We use the terrestrial ecosystem model (TEM), a process-based model, to investigate how interactions between carbon (C) and nitrogen (N) dynamics affect predictions of net primary productivity (NPP) for potential vegetation in North America. Data on pool sizes and fluxes of C and N from intensively studied field sites are used to calibrate the model for each of 17 non-wetland vegetation types. We use information on climate, soils, and vegetation to make estimates for each of 11,299 non-wetland, 0.5° latitude × 0.5° longitude, grid cells in North America. The potential annual NPP and net N mineralization (NETNMIN) of North America are estimated to be 7.032 × 1015 g C yr−1 and 104.6 × 1012 g N yr−1, respectively. Both NPP and NETNMIN increase along gradients of increasing temperature and moisture in northern and temperate regions of the continent, respectively. Nitrogen limitation of productivity is weak in tropical forests, increasingly stronger in temperate and boreal forests, and very strong in tundra ecosystems. The degree to which productivity is limited by the availability of N also varies within ecosystems. Thus spatial resolution in estimating exchanges of C between the atmosphere and the terrestrial biosphere is improved by modeling the linkage between C and N dynamics. We also perform a factorial experiment with TEM on temperate mixed forest in North America to evaluate the importance of considering interactions between C and N dynamics in the response of NPP to an elevated temperature of 2°C. With the C cycle uncoupled from the N cycle, NPP decreases primarily because of higher plant respiration. However, with the C and N cycles coupled, NPP increases because productivity that is due to increased N availability more than offsets the higher costs of plant respiration. Thus, to investigate how global change will affect biosphere-atmosphere interactions, process-based models need to consider linkages between the C and N cycles.