Most work in Amazonia has concentrated on dense lowland evergreen rain forest, a vegetation type with >40% cover. Large parts of southern Amazonia are covered by open evergreen lowland rain forest, physiognomically dominated by high abundance of palms. This vegetation type has received relatively little attention so far. Understanding the key predictors of above-ground biomass (AGB) across scales is important to accurately quantify the impacts of land cover change on the terrestrial carbon budget.Aims: We assessed the structure of southern Amazonian forests, Brazil, to quantify the relative importance of variation in AGB caused by the abundance/density of palm species and by forest structure.Methods: We stratified the landscape into homogeneous units in terms of vegetation types and elevation for using as a guide for plot establishment. We used the variation partitioning technique to decompose the relative contribution of forest structure and palm abundance.Results: The AGB(community) (including trees, palms and lianas) and AGB(tree) (excluding palms and lianas) significantly decreased with increasing abundance of palms. The Attalea speciosa, a large-leaved palm species, was the most important for explaining the variance of AGB. The total variance of AGB(tree) was partially explained by a redundant effect of A. speciosa and trees (28%) and by trees alone (62%), based on models of basal area. The redundant effect, alongside with additional analyses, indicated (1) competition between A. speciosa and Small trees and (2) covariation between A. speciosa and large trees.Conclusions: The abundance of palms plays a minor but significant role in predicting the AGB at the local scale in southern Amazonia.94409420David and Lucile Packard Foundation