Logging and understory fires are major drivers of tropical forest degradation, reducing carbon stocks and changing forest structure, composition, and dynamics. In contrast to deforested areas, sites that are disturbed by logging and fires retain some, albeit severely altered, forest structure and function. In this study we simulated selective logging using the Ecosystem Demography Model (ED-2) to investigate the impact of a broad range of logging techniques, harvest intensities, and recurrence cycles on the long-term dynamics of Amazon forests, including the magnitude and duration of changes in forest flammability following timber extraction. Model results were evaluated using eddy covariance towers at logged sites at the Tapajos National Forest in Brazil and data on long-term dynamics reported in the literature. ED-2 is able to reproduce both the fast (< 5yr) recovery of water, energy fluxes compared to flux tower, and the typical, field-observed, decadal time scales for biomass recovery when no additional logging occurs. Preliminary results using the original ED-2 fire model show that canopy cover loss of forests under high-intensity, conventional logging cause sufficient drying to support more intense fires. These results indicate that under intense degradation, forests may shift to novel disturbance regimes, severely reducing carbon stocks, and inducing long-term changes in forest structure and composition from recurrent fires.