Several studies have documented dominance by few species in Amazonian forests. Dominance patterns in contrasting forest habitats in lowland Amazonia have revealed that dominant species tend to be either locally abundant (local dominants) or regionally frequent (widespread dominants) but rarely both (oligarchs). However, the mechanisms underlying dominance remain unclear. Here, assuming that species traits reflect ecological processes that can lead to dominance, we asked across different habitat types whether: (i) dominance is defined by specific functional profiles and (ii) dominance patterns (local dominants vs. widespread dominants) are associated with different functional traits. We combined census data from 503 forest inventory plots across four lowland forest habitats in western Amazonia with trait information for _{2600 tree species, encompassing data collected in the focal plots and data from published sources. We considered traits that relate to leaf, wood, seed and whole-plant strategies: specific leaf area (SLA), leaf area (LA), N content per unit leaf mass (LN), wood density (WD), seed mass (SM), and maximum diameter at breast height (DBH}max~).  Our results reveal that dominant species display different trait combinations depending on the habitat type where they dominate. Moreover, taller dominant species exhibit higher regional frequency, associated with higher dispersal ability, and lower local abundance, likely due to negative density dependence. Greater SM contributes to higher regional frequency of dominant species via greater dispersal ability and seedling survival. Finally, traits related to resource conservation strategies, such as lower SLA, LA, LN and greater WD, favor higher local densities across most habitats, while the opposite pattern was linked to higher regional frequency. Synthesis: Our findings reveal that (i) dominance is associated with different functional traits depending on the habitat type, and (ii) different functional trait values define distinct dominance patterns. Our study exemplifies the potential of trait-based approaches to illuminate the ecological mechanisms that may underlie dominance in tropical forests. Finally, accounting for both local abundance and regional frequency when studying dominance is likely to improve our understanding and forecasting of how different species will respond to global change drivers in western Amazonia.

We used an extensive dataset consisting of 503 forest inventory plots, ranging from 0.025 to 0.213 ha, across different western Amazonian forests, from Colombia to Bolivia. Plots covered the four main habitat types found in western Amazonia proportionally to their extension in the region, with 76% in terra firme forests, 11% in floodplains, 7% in swamps and 6% in white sands. All trees ≥ 2.5 cm DBH were recorded and identified to species, representing a total of 93,719 individuals, and 2609 species. 

We defined dominant species as those that together account for 50% of the total relative abundance of each habitat type, following ter Steege et al. (2013) and Matas-Granados et al. (2024). We identified dominant species separately by habitat type. We defined two main aspects of dominance: (1) local abundance of dominant species, calculated as the mean local relative abundance (individuals of the species in a plot divided by total individuals in the plot), averaged across the plots where each dominant species occurred, and (2) regional frequency of dominant species, calculated as the number of plots where a species occurred divided by total plots in the habitat type.

We considered six plant functional traits as they related to resource acquisition, dispersal, defense, and competitive ability and can capture species differences in their ecological strategies: specific leaf area; leaf area; N content per unit leaf mass; maximum diameter at breast height as a proxy of maximum plant height; wood density; and seed mass

We used the plot inventories to determine the maximum diameter at breast height as the 95^th percentile of each species. Other trait data was compiled from various sources following standardized protocols, including (i) previous work by our research groups, (ii) publicly available trait databases such as TRY, funAndes, the Seed Information Database, and the Global Wood Density Database, and (iii) additional sources for seed mass. When more than one measure per species was available, we calculated the species mean trait value. The compiled trait dataset represented between 29% (for SM) and 100% (for DBH_max) of our 2609 species, and all species had information for at least one of the six traits. We log transformed all trait values (except WD) for subsequent analyses. As protocols can vary between studies, we compiled data for SLA and LA including and excluding petiole, and data for WD taken from the branch or from the sapwood and heartwood. Species coverages were greater when using SLA and LA measured without the petiole and WD measured from the branch, and they were highly correlated with SLA and LA measured with petiole and WD measured from the trunk, respectively.

To evaluate trait differences between species, considering their regional abundance across multiple and single dimensions, we conducted both multivariate and univariate analyses.

First, we performed principal component analyses (PCAs) with scaled functional trait values to characterize the main dimensions of functional variation among all species within each habitat type separately and to illustrate trait relationships. We extracted species scores for axes 1 and 2, as these together accounted for more than 55% of functional variation across all habitat types. To test the relationship between each main axis of functional variation and species regional abundance, we built two linear models (LMs) for each habitat type separately: one with axis 1 as the dependent variable and the other with axis 2. In both models, species regional abundance was the explanatory variable. Given that few species had values for all six traits (16% of all species), we repeated the analyses without SM (34% species had values for the remaining five traits) and presented these results in the main text. Second, for each trait separately, we conducted LMs to test the relationships between functional traits and species regional abundance.

To account for potential bias due to disproportionate sampling of species functional traits clustered in specific lineages (i.e., some lineages could be more represented than others), we subsampled one species per genus from the species list of each habitat type 100 times and performed all the multivariate and univariate analyses each time to compare the subsampled results to our observed results.

To explore the role of traits in the two variables of dominance measured to each dominant species (local abundance and regional frequency) in each of the four habitat types, we built Bayesian models. We built a unique model for each combination of dependent variable (local abundance and regional frequency) and single trait, resulting in 12 models. All traits were rescaled to facilitate model comparisons. We fitted all models with weakly informative priors. Model convergence was tested visually with trace and density plots and numerically estimating if Rhat was equal to one. Models usually converged after 4000 iterations. Model fit was evaluated using Bayesian R². We focus on the interaction between trait and habitat (i.e., the slopes of the relationships). All analyses were conducted in R v4.1.3.

Changes after Nov 14, 2025: We have included the compiled trait database with species mean trait values