Understanding the carbon cycle requires accurate quantification of today's fluxes, explanation of those fluxes in terms of specific mechanisms and incorporation of those mechanisms into predictive modeling. The carbon cycle includes significant components associated with physical, chemical, biological and human processes and mandates an unprecedented degree of integration between disciplines. Longterm changes to the carbon budget are dominated by slow changes to compartments (the deep ocean, soils, vegetation) with long overturning timescales, against a background of large high frequency variability. Data assimilation techniques are under development that incorporate many different types of carbon cycle observations (atmospheric, oceanographic and ecological) into a consistent framework for both quantification of fluxes and attribution of fluxes to specific mechanisms. Preliminary results from an optimal observing system design excercise will be shown.