Water temperatures are one of the most important factors of freshwater ecology and lake hydrodynamics. In order to evaluate the effect of water temperatures on fish in inland water bodies at the scale of France we undertook a study linking the advantages of models and satellite data. We simulated epilimnion and hypolimnion temperature for the period 1999-2013 at 480 French inland waterbodies, both natural and artificial, with a surface of >0.5 km2. We used three models: (i) the hydrodynamic model FLake, (ii) a modification of the model of Ottosson and Abrahamsson (1998) (OAmod), and (iii) the four-parameter model of Toffolon et al. (2014) (Tea2014, epilimnion temperature only). The Flake model requires no calibration. For the Tea2014 model, the parameter values were estimated using the regressions as a function of depth proposed in the original publication. The OAmod model was derived by the authors and had several calibration parameters. Water temperature data for calibration and validation came from three sources: 1) temperature profiles measured as part of the European Water Framework Directive monitoring of water quality in lakes; 2) continuous measurements at five water bodies of different characteristics using thermistor chains; 3) surface temperature measurements derived from Landsat 5 and Landsat 7 satellite thermal images processing for all studied lakes in the study period. These data was useful to analyse spatial surface water temperature variability and the characteristics of the annual thermal cycle of French water bodies. Meteorological data were obtained from Météo-France’s 8-km-resolution SAFRAN (Système d’Analyse Fournissant des Renseignements À la Neige) reanalysis at the nearest grid point. Satellite-image processing, which essentially consisted in the correction of atmospheric effects, provided surface temperatures with reasonable accuracy (root mean square error of 1-2 ºC), though they tended to underestimate surface temperatures. Variations in latitude, altitude and continentality explained well the spatial variability in epilimnion temperatures (mean annual temperature, average temperature in Januray-February, average temperature in July-August). Epilimnion temperature simulation was best for water bodies of between 1 m and 5 m depth. Simulation errors increased for shallower (< 1 m deep) and deeper (> 5 m deep) water bodies. For shallow water bodies (< 5 m), the three models simulated epilimnion temperature comparably well, but for deeper ones OAmod was best. Two models were used to simulate hypolimnion temperatures: OAmod and FLake. Error levels for the simulation of hypolimnion temperature were higher than for epilimnion temperature. For OAmod, simulation error decreased with water body depth, and it provided better simulations than FLake for water bodies deeper than 2-3 m. FLake gave netter predictions for shallow water bodies of less than 2-3 m, error increased greatly for deeper water bodies.