Now, a team from Harvard University, US, has found that changes in the variability of sunlight and precipitation could be as important for the carbon sink properties of forests in the northeastern US as changes in these factors' mean values. The more that sunlight and rainfall varied, the less carbon the forest ecosystem sequestered. In addition, broad-leaved deciduous trees thrived relative to their conifer cousins. Variability in temperature had only a minor impact, however.

"Our results show that terrestrial ecosystems and carbon budgets are highly sensitive to environmental variability, and so a knowledge of this variability is necessary for understanding ecosystem structure and functioning," David Medvigy, who is now at Princeton University, told environmentalresearchweb. "In particular, we showed that it was possible to determine a quantitative relationship between carbon uptake, sunshine variability and precipitation variability for one key temperate forest. We expect that similar relationships will hold at other forests."

According to the researchers, conifers are more affected by conditions in the spring than deciduous trees, which are only just beginning to come out of their dormant state. Typically, spring sees particularly variable conditions; this affects the amount of photosynthesis by conifers. Photosynthesis uses light less efficiently as solar irradiance increases.

To obtain their results, Medvigy and colleagues used the Ecosystem Demography biosphere model together with ten years' worth of carbon-flux measurements and hourly meteorological data from Harvard Forest in Massachussetts. The ecosystem model simulates the physiological functioning, growth, life, death and recruitment of individual plants, and of the whole forest ecosystem, they say. The team also used climate models to analyse how future changes in climate variability will affect carbon fluxes.

Climate change is projected to affect not just the mean values of factors such as temperature, precipitation and sunshine but also their variability. "Our work shows that projected changes in high-frequency variability may impact terrestrial ecosystems and carbon budgets as strongly as projected changes in the mean state," said Medvigy. "An important part of what will determine future land carbon uptake and in some cases ecosystem composition will be the degree of future high-frequency environmental variability. It is critical for climate models to simulate this variability as realistically as possible."

Modelling carbon uptake of ecosystems without sufficiently taking variability into account can lead to overestimates of as much as one-third, the team found.

Medvigy's original aim had been to understand the observed interannual variability in terrestrial carbon uptake at Harvard Forest. "To make this problem as simple as possible, I had carried out a few model simulations in which I left the interannual environmental variability intact while filtering out the high-frequency environmental variability," he explained. "To my surprise, these simulations suggested that that filtering [of] high-frequency variability had a really large effect on terrestrial carbon uptake. This really fascinated me, and so I decided to switch my focus from interannual variability to high-frequency variability."

A key next step will be to examine the sensitivity of other ecosystems, such as tropical forests, boreal forests and agricultural regions, to changes in high-frequency environmental variability, to assess the global response of the terrestrial biosphere.

The researchers reported their work in PNAS.