainties, and inadequate replication (32).

If atmospheric CO2 concentrations were low toward the close of the Permian, when a variety of geological data suggest deglaciation was well underway (23, 25, 26), then climatic forcing factors other than high levels of CO2 in the atmosphere may need to be examined to determine the cause of the Late Paleozoic deglaciation. Evidence for plate tectonic activity indicates that the southern Gondwana drifted across the South Pole during the Carboniferous and Permian, 320–260 Myr ago (33). This clockwise rotation of the plates would have exposed the Gondwana supercontinent ice masses to latitude-controlled warmer climates, leading directly to melting of the ice sheets (26), without the need to invoke changes in the concentration of greenhouse gases. Moreover, as Glossopteris floras spread into areas previously covered by ice (34), a positive feedback would have ensued due to seasonal reductions in land surface albedo, as the ice became replaced with a forested landscape. The switch in land cover from ice to forest has been found in contemporary climate modeling studies to be an effective promoter of strong high latitude climatic warming (35). In fact, numerical modeling simulations of early Permian climates lack representation of this vegetation feedback and it is notable that they have yet to satisfactorily reproduce the distribution of fossilized biota in the high southern latitudes, even when high (8 × times modern) atmospheric CO2 concentrations are prescribed (36).


I thank R. A. Berner, W. G. Chaloner, J. E. Francis, C. P. Osborne, D. L. Royer, C. Wellman, and F. I. Woodward for comments and discussion, A. Paul for access to the herbarium of the Natural History Museum (London), P. Taylor, P. Sheppard, and M. Howe for access to the Kidston collections at the British Geological Survey (Nottingham), and R. Dunn for stable isotope determinations. I gratefully acknowledge funding of this work through a Royal Society University Research Fellowship and the Leverhulme Trust.


stomatal index
million years


This paper was submitted directly (Track II) to the PNAS office.


1. Woodward F I. Nature (London) 1987;327:617–618.
2. Royer D L, Wing S L, Beerling D J, Jolley D W, Koch P L, Hickey L J, Berner R A. Science. 2001;292:2310–2313. [PubMed]
3. Retallack G J. Nature (London) 2001;411:287–290. [PubMed]
4. Crowley T J, Baum S K. Geology. 1992;20:507–510.
5. Berner R A, Kothavala Z. Am J Sci. 2001;301:182–204.
6. Ekart D, Cerling T R, Montanez I P, Tabor N J. Am J Sci. 1999;299:805–827.
7. Phillips T L, DiMichele W A. Ann Missouri Bot Gard. 1992;79:560–588.
8. McElwain J C, Chaloner W G. Ann Bot. 1995;76:389–395.
9. Benner R, Fogel M L, Kent Sprague E, Hodson R E. Nature (London) 1987;329:708–709.
10. Farquhar G D, Eherlinger J R, Hubrick K T. Annu Rev Plant Physiol Mol Biol. 1989;40:503–537.
11. Ehleringer J J. In: Ecophysiology of Photosynthesis. Schulze E D, Caldwell M M, editors. Berlin: Springer; 1994. pp. 363–392.
12. Mii H S, Grossman E L, Yancey T E. Geol Soc Am Bull. 1999;111:960–973.
13. Friedli H, Lötscher H, Oeschger H, Siegenthaler U, Stauffer B. Nature (London) 1986;324:237–238.
14. Broadmeadow M S J, Griffiths H. In: Stable Isotopes and Plant Carbon-Water Relations. Ehleringer J R, Hall A E, Farquhar G D, editors. San Diego: Academic; 1993. pp. 109–129.
15. Thomas B A. Ann Bot. 1967;31:775–782.
16. Thomas B A. J Nat Hist. 1968;2:425–428.
17. Thomas B A. J Nat Hist. 1967;1:53–60.
18. Thomas B A. Palaeontology. 1977;20:273–293.
19. Thomas B A. Palaeontology. 1970;13:145–173.
20. Wang Z, Chen A. Rev Palaeobot Palynol. 2001;117:217–243.
21. Wikstrom N, Kenrick P. Mol Phylogen Evol. 2001;19:177–186. [PubMed]
22. McElwain J C, Chaloner W G. Palaios. 1996;11:376–388.
23. Crowley T J, Berner R A. Science. 2001;292:870–872. [PubMed]
24. Hyde W T, Crowley T J, Tarasov L, Peltier W R. Climate Dynamics. 1999;15:619–629.
25. Frakes L A, Francis J E. Nature (London) 1988;333:547–549.
26. Crowell J C. Pre-Mesozoic Ice Ages: Their Bearing On Understanding The Climate System. Boulder, CO: Geological Society of America; 1999. , Memoir 192.
27. Royer D L, Berner R A, Beerling D J. Earth Sci Rev. 2001;54:349–392.
28. Mora C I, Driese S G, Colarusso L A. Science. 1996;271:1105–1107.
29. Ghosh P, Ghosh P, Bhattacharya S K. Palaeogeogr Palaeoclimatol Palaeoecol. 2001;170:219–236.
30. Cerling T E. Global Biogeochem Cycles. 1992;6:307–314.
31. Veizer J, Ala D, Azmy K, Bruckschen P, Buhl D, Bruhn F, Carden G A F, Diener A, Ebneth S, Godderis Y, et al. Chem Geol. 1999;161:59–88.
32. Kerp H. Nature (London) 2002;415:38. [PubMed]
33. Powell C, McA, Li Z. In: Permian-Triassic Pangean Basins and Foldbelts Along the Panthalassan Margin of Gondwanaland. Veevers J J, Powell C M, editors. Boulder, CO: Geological Society of America; 1994. , Memoir 184, pp. 5–9.
34. Chaloner W G, Lacey W S. Special Papers Palaeontol. 1973;12:271–289.
35. Bonan G B, Pollard D, Thompson S L. Nature (London) 1992;359:716–718.
36. Rees P M, Ziegler A M, Gibbs M T, Kutzbach J E, Behling P J, Rowley D B. J Geol. 2002;110:1–31.

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...