1. Models aim to predict phytoplankton dynamics based on observed initial conditions and a set of equations and parameters. However, our knowledge about initial conditions in nature is never perfect. Thus, if phytoplankton dynamics are sensitive to small variations in initial conditions, they are difficult to predict.
2. We used time-series data from indoor mesocosm experiments with natural phyto- and zooplankton communities to quantify the extent to which small initial differences in the species, functional group and community biomass in parallel treatments were amplified or buffered over time. We compared the differences in dynamics between replicates and among all mesocosms of 1 year.
3. Temperature-sensitive grazing during the exponential growth phase of phytoplankton caused divergence. In contrast, negative density dependence caused convergence.
4. Mean differences in biomass between replicates were similar for all hierarchical levels. This indicates that differences in their initial conditions were amplified to the same extent. Even though large differences in biomass occasionally occurred between replicates for a short time, dynamics returned to the same path at all hierarchical levels. This suggests that internal feedback mechanisms make the spring development of phytoplankton highly predictable.