Most urban lakes are shallow, e.g., former sand pits or retention basins. They provide varied amenities, including the increasingly demanded outdoor activities and bathing. But many of them are eutrophic and threatened by proliferations of phytoplankton, especially of toxic species. Law often encourages the reduction of the external nutrient load to these lakes. But unlike in deep lakes, phytoplankton proliferations can continue for many years after the reduction, since wind-induced mixing can release an internal load from the sediment many times in the year (polymicticity). To warn for harmful algal blooms or to assess how climate change may impact their frequency, a correct simulation of the alternation between stratification and mixing periods is required. We therefore investigated the mixing regime of a shallow urban lake, Lake Créteil, a former sand pit and now eutrophic lake (surface 0.40 km², mean depth 4.5 m, catchment area 1 km²). Meteorological variables and water temperature at 5 depths were measured in the deepest part of the lake (5.5 m) every 30 s from May to August in 2012, 2013 and 2014. Mixing and stratification periods were sorted on daily maximum and minimum differences between subsurface and bottom temperatures. The number of days of stratification (78, 81 and 60 days) and mixing (26, 34 and 47 days) and the number of stratification periods (6, 3 and 4) showed a strong inter-annual variability. We then tested whether a hydrodynamic model could reproduce these features. We already presented last year the calibration of Delft3D-FLOW on a one-month period and its performance on vertical and horizontal temperature heterogeneities, current speeds and internal waves. Here we only modified the wind drag coefficient to calibrate it on 2012 temperatures, and verified it on 2013 and 2014 temperatures. Simulation results reproduced well the observed inter-annual polymictic patterns for all three years, with less than 10 % error in the number of mixing or stratification days. Such a validated model, forced with outputs of a regional climate model, could be very useful to predict changes in mixing regime and phytoplankton dynamics in temperate shallow urban lakes.