The design (location and size) of sustaining, no-take reserves was investigated by combining realistic numerical simulations of larval dispersal from a sedentary marine species with a population dynamics model. The method explored, a priori: (1) the planktonic larval duration (PLD) of self-persistent populations within no-take reserves with radii from 1 to 20 km, (2) the size of a no-take reserve reaching self-persistent recruitment of the reserve population, and (3) offspring spillover to adjacent fisheries for PLDs from 1 to 6 weeks. In the Gulf of Lions (northwest Mediterranean), as the radius of a no-take reserve increased to 20 km, the median PLD of a self-persistent species within the reserve increased from 2 to 6.5 d. No unique relation between PLD and sustaining no-take reserve size could be established because of large spatial and temporal variabilities, thus precluding any general guidelines for marine protected area sizes. For species with mass spawning lasting < 3 d, variability due to spawning timing yielded twice the spatial variability, reflecting strong wind variability. In contrast, when spawning lasted more than 10 d, the spawning location became more important. Thus, a biological process (spawning duration) can trigger deterministic and stochastic effects of environmental variability. Finally, some unprotected areas (Narbonne to Agde and the Camargue) clearly appeared to be better locations than the existing no-take reserves for maximizing biodiversity persistence within a reasonable no-take reserve size (10 to 20 km) and for producing offspring spillover important for regional fisheries (80%).