The interplay between spatiotemporal overlap and morphology as determinants of microstructure suggests no 'perfect fit' in a bat-flower network

Abstract

Plant-pollinator interactions in diverse tropical communities are often predicted by a combination of ecological variables, yet the interaction drivers between flower-visiting bats and plants at the community level are poorly understood. We assembled a network between Neotropical bats and flowering plants to describe its macrostructure and to test the role of neutral and niche variables in predicting microstructure. We found a moderately generalized network with internally nested modules comprising functionally similar plant and bat species. Modules grouped bats and plants with matching degrees of specialization but had considerable overlap in species morphologies and several inter-module interactions. The spatiotemporal overlap between species, closely followed by morphology, and not abundance, were the best predictors of microstructure, with functional groups of bats also interacting more frequently with plants in certain vegetation types (e.g., frugivores within forests) and seasons (e.g., long-snouted nectarivores in the dry season). Therefore, flower-visiting bats appear to have species-specific niche spaces delimited not only by their ability to exploit certain flower types but also by preferred foraging habitats and the timing of resource availability. The prominent role of resource dissimilarity across vegetation types and seasons likely reflects the heterogeneity of Neotropical savannas, and further research in biomes beyond the Cerrado is needed to better understand the complexity of this system.

Figure 1

Spatiotemporal trends of bat capture rates of the most frequent floral visitors in the study site. Temporal trends in the relative frequency (in relation to all captured in a given month) of specialized nectarivorous bats (a) and mainly frugivorous bats (b) throughout the year. A dashed line separates the dry and rainy seasons. Spatial trends in the relative frequency of nectarivorous (c) and frugivorous bats (d) according to habitat type.

Figure 2

Interaction network between flower-visiting phyllostomid bats and plants in a savanna of central Brazil. Nodes represent species and lines, pairwise interactions. Line width corresponds to interaction weight (frequency) and node size to a species’ degree, or the sum of a species’ interactions. Plants are divided into chiropterophilous, non-chiropterophilous, or unknown syndromes. Modules in the network are divided by dashed lines and accompanied by a schematic illustration of the most important bat species in the module. Species codes are found in Supplementary Tables S1 and S2.

Figure 3

Density distribution of morphological variables from species in the network across modules. (a) Bats (RCR—rostrum-cranium ratio, and BCI—body condition index); (b) plants (FTL—floral tube length, and COD corolla outermost diameter). Module names correspond to those in Fig. 1. Solid red lines show the mean of each variable for all species pooled, and dashed lines indicate the standard deviation.

Figure 4

The likelihood of different interaction probability models of explaining pairwise interaction frequencies in the observed network. Model fit is expressed in their variation in the Akaike Information Criterion relative to the observed matrix. Models are organized from best fit (top) to worst fit (bottom). (a) Full network, (b) nectarivores only, and (c) other guilds only. Model labels—M: morphological specialization, A: relative abundance, S: spatial overlap, P: phenological overlap, Null: benchmark null model.