Rising atmospheric CO2 concentrations are the result of a long-term anthropogenic emissions against the background of a much larger natural carbon cycle. Exchanges among with huge carbon reservoirs in the ocean and land surface involve about 20% of all atmospheric CO2 molecules each year, and have the net effect of removing about half of the CO2 emitted by combustion of fossil fuel. This effective 50% emissions reduction is essentially a free service of the Earth system, yet a quantitative accounting of the locations and mechanisms responsible for these “sink” processes remains elusive after decades of research. Marine uptake is accomplished by both physico-chemical mechanisms (the “solubility pump”) and by biological fixation and sinking (the “biological pump”), and is limited primarily by rates of physical mixing into the deep ocean. Terrestrial uptake is the net result of differences in rates of globally integrated photosynthesis and decomposition, and is believed to be driven by CO2 fertilization, nutrient deposition, forest regrowth, invasion of woody shrubs, and the extension of boreal growing seasons due to climate change. Most of these mechanisms are expected to “saturate” in coming decades, leading most models to predict a reduced sink strength and enhanced growth rates of atmospheric CO2 in the future for given levels of anthropogenic emissions. Prediction and risk assessment with respect to global change requires that changes in carbon stocks and fluxes among natural reservoirs in the Earth system be quantified and that predictive, falsifiable models of source and sink processes be developed and tested.

Interactions between climate and the carbon cycle may produce substantial climate “surprises” in coming decades. A recent pair of global modeling studies involved 250-year simulations with fully coupled atmosphere-ocean-land surface GCMs. Fully prognostic carbon cycles were included in each model, and rather than specify CO2 concentrations, fossil fuel emissions were prescribed as boundary conditions. Changes in atmospheric CO2 were predicted as a residual after air-sea and air-land exchanges were accounted for, and the models were integrated from 1850-2100. Both models simulated strong sinks in both land and oceans, but one of the models simulates a dramatic reversal in the mide-21st century, with massive releases of CO2 from the land surface while the other does not. The difference results in almost 1000 ppm of CO2 in the global atmosphere by 2100 in one model, vs. only 750 ppm in the other, and a factor of two difference in resulting global warming. Carbon-climate feedbacks are therefore a first-order source of uncertainty in the simulation of future climate change, on a par with projected fossil fuel emissions or cloud-climate feedbacks.

A systematic program of in-situ measurements, remote sensing, and modeling, known as the North American Carbon Program, is being planned to address these questions. Basic processes contributing to carbon sinks will be evaluated, and regional extrapolation will be accomplished using multiple data sets. Mass-balance atmospheric constraints will be used to test model predictions at larger scales.