Using five Pacific open-sea data sets with five subsets, we linearly correlated observed column photosynthesis (Pt, mg C m-2 d-1, from 14C-uptake) with satellite pigment (Csat, mg m-3), simulated from in situ data at the same stations, and modeled Pt by a depth-integrated approach. The correlation (r2) of Pt with Csat and its accuracy improved after combining Csat with an in situ-based proxy for the light-saturated rate of photosynthesis, Poptb (mg C (mg chl d)-1), for the particular stations. Incorporating station Poptb always led to high r2 and accuracy of the regression of modeled on observed Pt. However, Poptb cannot yet be determined from space, so that in practice, geographic or seasonal means of Poptb have to be used. For our 10 sets, such means lead to more imprecise (lower r2 in 9 of 10 cases) and less accurate (for all) results, because Poptb and Pt are often correlated. We expect this error source to be widespread and to afflict all modeling of Pt using a term for light-saturated photosynthesis, but we cannot offer a solution to the quandary. On the basis of combining of data sets, we do not favor the use of a single algorithm employing light-saturated photosynthesis for calculating global Pt by a depth-integrated approach.