A Bayesian Synthesis Inversion of Carbon Cycle Observations: How Can Observations Reduce Uncertainties about Future Sinks?

The strengths of future carbon dioxide (CO 2) sinks are highly uncertain. A sound characterization of this uncertainty and how they might be reduced is crucial for the design of efficient carbon management strategies. Here a Bayesian synthesis inversion of historical carbon cycle observations is used to (i) estimate probability density functions (PDFs) of key carbon cycle parameters, (ii) derive statistically sound probabilistic predictions of future CO 2 sinks and (iii) assess the utility of hypothetical observation systems to reduce the prediction uncertainties. We find that the PDFs of model parameter estimates are not normally distributed, and that the residuals show a statistically significant correlation. Previous studies that assume normally distributed PDFs are likely biased, and those that neglect autocorrelation are overconfident in parameter estimates and predictions. The interannual variability of global temperature and the resulting CO 2 variations provide important information: Terrestrial parameter estimates are more sharply constrained if historical temperature data are assimilated rather than a smooth temperature timeseries generated with a simple energy-balance model. The expected values of terrestrial parameters also change significantly if observed temperatures are used. Although CO 2 observations provide a strong constraint on the total carbon sink, adding independent observations of terrestrial and oceanic fluxes has the potential to reduce uncertainty in predictions of this total sink more rapidly. Hypothetical annual observations of terrestrial and oceanic CO 2 fluxes with realistic uncertainties reduce predictive uncertainties about CO 2 sinks in the year 2050 by as much as a factor of two compared to observing CO 2 alone.