Large areas (2 million hectares per year) of Amazonian forests are selectively logged in a polycyclic harvest system. Modeling the post-logging dynamics of these production forests is thus of primary importance to assess their future carbon storage capacity as well as the structural and dynamic features of the forest that will be found in the next logging cycle. In this study, we used a network of 100 permanent sample plots in 10 sites spread across the Amazon basin to model three post-logging biomass fluxes (recruitment, growth and mortality). The temporal evolution of these biomass fluxes (recruitment, growth and mortality) for surviving trees and recruits and their relative importance in explaining biomass recovery through the Amazonian basin were modeled taking into account spatial as well as temporal autocorrelation in a mixed model framework. Incorporating both the environmental variability and the logging characteristics in the developed model indicate that the two key drivers of post-logging biomass fluxes are the relative biomass loss due to logging and the initial aboveground biomass. Overall, environmental factors had little additional weight in explaining the Amazon-wide variations of post-logging biomass fluxes. Our results stress the importance of developing specific modelling frameworks to account for the peculiar carbon cycle in managed tropical forests in order to better recognize their key role for climate regulation at the global scale. (Texte intégral)