Sources and sinks of atmospheric carbon dioxide estimated from batch least-squares inversions of carbon dioxide concentration measurements

Roughly 55% of the CO₂ produced by fossil fuel burning and deforestation does not stay in the atmosphere, but is taken up by the oceans and land biosphere. The nature of this sink—where and when it occurs—is poorly known, however. Atmospheric transport models have been used before to estimate the regional distribution of the sink from atmospheric CO₂ concentration measurements. These inverse methods generally agree that there should be more uptake in the north than the south, but disagree on the land/ocean partitioning and on which continents take up the most.; The regional CO₂ flux estimates given by time-mean CO ₂ flux inversions of atmospheric CO₂ measurements are found here to vary greatly depending on the time discretization used, the number of regions solved for, and the transport model used. On average, though, these inversions place over 2 GtC/yr of uptake in Eurasia and North America together, with more of it in Eurasia. The uptake by the oceans south of 15°S is under 1 GtC/yr, less than half that given by the Takahashi, et al. (1999) ‘1995’ model; the estimates for the tropical and northern oceans generally agree with that model. The tropical and southern land regions take up CO₂ overall, meaning that extra growth must be compensating for deforestation there.; The interannual variability of the regional CO₂ fluxes for 1980–1996 is also estimated here in a fully time-dependent batch inversion. The flux variability of the land biosphere is found to be twice that of the ocean. The tropical land regions are found to explain most of the global interannual variability. The amplitude of the tropical Pacific variability is small in comparison. Due to the sparse measurement network, however, the tropical ocean fluxes outside the Pacific cannot be distinguished well from the tropical land fluxes. Regularization methods (frequency truncation and SVD truncation) have been used to obtain robust results here, but O₂/N₂ and 13C measurements should be used in future inversions to help partition the tropical fluxes. Both the tropical land regions and the global oceans are found to release CO₂ during El Niño warm events of the 1980s; the decreased outgassing in the eastern tropical Pacific then is more than compensated for by solubility-related outgassing in the other ocean regions. The strong global uptake of 1992–1993 is attributed to the land, especially in the north where there were cooler temperatures at that time.; This study highlights the urgent need to reduce mixing errors in our transport models, and points to the importance of the pre-industrial carbon cycle when comparing the inversion results to anthropogenic CO₂ flux estimates.