Background/Question/Methods: Assembly of hyperdiverse tropical tree communities remains a subject of rich debate. Here we propose original methods to evaluate the relative importance of habitat filtering, species interactions, and dispersal limitation, in structuring tree communities across spatial scales in French Guianan forests. We first control for environmental filtering using a new statistical test of correlations between tree species density and environmental maps, by generating Monte Carlo simulations of random raster images that preserve autocorrelation of the original maps. We then employ a novel test based on Rao’s quadratic entropy to examine neighborhood community structure in terms of taxonomic, phylogenetic and functional diversity, at different spatial scales within homogeneous forest habitats. Results/Conclusions: Our results underline the importance of habitat filtering in structuring local tropical tree communities. Nearly three fourths of tree species in our datasets exhibit distributions constrained by relative elevation, revealing contrasted habitat preferences with soil hydrology and fertility. Nevertheless, other assembly processes such as dispersal limitation, negative density dependence and competitive interactions may contribute to local community structure at different spatial scales and across different forest plots. Our study provides methodological advances both to examine habitat preferences and to examine neighborhood composition at different spatial scales. In this way, our study will permit comparative analyses with the French Guianan plots presented here, of the relative contributions of habitat filtering, dispersal limitation and species interaction to local assembly and how these vary across broader geographic and environmental gradients