Amazonian forests are exposed to a range of threats, and forest dynamics models are needed to appraise their response to environmental changes, and to assess the likelihood of possible tipping points. However, these forest dynamics models need to be parameterized and validated based on the data produced by long-term ecological observatories. Although a good network of long-term forest inventory plots is available in French Guiana, understory trees (less than 10 cm dbh) have been overlooked, with serious implications for the modelling of tree regeneration, and also for our understanding of tree biodiversity and ecosystem functioning. The Understory project will address this knowledge gap. We will establish two 4-ha plots where all trees will be inventoried above 1cm dbh, one in each of the two main research stations of French Guiana (Nouragues and Paracou). Tree mapping will be facilitated by the novel use of digital terrain model obtained by airborne lidar and campaign of terrestrial lidar acquisition. All stems will be botanically identified using deep- learning plant species identification building on the previously established Pl@ntNet infrastructure. Combining our data with that of the Smithsonian Institution’s ForestGEO program, in central and western Amazonia, we will parameterize the individual-based forest simulator TROLL coupled with bioclimatic conditions. Finally, we will use the newly acquired data to advance understanding of forest structure, diversity and functioning by exploring spatial patterns of forest organization, through a comparative study of forest structure and diversity across the Amazon. This project has the potential to benefit the scientific community of Labex CEBA and will have long-lasting implications on forest science in French Guiana. By improving understanding of tree regeneration, it can also inform sylvicultural practices and benefit the French Guianese forest sector.