Human activities are largely to blame for the rise in atomspheric carbon dioxide that has seen global temperatures climb since the mid-twentieth century. But higher temperatures also cause more CO2 to be released into the atmosphere through the natural processes of the carbon cycle. This feedback loop may play an important part in amplifying anthropogenic warming. However, determining the magnitude of such an effect has been a challenge.

“We know that anthropogenic CO2 is having an effect on climate,” says David Frank, a climatologist at the Swiss Federal Research Institute WSL in Birmensdorf. This is mainly because CO2 traps heat from the Sun's rays, so the more we release into the atmosphere, the more heat is stored. But climate models have estimated that the feedback between the carbon cycle and climate could contribute anywhere between 0.1 °C and 1.5 °C per year on top of the rise in temperature due to direct anthropogenic emissions.

It has been difficult to precisely quantify the sensitivity of the carbon cycle to changes in temperature, partly, Frank says, because “during the past century the feedback relationship between temperature and CO2 has been obscured by the massive amounts of CO2 released by human activities. We therefore need to look at changes in temperature and CO2 over a longer timescale, before the industrial revolution.”

Other groups had already examined the relationship between temperature and CO2 during pre-industrial times, explains Frank, but they typically based their calculations on a single reconstruction of temperature over time and a single CO2 record. “It is as though you want to determine the average height of a population by measuring the height of one or two individuals,” says Frank. Just as variations in individuals' heights could skew the average one way or the other, preferentially using data sets that indicate a small temperature variation and a large change in CO2 would result in calculating a large feedback between the carbon cycle and temperature.

The solution that Frank and his colleagues came up with was to use every piece of data they could get their hands on. That meant combining data from nine large-scale temperature reconstructions and CO2 records obtained from three Antarctic ice cores. After poring over “a heck of a lot of data”, they were able to calculate more than 200,000 estimates of how CO2 varied in response to temperature between 1050 and 1800. “These estimates take into account the uncertainties of reconstructions,” explains Frank. “If the data were perfect, we might not need so many estimates.”

From these estimates, the researchers were able to calculate probability distributions and then determine a median value for the magnitude of the carbon cycle's sensitivity to temperature. That value, 7.7 parts per million by volume of CO2 per 1 °C, gives some idea of how much the ocean and terrestrial ecosystems will amplify anthropogenic actions by in the future (see page 527). “These values are similar to those obtained from climate models, although models with lower feedback might be slightly more accurate,” says Frank. But he adds a caveat. “We don't know whether additional processes that have not been operating or significant in the past will have an important role in the future.”