Small and shallow urban lakes provide many ecosystem services. In the Île-de-France region around Paris, 99% of the lakes have an area lower than 0.5 km2. In this region, the water quality degradation often leads to potentially toxic cyanobacteria blooms. Health risks generated by these blooms require a regular monitoring and warning systems when the human population is exposed. Besides the time variability of the biomass dynamics, a major difficulty in monitoring cyanobacteria comes from the heterogeneity of their vertical and horizontal distribution caused by their ecological strategies, by the alternation of thermal stratification and mixing and by surface and internal currents. Warning systems based on measuring buoys have been developed over the past decade. They do not provide information at the scale of the whole lake. In large water bodies, satellite remote sensing is sometimes used. In small water bodies, even if the image spatial resolution has greatly improved recently, the frequency of satellite pass limits an accurate detection of blooms. Furthermore, coupled hydrodynamic-ecological models can help in appraising cyanobacteria biomass at relevant space and time scales. At short time-scale, they can be embedded in early warning systems. In this paper, we present the implementation of a three-dimensional coupled hydrodynamic-ecological model, combined with continuous field measurements. The study site is a small and shallow lake (0.12 km2, average depth 2.3m, maximum depth 3m) located 20 km East of Paris, in a recreational park attended by the pupils of the neighbouring very densely urbanized county. Bathing in the lake has been repeatedly prohibited each summer because of cyanobacteria blooms. Two different datasets are available: (1) a summer survey, conducted since 2005 as required by bathing regulation, which includes the identification of phytoplankton genus and their relative abundance; (2) a scientific survey started in 2015 which includes high-frequency monitoring (5 min) of temperature and chlorophyll and bi-weekly vertical profiles of temperature and main phytoplankton groups. The main patterns of cyanobacteria dynamics and hydrodynamics are presented. The interannual variability of the seasonal cyanobacteria dynamics is outlined. The impact of the physical forcing is analyzed both at the seasonal scale over the 10-year period and at the daily scale in 2015. The modelling suite Delft3D (Deltares, 2014) was implemented for predicting the cyanobacteria biomass over a 5-day horizon and its transport towards regions of interest, for example a beach. At the short time-scale of the prediction, a main assumption is that cyanobacteria distribution is mainly driven by hydrodynamics. Therefore, in the ecological model, no growth rate but only a global decay rate of cyanobacteria is considered. The hydrodynamic model achieved very good results. The forecasts of cyanobacteria biomass for some typical episodes are presented and discussed.