Eutrophication of lakes and harmful algae blooms are recently widespread water ecological environmental issues all over the world. Lake Taihu is a typical large shallow and eutrophic lake, and the cyanobacteria blooms induced by eutrophication have always been presenting high spatial-temporal variability and instability that made it very difficult to monitor and predict. Many previous publications have related to the eutrophication and cyanobacteria bloom with the most attention paid to phytoplankton growth, biomass increase and dominant species, while a few works addressed the appearance/disappearance of cyanobacteria bloom; however, none of them could explain the swift shift of cyanobacteria bloom in time and space in Lake Taihu. Based on the long-term field observation data and simulation experiments in Lake Taihu, here we presented a physical process controlled cyanobacteria bloom formation mechanism. In Lake Taihu, the visible cyanobacteria bloom occurrence was mainly controlled by the hydrodynamic intensity. When the cyanobacteria biomass accumulated in the water column, the large size colonies would increase during the process of cell division and proliferation, cells and colonies collisions, colonies aggregation (due to the stick extracellular polysaccharides) and disaggregation (due to the intensive turbulence). The bloom would occur if the wind influence decline and the colonies suspended at water column were able to float to the water surface to form bloom and scum. These colonies floating at the surface couldnot migrate downward because of the large size and great buoyancy; furthermore, they would drift to the downwind zone. But even the hydrodynamic process played the key role in the cyanobacteria bloom formation in this large shallow lake; it couldnot exclude the existence of other dynamics determining the cyanobacteria bloom formation. During the calm weather periods, the light controlled colonies diurnal migration, the zooplankton predation induced colony formation, and toxic materials induced colony aggregation, all would promote the cyanobacteria bloom formation. Thus these physiological induced cyanobacteria bloom occurrence modes and physical process controlled modes would alternatively take place in this large shallow lake. This revised cyanobacteria bloom occurrence dynamics in large eutrophic lake provided the possibility for forecasting and preventing the cyanobacteria bloom. Because Lake Taihu has been functioning as drinking water source for millions of around people, precise prediction and precaution of the bloom could increase the efficiency of cyanobacteria bloom collection, which had important practical significance for reducing the nutrient loading for reuse by cyanobacteria and decreasing the risk of cyanobacteria scum decaying induced drinking water pollution. © 2016, Science Press. All right reserved.