Current aquatic ecosystem models accommodate increasing amounts of physiological detail, but marginalize the role of biodiversity by aggregating multitudes of different species. We propose that at present, understanding of aquatic ecosystems is likely to benefit more from improved descriptions of biodiversity and succession than from incorporation of more realistic physiology. To illustrate how biodiversity can be accounted for, we define the system of infinite diversity (SID), which characterizes ecosystems in the spirit of complex adaptive systems theory as single units adapting to environmental pressure. The SID describes an ecosystem with one generic population model and continuity in species-characterizing parameters, and acquires rich dynamics by modeling succession as evolution of the parameter value distribution. This is illustrated by a four-parameter phytoplankton model that minimizes physiological detail, but includes a sophisticated representation of community diversity and interspecific differences. This model captures several well-known aquatic ecosystem features, including formation of a deep chlorophyll maximum and nutrient-driven seasonal succession. As such, it integrates theories on changes in species composition in both time and space. We argue that despite a lack of physiological detail, SIDs may ultimately prove a valuable tool for further qualitative and quantitative understanding of ecosystems.Biodiversity poses a perennial problem for ecosystem modelers. Confronted with a reality fraught with species, dependencies, and physiological detail, one cannot help but think that simple models cannot do it justice (Anderson 2005). Simple models aggregate large numbers of species into single state variables, and by doing so they lose the ability to reproduce ranges of behavior shown by detailed species-explicit models (Raick et al. 2006). Also, the use of aggregation puts models at a greater distance from empirical results, first because assimilation of empirically determined, species-specific parameter values to parameters of virtual aggregates of species is a difficult and largely subjective process; second because aggregate models provide only indirect information about individual species observed in the field. Not surprisingly, large ecosystem models that describe many classes of species explicitly have recently gained in popularity (Baretta et al. 1995; Quéré et al. 2005). However, continued diversification of functional groups may create more problems than it solves. Increasing the number of groups within ecosystem models dilutes the available empirical information per model unit, and therefore increases the uncertainty per parameter. Considering the substantial uncertainty already associated with parameters of moderate-size ecosystem models, this route seems unappealing. Also, it is easy to overlook that as the number of variables within an ecosystem model increases, so does the amount of information needed to initialize the model. A utopian species-complete model would require ini...