In situ bio-optical measurements from several oceanographic campaigns have been analyzed to derive a direct relationship between water column primary production Pt ocean color as expressed by the ratio of reflectances R1 at 440 nm and R3 at 550 nm and photosynthetically available radiation (PAR). The study is restricted to the Morel case I waters for which the following algorithm is proposed: log(Pt)=-4.286-1.390 log (R1/R3)+0.621 log (PAR), with Pt in g C m-2 d-1 and PAR in J m-2 d-1. Using this algorithm the rms accuracy of primary production estimate is 0.17 on a logarithmic scale, i.e., a factor of 1.5. Using spectral reflectance measurements in the entire visible spectral range, the central wavelength, spectral bandwidth, and radiometric noise level requirements are investigated for the channels to be used by an ocean color space mission dedicated to estimating global marine primary production and the associated carbon fluxes. Nearly all the useful information is provided by two channels centered at 440 nm and 550 nm, but the accuracy of primary production estimate appears weakly sensitive to spectral bandwidth, which, consequently, may be enlarged by several tens of nanometers. The sensitivity to radiometric noise, on the contrary, is strong, and a noise equivalent reflectance of 0.005 degraded the accuracy on the primary production estimate by a factor 2(0.14--0.25 on a logarithmic scale). The results should be applicable to evaluating the primary production of oligotrophic and mesotrophic waters, which constitute most of the open ocean. ¿ American Geophysical Union 1994