Plant populations can undergo very localized adaptation, allowing widely distributed populations to adapt to divergent habitats in spite of recurrent gene flow. Neotropical trees –whose large and undisturbed populations often span a variety of environmental conditions and local habitats – are particularly good models to study this process. Here, we explore patterns of adaptive divergence from large (i.e. regional) to small (i.e. microgeographic) spatial scales in the hyperdominant Amazonian tree Eperua falcata Aubl. (Fabaceae) under a replicated design involving two microhabitats (~300 m apart) in two study sites (~300 km apart). A three-year reciprocal transplant illustrates that, beyond strong maternal effects and phenotypic plasticity, genetically driven divergence in seedling growth and leaf traits was detected both between seedlings originating from different regions, and between seedlings from different microhabitats. In parallel, a complementary genome scan for selection was carried out through whole-genome sequencing of tree population pools. A set of 290 divergence outlier SNPs was detected at the regional scale (between study sites), while 185 SNPs located in the vicinity of 106 protein-coding genes were detected as replicated outliers between microhabitats within regions. Outlier-surrounding genomic regions are involved in a variety of physiological processes, including plant responses to stress (e.g., oxidative stress, hypoxia and metal toxicity) and biotic interactions. Together with evidence of microgeographic divergence in functional traits, the discovery of genomic candidates for microgeographic adaptive divergence represents a promising advance in our understanding of local adaptation, which likely operates across multiple spatial scales and underpins divergence and diversification in Neotropical trees.