Functional traits help understand biological diversity and the mechanism by which ecological communities are structured and how they respond to the environment. For example, the high tree species diversity within tropical forests can be grouped into a few functional attributes like wood density, size, and dependence on animal pollination or seed dispersal. However, little is known about how these traits influence animal taxonomic and functional diversity. We carried out a vegetation census on six plots (20 x 100 m) within the National Forest of Carajás (Amazon biome) to identify forest canopy species and their functional traits. Within the same plot, we also applied three bee sampling methods (entomological nets, honey traps, and scent traps). By characterizing the functional traits of trees and bees, we were able to predict bee functional diversity better than with taxonomic diversity alone via combinations of tree traits like size, wood density, dependence on pollinators, and extinction risk. We found that the basal area of trees with low wood density was negatively associated with small, eusocial tree cavity-nesting bees. The richness of medium-sized solitary bees was positively associated with the richness and abundance of trees with extinction risk. The community dominance (average diameter at the basal area) of pollinator—dependent trees was negatively associated with the richness of aboveground and cavity-nesting bees. Our findings suggest that tree community composition limits the availability of nesting resources for specific bee groups. Moreover, the presence of trees with high conservation value was associated with a greater variety of bee traits and was the only metric associated with overall bee richness. As expected, functional traits shed light on the mechanism that might drive high diversity within tropical forests. Moreover, there seems to be complementarity in terms of conservation value and carbon stock potential, as areas harboring tree species with extinction risk and higher wood density are also those with overall greater bee and functional diversity. Finally, our study can contribute to the restoration of plant—pollinator community by providing an understanding of the vegetation community that contributes to biodiversity maintenance.   

1. Vegetation sampling

The Carajas National Forest protected area comprises a set of mountain ranges covered with pristine Amazon Forest. Within the National Forest, we implemented vegetation census on 6 plots. Each plot consisted of a 20 m x 100 m area in which all the plants whose diameter at breast height (DBH) was greater than 10 cm were identified to the species level, and their DBH was measured. The minimum distance between plots was more than 2.5 km. For all trees, palms, and vines, we measured height, and DBH and assessed wood density and dependency on bee pollination from the literature (Table S3). For species with no data available (20%) we estimate wood density using the average wood density among the genera of that same individual for which wood density was missing. As we focused on trees, palms and vines, one vegetation census was performed in each plot over two years (2021 and 2022) of sampling, and the vegetation census occurred independently of bee sampling (2022).  

Additionally, we classified species according to their vulnerability to extinction status, based on the International Union for Conservation of Nature (IUCN) Red List. Species were considered threatened if they were classified as endangered, vulnerable, or near threatened. The species classification is based upon one or multiple criteria, including population size decline, geographic range, and very small populations. Species classification is based upon one or multiple criteria, including population size decline, geographic range, and very small populations. Although threatened to extinction is not a function associated with a process, it can be considered a function in terms of a species’ shared susceptibility to stressors (all traits are functional traits).

We considered five functional traits: i) Plant size (DBH), ii) wood density, iii) height, iv) melittophily and v) extinction risk (threatened or not). To characterize the community considering categorical traits, we considered the abundance of each level (e.g., the abundance of melittophilous plants) and the aggregated plant size (sum of the DBH of individuals with a specific trait level) to characterize the dominance of functional traits in the community.  For wood density, for example, we first categorized trees using a wood density threshold of 0.68 g/cc (median), grouping species as either having low wood density or high wood density. Then, for each category, we calculated the average, maximum value and accumulated total plant size (sum) of low- and high—wood—density trees. The same was done for all bee pollination dependent species and for all tree species with some level of extinction threat.    

2. Bee sampling

We applied three complementary protocols for sampling bees in each vegetation sampling plot: 1) We actively collected bees using entomological nets by sampling across flowering plants in the understory. We collected bees from flowers by walking in the sampling area (plot) for two non-consecutive hours, one-hour sampling, at 30-minute intervals, followed by a second hour of sampling. Sampling was performed by one researcher in each area between 7 and 10 am. Three independent researchers collected the samples at the same time; thus, all areas were sampled on two consecutive days. 2) We used a honey trap, which consisted of a mixture of honey and water (1:2, i.e. 250 ml of honey and 500 ml of water) that was sprayed on a square meter on a plant with broad leaves chosen at random at the same sampling point. We performed two 10-minute bee samples (using entomological nets) from the honey traps (first between 8 and 9 am, and second between 10 and 11 am), with one-hour intervals between each sample. Finally, 3) we used a scent trap (four different scents were used), which was most effective for attracting bees from the Euglossini group (orchid bees). We used two sets of traps (with four scents: eucalyptol, eugenol, methyl salicylate and vanillin), and each set was placed at 2- and 20 m in height and remained in place for 48 h. Each of the sampling protocols was applied three times in each sampling area. Honey and scent traps were placed in the center of the plot, and flower visitors were actively collected through the plot. In total, we applied three sampling protocols (flower visitors, honey and scent traps) three times (March, May and August of 2022) in each area. We identified all bees to the species level, and we deposited them in Museu Paraense Emilio Goeldi (Belém, Pará, Brazil).

Using literature information, specimens’ measurements, and the species taxonomy, we were able to classify the bees accordingly to their nesting traits (type and location), body size and sociality. Nesting types were either below- or aboveground. Nesting location refers to whether a bee nested in tree cavities or not (other types were not considered due to the low number of species, cleptoparasitic and exposed). Body size was classified into three size classes according to the intertegular distance: small (0.9–2.1 mm), medium (2.2–3.9 mm) and large (>3.9 mm), following the classification from Borges et al., 2020. Moreover, we grouped them accordingly to their social behavior (eusocial, solitary and semisocial).

3. Functional trait richness

For each functional trait categorized either for plants (DBH, wood density, height, melittophily, or extinction risk) or bees (nesting type, nesting location, body size, and sociality) we also calculated the taxonomic richness within each categorical functional trait presented. For instance, the richness of bee-pollinated plants in each plot or the richness of plants with extinction risk per plot. Additionally, we calculated the Shannon index for each plot based on plant species abundance. For continuous variables, we also calculate average and total plot values within each category. The total plot value of plant size is aggregated per functional trait; for example, the size (DBH) of plants pollinated by bees is considered a proxy of how much of the plot’s canopy is dominated by plants that depend on bees for pollination. In the case of bee traits, a similar approach was implemented: we calculated the richness of small bees, for example, or the richness of bee that nest aboveground, or that are classified as solitary. Additionally, we calculated functional diversity metrics using the categorical and continuous abovementioned functional traits, with the functional diversity function of the FD package in r (R core Team, 2023).