We trapped a total of 648 bees, including 46 different species
representing two families, the Halictidae (72.2%) and Apidae
(27.8%) (Supplementary material, Appendix S2). The observed species
accumulation curve (Fig. 3) approached but did not reach a
plateau, and the estimated species richness in the region was 59
species. For the entire bee community, the most critical habitat
variables for predicting abundance were tree species richness
(site-stratified p-value = 0.020) and percent canopy cover
(Fig. 4a). Overall bee species richness was most impacted by coffee
bush density (site-stratified p-value = 0.045) and flowering tree
species richness (Fig. 4b). Overall bee community composition
(tribe level) did not vary substantially based on a gradient of forest
cover or tree species richness, but did show a weak trend of
increasing tribe richness with increasing tree species richness
(Fig. 5). Given the diverse composition of the sampled bee community,
and the likely divergence in responses to agroforestry management
depending on (1) sociality, (2) evolutionary history
(tribe), and (3) nesting guild, we separately examined bee abundance
in response to habitat variables within these three groups.
While these analyses are not completely independent of one another
(e.g., many social bees are wood-nesting), we feel that it is
valuable to examine the responses of conventionally defined
groups in order to compare the findings with those of previous
studies. Given that grouping based on tribes by definition clusters
individuals that share an evolutionary history, we expect that this
grouping will have the most predictive power.