Temporal changes of assemblages may result from environmental variability and reflect seasonal dynamics of their ecosystem. In the subtropics, the hydrological regime is usually characterized by well-defined wet and dry seasons, regulating discharge and influencing a series of environmental variables that affect phytoplankton persistence. Therefore, we may expect that dry seasons are environmentally more stable than wet seasons. We analysed interannual phytoplankton assemblage variability (or, inversely, persistence) in a subtropical reservoir sampled every austral summer and winter during 5 years. We tested (i) if phytoplankton assemblage structure differed between the dry (summer) and wet (winter) seasons; (ii) if assemblage persistence differed between the seasons; (iii) if assemblage persistence was related to environmental stability; and (iv) if assemblage dissimilarity increased over time. Phytoplankton assemblages differed between the summer and winter seasons.Winter indicator species were mostly Bacillariophyceae or Cryptophyceae, whereas Cyanophyceae and Chlorophyceae taxa were more frequent and abundant in summer. Assemblages in the dry season were more persistent among years than those occurring during rainy periods. Similarly, environmental variability tended to be lower among dry than among rainy seasons. The relation between the phytoplankton temporal cycle and the temporal patterns of environmental variability supports our prediction that high environmental stability results in more persistent assemblages. Assemblage dissimilarity increased as sampling years were farther apart, for both seasons. Additionally, assemblages in the rainy periods showed a more pronounced increase in dissimilarity, as their changes among years were less predictable. We found a clear temporal pattern and an increased dissimilarity over time in the phytoplankton assemblage structure. Unravelling these temporal patterns may improve our understanding of phytoplankton temporal dynamics, and may have implications for management and monitoring programs. High dissimilarity of assemblages among years, particularly among rainy periods, can obscure human impacts, and monitoring programs should take this into account.