Introduction
Reconstructing global climate and atmospheric carbon dioxide concentrations on multi-million year timescales has been an elusive goal for geoscientists. Various proxies and models suggest that warm climate periods and high levels of atmospheric CO2 coincide whereas cool periods correspond to lower atmospheric CO2 levels1, 2, 3, 4. This has been taken as evidence that the climate evolution over million-year timescales is mainly controlled by natural variations in CO2 production and consumption. However, some temperature reconstructions have contradicted this view5, 6. The most striking example for the suggested temperature–CO2 decoupling lies in the late Jurassic to early Cretaceous periods, when the records indicate cool conditions7, despite proxies and models predicting a warm greenhouse climate accompanied by high atmospheric CO2 concentrations1. On page 43 of this issue, Fletcher and co-workers8 analyse the carbon isotopic ratios of fossil plants and suggest that late Jurassic atmospheric CO2 levels are much lower than previously believed.
Stable carbon isotope measurements from fossilized soils formed in the Jurassic and Cretaceous periods suggest high levels of atmospheric CO2 (ref. 3). Models of the global carbon cycle confirm this interpretation and produce a 100 million year period of warm stable climate1. However, the distribution of coals, desert and salt deposits, glacial materials, and tropical soils suggest a much more variable climate with at least one pronounced cool episode6. Marine records seem to confirm the occurrence of a cold interval during the late Jurassic and early Cretaceous5, 9. This apparent decoupling of atmospheric CO2 and climate has led to alternative explanations for climate forcing during this time, such as changes in the flux of galactic cosmic rays reaching the Earth5.
However, climate–CO2 decoupling need not be assumed as there is a weak link in the chain of arguments for the Jurassic–Cretaceous high-CO2 greenhouse. Over geological timescales, levels of atmospheric CO2 (pCO2) are reconstructed with models based on the marine record of 87Sr/86Sr ratios. Strontium ratios respond to the hydrothermal and volcanic release of CO2 as well as its consumption through silicate-rock weathering — the main source and sink mechanisms, respectively — with high 87Sr/86Sr ratios suggesting a low pCO2 and vice versa. The models yield high pCO2 values during the Jurassic period1, 9, in part because the marine 87Sr/86Sr record over the past 545 million years reaches a well-defined minimum during the Jurassic period that could indicate high rates of volcanic emissions or low rates of rock weathering. The low 87Sr/86Sr ratios could, however, also be produced by weathering of young volcanic deposits10, 11. In that case, the low Jurassic 87Sr/86Sr ratios would indicate a low, rather than high, pCO2 because CO2 is rapidly consumed by the weathering of young volcanic rocks.
Fletcher and co-workers8 take the effect of the weathering of young volcanic rocks on pCO2 and the marine strontium record into account in their solution to the Jurassic paradox. They used the carbon isotopic ratios of fossil liverworts to reconstruct changes in pCO2 over the Jurassic and Cretaceous periods. On the basis of the fossils, they conclude that in the late Jurassic to early Cretaceous, pCO2 was much lower than previously believed. The group also independently simulated Jurassic and Cretaceous pCO2 levels using a model based on the 87Sr/86Sr record that accounts for the effect of weathering of young volcanic rocks on the strontium records. This also results in lower than expected pCO2 values during the late Jurassic and early Cretaceous periods, in agreement with earlier modelling efforts10, 11. Of the existing pCO2 records, those from liverwort most closely match the simulation, although a reconstruction of CO2 from the pores of fossil leaves also yields largely similar results5.
The liverwort CO2 reconstruction, when converted into temperature, is broadly consistent with the carbonate-based surface temperatures that had sparked earlier debate5, and it shows a positive correlation with surface temperatures derived from pH-corrected 
18O values measured in marine carbonate fossils4, 9 (Fig. 1). This correlation might be fortuitous, or it may suggest that ancient pCO2 records are better preserved in liverworts than in other archives such as fossilized soils and leaves that have previously been used to reconstruct atmospheric compositions3.
Figure 1: Jurassic and Cretaceous surface temperature reconstructions.
Temperatures derived from the liverwort-based reconstruction of atmospheric carbon dioxide levels (open squares)8 are substantially lower than those from fossil soils (closed triangles)3, when both sets of pCO2 proxy data were converted into average global surface temperatures using the GEOCARB III equation1. Temperatures from the pH-corrected marine oxygen isotope records (red solid line4 and blue dotted line9) broadly agree with the liverwort reconstruction. Relatively warm and cool periods are shown in the bar at the top6; the timescale is taken from ref. 13.
Full size image (20 KB)Carbon isotope reconstructions from liverworts yield much lower pCO2 values and lower surface temperatures than from fossilized soils over the Jurassic to Cretaceous periods. If they prove to be a reliable archive for ancient pCO2 values, the new results raise questions about the validity of the previously used proxies. Deviations between reconstructions from liverwort and fossil soils could, however, be due to rapid changes in pCO2 that are as yet beyond the temporal resolution of the records. For example, the Jurassic and Cretaceous periods could have generally experienced a warm greenhouse climate that was punctuated by multiple discrete cooling events, probably accompanied by rapid changes in pCO2 (ref. 3). More liverwort data need to be generated to cross-check the new method with other more established proxies.
Fletcher and co-workers' reconstructions suggest that the Earth's long-term climate history is indeed controlled by atmospheric CO2 concentrations. However, the various proxies for both CO2 levels and temperatures over million-year timescales do not yet provide an unambiguous picture, leaving models of the global carbon cycle poorly constrained. This is not necessarily surprising; after all, even the modern global carbon cycle is not fully understood. Important feedbacks that may have far reaching consequences for our understanding of past and future climate change12 are still being discovered.
It is important for geoscientists to continue to improve proxies and models that help reconstruct the ancient global carbon cycle, while taking into consideration the results of those working on modern carbon cycling and future climate change.
