Understanding different dominance patterns in western Amazonian forests

Dominance of neotropical tree communities by a few species is widely documented, but dominant trees show a variety of distributional patterns still poorly understood. Here, we used 503 forest inventory plots (93,719 individuals ≥ 2.5 cm diameter, 2,609 species) to explore the relationships between local abundance, regional frequency, and spatial aggregation of dominant species in four main habitat types in western Amazonia. Contrary to the widely supported positive abundance-occupancy relationship in ecology, we found that among dominant Amazonian tree species, there is a strong negative relationship between local abundance and regional frequency and/or spatial aggregation across habitat types. Our findings suggest an ecological trade-off whereby dominant species can be locally abundant (local dominants) or regionally widespread (widespread dominants), but rarely both (oligarchs). Given the importance of dominant species as drivers of diversity and ecosystem functioning, unraveling different dominance patterns is a research priority to direct conservation efforts in Amazonian forests.Funding provided by: Ministerio de Economía y CompetitividadCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100003329Award Number: CGL2016–75414–PFunding provided by: Ministerio de Ciencia e InnovaciónCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100004837Award Number: PID2019-105064GB-I00Funding provided by: Ministerio de Economía y CompetitividadCrossref Funder Registry ID: http://dx.doi.org/10.13039/501100003329Award Number: CGL2015-72431-EXPWe used data from 503 forest inventory plots spread across western Amazonia, from Colombia to Bolivia. A total of 363 plots had an area of 0.1 ha, 134 plots were smaller than 0.1 ha (range from 0.025 to 0.08 ha), and 6 plots were larger (range from 0.128 to 0.213 ha). Plots are included in the RedGentry network. Across all plots, we measured stems with a diameter at breast height ≥ 2.5 cm within the plot limits. Plots covered the main four habitat types in western Amazonia: 383 in terra firme (76%), 54 in floodplain (11%), 35 in swamp (7%) and 31 in white sand (6%) forests. We excluded all individuals not identified to species level (mean 14% of individuals per plot), since plot data came from different projects and morphospecies were not cross-checked. We also excluded individuals from doubtful identifications, e.g. 'cf.' and 'aff.' (mean 3% of individuals per plot). To the remaining individuals, we checked species names for synonym and spelling mistakes, using the R package 'Taxonstand'. Identifications that were difficult to designate to a species were considered morphospecies and were also removed. Finally, we cross-checked our species names list against the most recent checklists of Amazonian species. Species not found in these checklists (635 species) were compared with collection records in the Tropicos database, and were excluded because: 572 species of them were growth forms not consistently included in all datasets (epiphytes, lianas, herbs and ferns), 25 were illegitimate Amazonian species with ranges outside of our region and 38 species were considered wrong identifications because they do not have recorded collection since their descriptions. After these filters, 2,609 species and 93,719 individuals remained available for our analyses. Since plot size varied among datasets, we transformed abundances into relative abundances (i.e., number of individuals per species/total individuals per plot). Then, we defined dominant species as those species that together accounted for 50% of the total relative abundance of all individual trees in each habitat. We analyzed separately dominant species by habitat type. Since our plots are not evenly distributed in space, identifying dominant species considering all plots in each habitat type could favorize the selection of spatially clumped species. To explore the effect of this potential bias, we extracted 100 random subsamples from the complete set of plots in each habitat type at minimum pairwise plot distances of 0.5 km, 1 km, 2 km, and 5 km. We identified dominant species in each subsample at each minimum pairwise plot distance. To test the relationship between local abundance and regional frequency of dominant species by habitat type, we built beta regression models with a logit link function. We used the mean local relative abundance of each dominant species as the dependent variable and both the regional relative frequency and the habitat type as predictors. We built species-level rank abundance distribution graphs within each habitat type to explore if local abundance in each plot of each dominant species gave similar information that their mean local abundance. We conducted these analyses for: i) the complete dataset, including all plots of each habitat type; and ii) for each subsample at each minimum pairwise plot distance. To study the relative spatial aggregation of species, we analyzed the co-dominance of each species at each spatial extent and habitat. To do so, we calculated the F index related to each geographical distance between plots to all species and relativized these values to the community-level aggregation curve