In West Africa, little is known about the restoration trajectories of secondary forests, and even less about the origin of the variability observed in recovery rates. We installed 8 chronosequences on a typical North-South climatic gradient in West Africa and inventoried, for each chronosequence, all trees over 2.5cm DBH on 20 plots aged 0-50 years, including controls. These data allowed us to quantify, and model in a hierarchical Bayesian framework, the recovery trajectories of biodiversity, aboveground biomass and floristic composition. The relative effects of local (plot history and landscape context) and regional (climate and soil) variability on these trajectories were then estimated. The results show that diversity recovers faster than composition, while above-ground biomass is the slowest. Seasonality of rainfall, soil hydromorphism and the number of remnant trees had a major impact on all restoration trajectories, while the duration of the previous agricultural cultivation negatively influenced the speed of recovery of aboveground biomass and diversity. In conclusion, the variability of regional pedoclimatic conditions certainly plays a role in the reconstitution rates of secondary forests, but local conditions linked to the history of the plot and to the landscape context are very important to consider in a decision-making framework for the choice of more or less active restoration techniques.