We don’t have direct CO2 records earlier than 800,000 years ago. So scientists use what are called 'proxies' – indicators that give us indirect estimates of the amount of carbon dioxide in the atmosphere. Advances in scientific techniques have given us increasingly good proxies, meaning we can be more confident that our estimates of CO2 levels hundreds of millions of years ago are about right. There are anomalies: the various different proxies don’t always provide similar results. A paper published this week goes a long way to removing one important anomaly. It shows that the proxy that uses estimates of CO2 concentrations based on isotope levels in ancient soil carbonates may have over-recorded atmospheric carbon dioxide levels at some periods. This brings the figures into line with other measures. Why does this matter? In Mesozoic times, from about 250 to 80 million years ago, the soil carbonate proxy has previously suggested very high CO2 levels in the atmosphere. Some studies had concluded that the air had several thousand parts per million of carbon dioxide, peaking perhaps at fifteen times higher levels than today. But records suggest that the temperature was only a maximum of 10 degrees Celsius higher than today. If the standard 'carbon cycle' models are correct, the estimated CO2 concentrations ought to have produced warmer conditions. The importance of this new research is that it shows that adjusted soil carbon proxies indicate that Mesozoic CO2 concentrations probably never rose above 1,500 parts per million, a figure consistent with other proxies and with the probable temperature levels at time. This is one more important indication that ancient CO2 levels were strongly correlated with climate.
We have a reliable record of atmospheric carbon dioxide dating back the best part of a million years. Ice cores from the Antarctic contain trapped air bubbles that can be analysed to show levels of CO2 right back to about 800,000 years ago. Ice core analysis isn’t perfect but Antarctic data gives us estimates of recent CO2 levels that are very similar to other techniques. We can be reasonably confident that the ice cores are giving us a good fix on atmospheric CO2.
But the ice cores can only take us back so far. What about ten or a hundred million years ago? We can't directly measure CO2 in the atmosphere of the time in the way that we can with bubbles in ice. We have to use other techniques, such as examining the stomata of leaves left in fossil records. Stomata breathe in the CO2 that plants need for photosynthesis but they also increase the rate of water loss. When CO2 is abundant, plants don’t need as many of these pores and so reduce the number in their leaves. Stomata density on fossil leaves is therefore an increasingly important proxy for atmospheric CO2 concentrations.
Another of the four or five standard proxy measures is provided by the ratio of carbon 12 to the carbon 13 isotope embodied in calcium compounds in ancient soil. This ratio depends on the relative levels of CO2 in the atmosphere and on the amounts of carbon dioxide being respired by organic matter in the soil itself. Previous research had not adequately studied how the CO2 in the tiny pores of the soil – a mixture of carbon dioxide from the atmosphere and from the biological processes in the soil – varied during the seasons of the year. The new paper gives the results from experiments that showed how carbon is only absorbed into calcium compounds during periods when the soil is dry and the temperatures high. The effect is to reduce the assumed typical atmospheric levels of CO2 during the Mesozoic period. This work appears to be quite simple: science should be slightly embarrassed that it has taken until today to provide a good calibration of the effect of atmospheric CO2 levels on the rate of take-up of the two isotopes carbon 12 and carbon 13 in soil carbonates in varying soil conditions.
The new research gives evidence that previous measures of soil carbonate overstated the likely level of CO2 in the atmosphere hundreds of millions of years ago. The authors provide an estimate that maximum carbon dioxide levels never exceeded about 1,500 parts per million, slightly less than four times less than at present and five times pre-industrial levels. This is broadly in line with estimates from other proxies such as leaf stomata and far more plausible than the figures of 5,000 ppm or so estimated from previous work using soil carbonate levels.
Most climate models assume that a doubling of CO2 in the atmosphere will add about 3 degrees Celsius to average global temperatures. Therefore a four times multiplication will add about 6 degrees. Add in the temperature rise already embedded because of the CO2 increase since the start of industrialization and the new research suggests that typical temperatures in the later Mesozoic area would be estimated to be about 7 to 8 degrees higher than today. The actual figure was probably slightly higher, with temperature levels at the poles even more elevated. The paper provides corroboration for the other estimates of CO2 levels in ancient times.
If supported by further research, this paper removes an important anomaly. The late Mesozoic average temperature no longer appears to be lower than would be expected. As time passes, advances in experimental science are consistently showing that the temperature record conforms closely to what the climate models suggest. The wider importance of a result such as the ones in this paper is that the alternative explanations for climate variability – cosmic ray incidence, variations in the earth's orbit, changes in the sun's energy output, etc. – all appear to be increasingly less plausible as hypotheses explaining the gross changes in climate in the past. No one doubts that they all have some impact; it is just that CO2 variations increasingly appear to be a fuller explanation.
 D. O. Breecker, Z. D. Sharp, and L. D. McFadden, 'Atmospheric CO2 concentations during ancient greenhouse climates were similar to those predicted for A.D. 2100', Proceedings of the National Academy of Sciences, 106.52 (29 December 2009) http://dx.doi.org/10.1073/pnas.0902323106 (doi:10.1073/pnas.0902323106).