An adjoint model was developed for the conservation of a tracer and used to assimilate phosphate observations from the North Pacific. The adjoint model estimated optimal values for new production, remineralization length scale of particulate organic matter (POM), phosphate field, and circulation field. Allowing no modifications to the circulation field computed by the Hamburg large-scale geostrophic (LSG) model, we could not produce optimal estimates of new production and phosphate concentrations that were consistent with the observations. However, by allowing modifications to the LSG model's circulation field the adjoint model demonstrated that a model with only POM transport of organic matter produced results which were consistent with observations of new production and phosphate concentrations. In this model only small modifications to the circulation field were required to produce consistent estimates of new production and phosphate concentrations. This showed that the modeled phosphate field and new production were sensitive to small changes in the circulation. Furthermore, the data assimilation model implied that dissolved organic phosphorus did not necessarily play an important role in the cycling of phosphate in the North Pacific. ¿ American Geophysical Union 1995