We evaluated annual productivity and carbon fluxes over the Fontainebleau forest, a large heterogeneous forest region of 17,000 ha, in terms of species composition, canopy structure, stand age, soil type and water and mineral resources. The model is a physiological process-based forest ecosystem model coupled with an allocation model and a soil model. The simulations were done stand by stand, i.e., 2992 forest management units of simulation. Some input parameters that are spatially variable and to which the model is sensitive were calculated for each stand from forest inventory attributes, a network of 8800 soil pits, satellite data and field measurements. These parameters are: (1) vegetation attributes: species, age, height, maximal leaf area index of the year, aboveground biomass and foliar nitrogen content; and (2) soil attributes: available soilwater capacity, soil depth and soil carbon content. Main outputs of the simulations are wood production and carbon fluxes on a daily to yearly basis. Results showed that the forest is a carbon sink, with a net ecosystem exchange of 371 g Cm– 2 year – 1.Net primary productivity is estimated at 630 g C m–2 year –1 over the entire forest. Reasonably good agreement was found between simulated trunk relative growth rate (2.74%) and regional production estimated from the National Forest Inventory (IFN) (2.52%), as well as between simulated and measured annual wood production at the forest scale (about 71,000 and 68,000m3 year –1, respectively). Results are discussed species by species.