Researchers in Switzerland and Germany have analysed data stretching back 1000 years to get the best estimate yet of how changes in global temperature affect the biosphere's ability to soak up carbon dioxide. The team found that this feedback coefficient is about five times smaller than previously expected – which suggests that the amplification of manmade global warming by carbon-cycle feedback will be less than previously thought. Furthermore, the reduced uncertainty of this latest result could lead to better predictions of climate change caused by increasing amounts of carbon dioxide in the atmosphere.

To understand climate change, scientists need to know how changes in global temperature affect the amount of carbon dioxide in the atmosphere. Rising temperatures could, for example, turn a green landscape into a desert, which would reduce that region's ability to absorb carbon dioxide. Conversely, a warmer climate could lengthen the growing season in mid and high latitudes, increasing the absorption of carbon dioxide in these places. Changes in temperature could also affect the amount of carbon dioxide produced by the vast numbers of micro-organisms in soil.

The overall effect of this "feedback" is expected to be positive – higher temperatures leads to less carbon dioxide being absorbed, which means more of the gas in the atmosphere, which in turn makes the climate even warmer. However, scientists have struggled to get a precise value for the feedback coefficient – a process that involves studying historical carbon dioxide and temperature data.

Large uncertainty

The best estimate had been that a rise in the mean global temperature of one degree boosts the carbon dioxide concentration by about 40 parts per million by volume (40 ppmv/°C) – but this could be off by 30 ppmv/°C or more. Such a large uncertainty makes the prediction of future carbon dioxide levels – and therefore future temperatures – all the more difficult. Indeed, 40% of the uncertainty in such predictions can be attributed to carbon dioxide feedback.

Now, David Frank and colleagues at the Swiss Federal Research Institute in Birmensdorf, the University of Bern and the Gutenberg University in Mainz have performed the most comprehensive analysis of carbon dioxide and temperature data yet. The team studied the period 1050–1800 AD, when manmade emissions were small enough to be ignored. Carbon dioxide levels were determined from three Antarctic ice cores. Average temperatures in the northern hemisphere were derived from nine different "proxy reconstructions" of temperature – average temperatures derived mostly from tree rings and the isotopic content of ice cores.

Two distinct periods

Frank and colleagues conclude that the feedback coefficient is probably about 2–21 ppmv/°C, with 8 ppmv/°C being the median value. The team also found that the coefficient appears to be significantly different in the periods 1050–1549 and 1550–1800 – when it was about 4 and 16 ppmv/°C respectively. The former era corresponds roughly to the "medieval warm period" and 1550–1800 to the "little ice age" – which experienced different patterns of global temperature and precipitation. As a result, Frank and colleagues believe that the shift from one period to the next could have decreased the carbon storage capabilities of some parts of the globe.

The research excludes the previously accepted value of 40 ppmv/°C with a confidence of 95% and provides further evidence that the coefficient is positive rather than negative. The latter is important because it suggests that the biosphere will not be able to soak up manmade emissions of carbon dioxide, which are believed to contribute to global warming.

The work is described in Nature 463 527.