Besides its extreme climate conditions, the Caatinga (a type of tropical seasonal forest) hosts an impressive faunal and floristic biodiversity. In the last 50 years there has been a considerable increase in the number of studies in the area. Here we aimed to present a review of these studies, focusing on four main fields: vertebrate ecology, plant ecology, human ecology, and ethnobiology. Furthermore, we identify directions for future research. We hope that the present paper will help defining actions and strategies for the conservation of the biological diversity of the Caatinga.

Aim: Among the world's three major nectar-feeding bird taxa, hummingbirds are the most phenotypically specialized for nectarivory, followed by sunbirds, while the honeyeaters are the least phenotypically specialized taxa. We tested whether this phenotypic specialization gradient is also found in the interaction patterns with their floral resources.Location: Americas, Africa, Asia and Oceania/Australia. Methods:We compiled interaction networks between birds and floral resources for 79 hummingbird, nine sunbird and 33 honeyeater communities. Interaction specialization was quantified through connectance (C), complementary specialization (H 2 0 ), binary (Q B ) and weighted modularity (Q), with both observed and null-model corrected values. We compared interaction specialization among the three types of bird-flower communities, both independently and while controlling for potential confounding variables, such as plant species richness, asymmetry, latitude, insularity, topography, sampling methods and intensity.Results: Hummingbird-flower networks were more specialized than honeyeaterflower networks. Specifically, hummingbird-flower networks had a lower proportion of realized interactions (lower C), decreased niche overlap (greater H 2 0 ) and greater modularity (greater Q B ). However, we found no significant differences between hummingbird-and sunbird-flower networks, nor between sunbird-and honeyeaterflower networks.

Functional traits can determine pairwise species interactions, such as those between plants and pollinators. However, the effects of biogeography and evolutionary history on trait‐matching and trait‐mediated resource specialization remain poorly understood. We compiled a database of 93 mutualistic hummingbird–plant networks (including 181 hummingbird and 1,256 plant species), complemented by morphological measures of hummingbird bill and floral corolla length. We divided the hummingbirds into their principal clades and used knowledge on hummingbird biogeography to divide the networks into four biogeographical regions: Lowland South America, Andes, North & Central America, and the Caribbean islands. We then tested: (a) whether hummingbird clades and biogeographical regions differ in hummingbird bill length, corolla length of visited flowers and resource specialization, and (b) whether hummingbirds' bill length correlates with the corolla length of their food plants and with their level of resource specialization. Hummingbird clades dominated by long‐billed species generally visited longer flowers and were the most exclusive in their resource use. Bill and corolla length and the degree of resource specialization were similar across mainland regions, but the Caribbean islands had shorter flowers and hummingbirds with more generalized interaction niches. Bill and corolla length correlated in all regions and most clades, that is, trait‐matching was a recurrent phenomenon in hummingbird–plant associations. In contrast, bill length did not generally mediate resource specialization, as bill length was only weakly correlated with resource specialization within one hummingbird clade (Brilliants) and in the regions of Lowland South America and the Andes in which plants and hummingbirds have a long co‐evolutionary history. Supplementary analyses including bill curvature confirmed that bill morphology (length and curvature) does not in general predict resource specialization. These results demonstrate how biogeographical and evolutionary histories can modulate the effects of functional traits on species interactions, and that traits better predict functional groups of interaction partners (i.e. trait‐matching) than resource specialization. These findings reveal that functional traits have great potential, but also key limitations, as a tool for developing more mechanistic approaches in community ecology. A free Plain Language Summary can be found within the Supporting Information of this article.

No abstract