Secondary forests recovering after disturbances currently comprise about half of the world′s tropical forests. A better understanding of the recovery of species composition, stand structure, and biomass stocks of these secondary forests is critical for the sustainable management of these ecosystems. The main objective of this research was to develop the successional trends in forest biomass, tree diversity, and species composition recovery. We monitored the changes in tree species composition, stand structural characteristics, and biomass growth in a chronosequence of tropical secondary forests in the Yucatan Peninsula of Southeastern Mexico. We used both linear and non-linear regressions to develop biomass growth models during forest succession. We calculated tree species diversity and similitude indices between forest successional groups (young, medium-aged, and old-growth forests). The aboveground biomass stocks ranged from 5.2 to 121 Mg C ha-1, which increased gradually with forest age. We observed a divergent trend of biomass accumulation between soft- and densewood species during succession. The contribution of densewood species on tree biomass stock increased (from 32 % to 81 %) with the increasing forest ages. However, the tree diversity indices did not change significantly with forest age and were a poor predictor of aboveground biomass growth during succession. Secondary forests (SFs) of advanced age were more similar to old-growth forests (OF) in their species composition compared to young ones, as evidenced by the Chao-Jaccard index. However, it may take more than a century for these secondary forests to recover to OF-level species composition. Furthermore, the structural equation model allowed us to infer the relative importance of forest age, tree diversity, soil properties, and stand structural attributes, including wood density on biomass growth along the forest chronosequence. Forest age-mediated effect of species composition shift and stand structural changes was a better predictor of biomass accumulation than the changes in tree diversity or soil properties during succession. However, the soil properties as a construct (latent variable), had a significant positive covariance with stand structure, indirectly influencing biomass growth. These biomass and species composition recovery models could be useful in forest management for biodiversity conservation, landscape restoration, and ecosystem services, including carbon sequestration.