Adaptive physiological traits of cyanobacteria to support their competitiveness have been widely recognized. However, in conventional cyanobacteria models, state variables respond directly to exogenous factors over the current time step only, without incorporating the ‘memory’ of environmental conditions experienced by cells during preceding time steps. Here, we developed a mechanistic, individual-based model (IBM) that transfers information on light exposure history and cell physiology through successive time steps. This IBM also incorporates cyanobacteria growth, respiration, buoyancy, mixing and transport. The model captures the light-induced fluorescence suppression and nonphotochemical quenching relaxation kinetics to deconvolve the cyanobacteria biomass variability driven by physical processes from the variability driven by physiological processes. The IBM was coupled with a three-dimensional hydrodynamic model to simulate a cyanobacterial bloom event and its collapse in a temperate lake (Peter Lake, Michigan, USA) under experimental nutrient enrichment. Model simulations showed that the thermal stratification caused an increase in cyanobacterial biomass at the water surface because of filament buoyancy and high growth rates. In contrast, mixing caused by a change in weather led to a rapid reduction in biomass due to the entrainment of filaments through a deep mixed layer as well as low growth rates. High daytime fluorescence quenching was followed by long photo-physiological relaxation periods during stratification, while low daytime quenching and rapid relaxation occurred in response to low light exposure history of filaments as the mixed layer deepened. Our results are important for understanding light-induced quenching and interpreting output from fluorescence probes used to estimate cyanobacteria and phytoplankton biomass. Conventional models are poorly adapted to capturing physiological responses of cyanobacteria to light and nutrient history. Our model improves current representations of cyanobacterial bloom dynamics at a whole lake scale by incorporating adaptive physiological traits, enabling much greater predictive accuracy than previously possible.