Attempts are now underway to estimate primary productivity on global scales from a combination of satellite color imagery and manned or unmanned platforms. Regardless of the observing network, productivity over basin scales will be estimated from plankton biomass, estimated in turn from composite (or averaged) satellite images. We show mathematically that the interaction of two-dimensional horizontal turbulent diffusion with an annual cycle in plankton growth induces an intrinsic large-scale patchiness in the resulting plankton biomass distributions. We assume, as typical scales, a root-mean-square horizontal turbulent velocity of 0.05 m s-1 and a turbulent mixing length of 50 km. The corresponding length scale for the resulting patchiness in the annual cycle is of the order of 700 km, or at least 10 times the mixing length. Such patchiness can occur even when the growth rate is spatially uncorrelated; in fact, we show that small-scale grazing by zooplankton could actually enhance the patchiness at a length scale of 700 km. The existence of the patchiness is not contingent upon assuming a given, finite breadth for the region of fertile water, unlike the earlier ''KISS'' models, nor upon spatial forcing of any type, but depends only on the existence of an annual cycle in primary productivity. The implication for numerical modeling is obvious. The most plausible physical requirement might be a turbulent quasi-geostrophic motion field at or approaching statistical stationarity, but the first priority for the biological component embedded in the modeled motion field ought to be a model of the onset of the spring bloom and the subsequent development of a seasonal cycle. ¿ American Geophysical Union 1988