Surface waterblooms of toxic cyanobacteria (scums) interfere with the use of lakes, for instance in the production of drinking water or for recreation. Routine monitoring data are not sufficient for early warning due to the large temporal and spatial variability in the occurrence of surface waterblooms, and the time lag between the formation of the scum and the availability of relevant information for risk management. We combined a "traditional" dynamic simulation model based upon differential equations with fuzzy logic to describe the three main conditions governing surface waterbloom formation: (1) a preexisting population of cyanobacteria, (2) buoyancy of the cells, and (3) stability of the water column. The attributes and membership functions of the fuzzy model were based on earlier field studies of diel changes in buoyancy and vertical distribution of cyanobacteria. The model was applied without further calibration to the large lake IJsselmeer (1200 km(2)) in the Netherlands, and we validated the model output using 12 years of NOAA-AVHRR (National Oceanic and Atmospheric Administration-Advanced Very High Resolution Radiometers) satellite images on which surface blooms are discernible as an enhanced vegetation index or increased surface water temperature. Existing surface blooms were predicted with high accuracy, but additional blooms were also predicted. A statistical test (Cohen's Kappa) showed that correct predictions of the absence or presence of surface blooms were highly unlikely to have occurred by chance only. The model can be used to predict the occurrence of surface waterblooms in advance on the basis of the long-term weather forecast, leaving time for appropriate management of the problem. The lake management has expressed interest in converting the present model into a fully operational-online-early warning system.