Abstract: A major challenge remains to understand the relative contributions of history, dispersal, and environmental filtering to the assembly of hyperdiverse communities across spatial scales. Here, we examine the extent to which biogeographical history and habitat specialization have generated turnover among and within lineages of Amazonian trees across broad geographic and environmental gradients. We replicated standardized tree inventories in 102 0.1-ha plots located in two distant regions—the western Amazon and the eastern Guiana shield. Within each region, we used a nested design to replicate plots on contrasted habitats: white-sand, terra firme, and seasonally flooded forests. Our plot network encompassed 26,386 trees that together represented 2,745 distinct taxa, which we standardized across all plots and regions. We combined taxonomic and phylogenetic data with detailed soil measurements and climatic data to: (1) test whether patterns of taxonomic and phylogenetic composition are consistent with recent or historical processes, (2) disentangle the relative effects of habitat, environment, and geographic distance on taxonomic and phylogenetic turnover among plots, and (3) contrast the proportion of habitat specialists among species from each region. We found substantial species turnover between Peru and French Guiana, with only 8.8% of species shared across regions; genus composition remained differentiated across habitats and regions, whereas turnover at higher taxonomic levels (family, order) was much lower. Species turnover across plots was explained primarily by regions, but also substantially by habitat differences and to a lesser extent by spatial distance within regions. Conversely, the composition of higher taxonomic levels was better explained by habitats (especially comparing white-sand forests to other habitats) than spatial distance. White-sand forests harbored most of the habitat specialists in both regions, with stronger habitat specialization in Peru than in French Guiana. Our results suggest that recent diversification events have resulted in extremely high turnover in species and genus composition with relatively little change in the composition of higher lineages. Our results also emphasize the contributions of rare habitats, such as white-sand forests, to the extraordinary diversity of the Amazon and underline their importance as conservation priorities.

Abstract: Across the evolutionary history of insects, the shift from nitrogen-rich carnivore/omnivore diets to nitrogen-poor herbivorous diets was made possible through symbiosis with microbes. The herbivorous turtle ants Cephalotes possess a conserved gut microbiome which enriches the nutrient composition by recycling nitrogen-rich metabolic waste to increase the production of amino acids. This enrichment is assumed to benefit the host, but we do not know to what extent. To gain insights into nitrogen assimilation in the ant cuticle we use gut bacterial manipulation, 15N isotopic enrichment, isotope-ratio mass spectrometry, and 15N nuclear magnetic resonance spectroscopy to demonstrate that gut bacteria contribute to the formation of proteins, catecholamine cross-linkers, and chitin in the cuticle. This study identifies the cuticular components which are nitrogen-enriched by gut bacteria, highlighting the role of symbionts in insect evolution, and provides a framework for understanding the nitrogen flow from nutrients through bacteria into the insect cuticle.

Abstract: Ants are a major ecological group in tropical rainforests. Few studies in the Neotropics have documented the distribution of ants from the ground to the canopy, and none have included the understorey. A previous analysis of an intensive arthropod study in Panama, involving 11 sampling methods, showed that the factors influencing ant β diversity (i.e., changes in assemblage composition) were, in decreasing order of importance, the vertical (height), temporal (season), and horizontal (geographic distance) dimensions. In the present study, we went one step further and aimed (1) to identify the best sampling methods to study the entire ant assemblage across the three strata, (2) to test if all strata show a similar horizontal β diversity and (3) to analyze the functional structure of the entire ant assemblage. We identified 405 ant species from 11 subfamilies and 68 genera. Slightly more species were sampled in the canopy than on the ground; they belonged to distinct sub-assemblages. The understorey fauna was mainly a mixture of species found in the other two strata. The horizontal β diversity between sites was similar for the three strata. About half of the ant species foraged in two (29%) or three (25%) strata. A single method, aerial flight interception traps placed alongside tree trunks, acting as arboreal pitfall traps, collected half of the species and reflected the vertical stratification. Using the functional traits approach, we observed that generalist species with mid-sized colonies were by far the most numerous (31%), followed by ground- or litter-dwelling species, either specialists (20%), or generalists (16%), and arboreal species, either generalists (19%) or territorially dominant (8%), and finally army ants (5%). Our results reinforce the idea that a proper understanding of the functioning of ant assemblages requires the inclusion of arboreal ants in survey programs.

Abstract: Automatic tree crown segmentation from remote sensing data is especially challenging in dense, diverse, and multilayered tropical forest canopies, and tracking mortality by this approach is even more difficult. Here, we examine the potential for combining airborne laser scanning (ALS) with multispectral and hyperspectral data to improve the accuracy of tree crown segmentation at a study site in French Guiana. We combined an ALS point cloud clustering method with a spectral deep learning model to achieve 83% accuracy at recognizing manually segmented reference crowns (with congruence >0.5). This method outperformed a two-step process that involved clustering the ALS point cloud and then using the logistic regression of hyperspectral distances to correct oversegmentation. We used this approach to map tree mortality from repeat surveys and show that the number of crowns identified in the first that intersected with height loss clusters was a good estimator of the number of dead trees in these areas. Our results demonstrate that multisensor data fusion improves the automatic segmentation of individual tree crowns and presents a promising avenue to study forest demography with repeated remote sensing acquisitions.