Express the quantity of biomass recovered over time in primary exploited and secondary forest plots, by including co-variables (landscape, climate, topography, soil) and developing our model in a Bayesian framework, which is particularly adapted when there is little data, thanks to the addition of information that could be described as "non-pure data" or "priors". These priors can be based on previous studies or expert knowledge: they are a way to make sure that our predictions are within the range of acceptable values given our prior knowledge on similar processes, and are thus especially important when data is scarce. Moreover, the Bayesian approach allows a rigorous estimation of parameters correlation and uncertainty.