In this paper we use numerical models of coupled biological-hydrodynamic processes to search for general principles of bloom regulation in estuarine waters. We address three questions: What are the dynamics of stratification in coastal systems as influenced by variable freshwater input and tidal stirring? How does phytoplankton growth respond to these dynamics? Can the classical Sverdrup Critical Depth Model (SCDM) be used to predict the timing of bloom events in shallow coastal domains such as estuaries? We present results of simulation experiments which assume that vertical transport and net phytoplankton growth rates are horizontally homogeneous. In the present approach the temporally and spatially varying turbulent diffusivities for various stratification scenarios are calculated using a hydrodynamic code that includes the Mellor-Yamada 2.5 turbulence closure model. These diffusivities are then used in a time- and depth-dependent advection-diffusion equation, incorporating sources and sinks, for the phytoplankton biomass. Our modeling results show that, whereas persistent stratification greatly increases the probability of a bloom, semidiurnal periodic stratification does not increase the likelihood of a phytoplankton bloom over that of a constantly unstratified water column. Thus, for phytoplankton blooms, the physical regime of periodic stratification is closer to complete mixing than to persistent stratification. Furthermore, the details of persistent stratification are important: surface layer depth, thickness of the pycnocline, vertical density difference, and tidal current speed all weigh heavily in producing conditions which promote the onset of phytoplankton blooms. Our model results for shallow tidal systems do not conform to the classical concepts of stratification and blooms in deep pelagic systems. First, earlier studies (Riley, 1942, for example) suggest a monotonic increase in surface layer production as the surface layer shallows. Our model results suggest, however, a nonmonotonic relationship between phytoplankton population growth and surface layer depth, which results from a balance between several ''competing'' processes, including the interaction of sinking with turbulent mixing and average net growth occurring within the surface layer. Second, we show that the traditional SCDM must be refined for application to energetic shallow systems or for systems in which surface layer mixing is not strong enough to counteract the sinking loss of phytoplankton. This need for refinement arises because of the leakage of phytoplankton from the surface layer by turbulent diffusion and sinking, processes not considered in the classical SCDM. Our model shows that, even for low sinking rates and small turbulent diffusivities, a significant percentage of the phytoplankton biomass produced in the surface layer can be lost by these processes.