A coupled physical biological optical model was developed to investigate the causes of phytoplankton variability in the spring 1979 Mid-Atlantic Bight (MAB) as observed in Coastal Zone Color Scanner (CZCS) imagery, and to estimate the magnitude and variability of primary production. The model incorporated advection, mixing, sinking, growth as a function of light, temperature, and nutrient availability, and death as a function of ingestion. These variables were assumed to determine the large-scale, low-frequency variability of phytoplankton distribution. The model also contained two phytoplankton groups, netplankton and nanoplankton, which differed in maximum growth rate, sinking rate, and specific light absorption. The model produced chlorophyll concentrations within the first attenuation depth within 1 standard deviation of CZCS imagery on large scale (i.e., over regions about 50 km in width). This suggested the model represented the physical/biological coupling often observed in imagery. The two phytoplankton groups used in the model were initialized to equal distributions throughout the model domain and organized in the course of the model run into distributions corresponding to long-term observations in the MAB: netplankton dominated the coast and nanoplankton dominated the slope. This suggested that the model incorporated the mechanisms causing these distributions in the real ocean, namely, the interaction between mixing, differential growth rates, and differential sinking rates. Finally, primary production estimates were within reasonable agreement with those measured in situ, suggesting the applicability of the model in estimating regional-scale primary production. ¿ American Geophysical Union 1992