The concentration of carbon dioxide in the atmosphere undergoes seasonal, cyclic variation, the amplitude of which has increased by up to 50% in the Northern Hemisphere over the past 50 years1,2. Several factors have been proposed to explain this increase3,4,5, including the response of the terrestrial biosphere to climate change, increased fossil-fuel emissions, and changes in oceanic fluxes and atmospheric transport of CO2, but the relative magnitude and latitudinal contribution of each are still debated. In two studies published in this issue, Gray et al.6 (page 398) and Zeng et al.7 (page 394) reveal that intensification of agriculture has contributed substantially to this trend.
The atmospheric CO2 concentration has increased at an unprecedented rate during the past few decades. We know from a global network of atmospheric CO2 measurements that roughly only half of the emissions associated with fossil-fuel use and land-use change remain in the atmosphere8. The ocean and land surface must therefore act as a global carbon sink, although its magnitude and location — and the mechanisms driving it — remain uncertain because of the difficulty of measuring and modelling carbon stocks and fluxes at large scales. Improving our knowledge of the driving mechanisms is essential for accurate projections of the global carbon budget under future climate and land-use changes.
Atmospheric CO2 data can provide an integrated, albeit indirect, measure of the global carbon budget, and so it is crucial to understand the causes of spatiotemporal variability in these data. Much focus has been put on the growth rate of the annual mean CO2 concentration and its year-to-year variability. By contrast, less attention has been paid to the observed increase in the amplitude of the seasonal CO2 cycle in the extratropics of the Northern Hemisphere (regions at latitudes of 30° to 90° N), which results from higher carbon uptake in the summer and greater release in the winter.
Agricultural productivity has previously been proposed as a possible cause4. Crops can have a stronger impact on carbon uptake than can natural vegetation, because of their high productivity. The widespread use of fertilizers, irrigation and high-yield crop cultivars has led to a threefold growth in global agricultural production in the past 50 years, with only a small expansion of cropland area9 (Fig. 1). Gray et al. and Zeng et al. are the first to demonstrate that agricultural productivity really has affected the amplitude of the annual CO2 cycle.
Gray and colleagues used a carbon-accounting method and crop-production statistics published by the Food and Agriculture Organization of the United Nations to calculate how much carbon was taken up by four major crop types — maize (corn), wheat, rice and soya beans (collectively called MWRS) — in the northern extratropics each year from 1961 to 2008. They found that the annual exchange of carbon between crops and the atmosphere increased by 0.33 petagrams (1 petagram is 1015 grams) during this period, mainly because of farming in northern China and the midwestern United States. The authors conclude that the rise in MWRS production is responsible for 17–25% of the increase in the seasonal carbon flux required to explain observed changes in atmospheric CO2 seasonality2, with maize alone accounting for 66% of this increase.
Zeng and co-workers followed a more 'bottom-up' approach, adapting a terrestrial biosphere model known as VEGAS to include a simple representation of changing agricultural management practices for a generic crop functional type (a single description that represents an average of the growth characteristics of all crops). According to their study, enhanced agricultural productivity in the mid-latitudes contributes about 45% of the increasing amplitude of global net surface carbon fluxes between 1961 and 2010, compared with 29% from climate change and 26% from CO2 fertilization (increased photosynthesis caused by rising atmospheric CO2 levels).
Although both studies highlight the influence of agricultural intensification, they calculate considerably different values for its contribution to the increasing amplitude. Why is this? Gray et al. focused on the change in productivity in the extratropics, where MWRS accounts for only 68% of dry biomass production from crops — which, as they point out, may lead to a substantial underestimate in their proposed contribution. Zeng and colleagues, however, performed a global simulation with a generic crop model and assumed that crop growth is driven solely by favourable climate conditions. This may bias their results towards higher carbon uptake, because they do not account for winter wheat varieties that are commonly grown during the period of net carbon release.
So is the contribution of agriculture to the increasing seasonal amplitude of atmospheric CO2 closer to 20%, as Gray and co-workers estimate, or around 50%, in line with Zeng and colleagues' result? The jury is still out. 'Top-down' data-driven approaches, such as those used by Gray et al., conceivably provide the best available crop-specific estimates. Process-based modelling frameworks are complementary; their strength lies in their potential to examine the relative influence of all possible causal mechanisms, as undertaken by Zeng and co-workers. This requires the processes to be accurately represented, but current-generation terrestrial biosphere models vary in their sensitivity to temperature, precipitation and CO2 fertilization8. Moreover, the effects of nutrient limitation, and of changes in the age distribution and management of forests, are often missing or inadequately represented in models8. All of these issues may affect simulations of the temporal dynamics of carbon fluxes.
The terrestrial biosphere is thought to be the main driver of changes in atmospheric CO2 seasonality in the Northern Hemisphere2,5. However, we have not yet clearly differentiated between the many contributory effects, such as increased growing-season length1,10 and changing rates of respiration11 due to warmer temperatures; enhanced plant growth caused by climate change, CO2 fertilization and/or the deposition of nitrogen compounds from the atmosphere4,5; and human-induced disturbance of the natural ecosystem, for example from fire or grazing12. The intensification of agricultural productivity must now join the list.
Finally, an atmospheric-transport model that accounts for complex mixing processes is necessary to properly assess the different contributions to increased seasonality of atmospheric CO2 concentrations and their spatial distribution. Shifts in the seasonal variations of fossil-fuel emissions and ocean CO2 fluxes may have been overlooked, and the influence of tropical regions, although less seasonal, should be considered in future studies.