The abilities of the Gent and McWilliams (1990) (GM) and Horizontal Diffusion (HD) eddyparameterizations to represent the mesoscale effects relevant for primary production are compared and analyzed. Following Lévy et al. (1999a), this is done in the case study of the spring bloom that follows the formation of a dense water patch in the northwestern Mediterranean Sea. It is shown that, unlike HD, the use of the GM parameterization can capture many aspects of the primary production enhancement associated with the restratifying action of mesoscale eddies. However, predicted primary production, when using the GM parameterization, is sensitive to the GM's parameter set, and particularly to the maximum value of the lateral mixing coefficient, kmax .