The processes that underlie forest productivity and C allocation dynamics in trees are still poorly understood. Forest growth has for long been thought to be C limited, through a hypothesized causal link between C supply and growth (source control). This C-centric paradigm underlies most of the C allocation rules formalized in process-based models (PBMs). However, the source limitation of growth has been questioned by several authors, arguing that meristem activities are more sensitive than C assimilation to environmental stresses (e.g., water deficit and low temperatures). Moreover, the effect of management, which strongly affects forest functioning and wood growth, is not accounted for in most of the PBMs used to project the future terrestrial C sink. Our main objective in this thesis was to move forward into our understanding of the constraints that affect - or will affect - the wood productivity in European forests, from present to the end of the 21st century. We addressed this objective through the improvement of the representation of the forest productivity and C allocation in the CASTANEA PBM, building on a detailed analysis of the key drivers of annual wood productivity in French forests over the last 30 years (the species studied are Fagus sylvatica, Quercus ilex, Quercus petraea, Quercus robur and Picea abies). Our results supported the premise that the annual wood growth of the studied species is under a complex control including both source and sink limitations. The inter-site variability in the fraction of C allocatedto stand wood growth was predominantly driven by an age-related decline. At the tree level, we showed that annual wood growth was well predicted by the individual size. The size-asymmetry of growth, i.e., the advantage of big trees in the competition for resources, increased consistently with the whole stand productivity at both inter-site and inter-annual scales. On the basis of our findings, we developed a new C allocation scheme in the C