We couple a model of oxygen which includes surface gas exchange and net biological production with an extant model for upper-ocean physics that responds to surface inputs of seasonal buoyancy flux and wind stress. This model predicts the time-dependent depth of the mixed layer and vertical fluxes due to turbulent mixing below the mixed layer. Using this model, we simulate the seasonal thermal and oxygen cycles at a site in the subtropical North Atlantic and compare the simulation results with a composite year of temperature and oxygen created by averaging observations of temperature and oxygen from 1961 to 1970. The magnitude of diffusive vertical mixing is constrained to a narrow range by the thermal cycle. We then use this model diagnostically to estimate the biological new production by adjusting the levels of annual production of oxygen in the model until the simulation results match the composite year. The estimate of production is sensitive to the magnitude of turbulent vertical mixing (presumed known from the seasonal thermal cycle), the amount of near-surface production, and the surface gas exchange rate. In consideration of the role of air injection due to bubbles and the lack of nitrate in the surface waters, we conclude that production in the surface waters is negligible. The major uncertainty in the estimate of the production is due to the poorly constrained relationship between the gas exchange rate and the wind speed. Within the bounds provided by the laboratory and field experiments of gas exchange, we estimate that the vertically integrated, annual oxygen production is in the range of 3--4 mol m-2, which is slightly less than that estimated by Jenkins and Goldman (1985). ¿ American Geophysical Union 1988