A nitrogen balancing scheme for ocean color data assimilation in general circulation models is described and its potential for improving air-sea CO2 flux estimates is demonstrated. Given increments for phytoplankton, obtainable from a univariate surface chlorophyll analysis, the scheme determines mixed layer concentration increments for the other nitrogen pools: zooplankton, detritus and dissolved inorganic nitrogen (DIN). The fraction of the phytoplankton increment to be balanced by changing DIN varies dynamically with the likely contributions of phytoplankton growth and loss errors to the error in the background state. Further increments are applied below the mixed layer wherever positive DIN increments in shallower layers would otherwise cause the creation of unrealistic sub-surface minima. Total nitrogen at each grid point is conserved where possible. The scheme is evaluated by 1-D twin experiments for two contrasting locations in the North Atlantic, in which synthetic chlorophyll observations are assimilated in an attempt to recover known system trajectories generated by perturbing model parameters. Dissolved inorganic carbon (DIC) and alkalinity tracers, controlled by the nitrogen dynamics, determine the biological modification of sea-water pCO2 at the ocean surface. Assimilation affects DIC and alkalinity directly, the increments being inferred from the nitrogen increments, as well as having a post-analysis effect via the dynamics. It gives major improvements in surface pCO2 at 50N but less improvement at 30N where errors in the phytoplankton nitrogen:chlorophyll ratio cause it to have a detrimental effect in summer. Beneficial effects of nitrogen balancing are demonstrated by comparison with experiments in which only phytoplankton and DIC are updated in the analysis.