1. Mega dams in lowland tropical forests often create large archipelagos, leading to biodiversity decay and disruption of ecosystem functioning in remnant habitat islands.

2. We investigated the functional diversity and functional trait filtering of aerial insectivorous bats in both insular forest patches created by a vast ~30-yr-old hydropower reservoir and the adjacent mainland continuous forest in Central Amazonia.

3. Bats were surveyed using passive bat recorders across 34 forest sites. Based on a set of morphological traits derived for each species recorded, we estimated both the bat functional richness, functional evenness and functional dispersion at each surveyed site. We further assessed the effects of local vegetation, patch and landscape features on patterns of functional diversity. The interaction between functional traits, environmental characteristics, and species distribution was investigated using a combination of RLQ and fourth-corner analyses.

4. We found that mainland sites retained higher functional richness and lower functional evenness compared to forest islands, indicating a more complete functional assemblage in the mainland. Additionally, species composition was affected by local vegetation structure and forest area, with small isolated islands exhibiting pervasive loss of functional traits. RLQ and fourth-corner analyses showed that larger understorey foraging species with greater dispersal capacity, constant frequency-frequency modulated calls, and higher frequency of maximum energy were associated with more isolated small islands. Conversely, forest subcanopy species, exhibiting quasi-constant frequency calls and presenting low dispersal capacity were associated with continuous forests and islands with greater forest area, and were therefore more sensitive to habitat insularization.

5. Our study calls attention to the pervasive impacts induced by large dams on the functional diversity of tropical insectivorous bats. We recommend that future assessments of the effects of habitat fragmentation on mammals should include traits linked to ecosystem services. In designing and licensing new dams, we suggest the creation of extensive protected areas surrounding mainland forests to minimize the detrimental impacts of small isolated islands and safeguard the full complement of key ecological functions provided by insectivorous bats.

Bat surveys

Between July and November 2016, we sampled aerial insectivorous bats using a Song Meter 2 BAT+ ultrasound recorder, connected to an SMX-US ultrasonic omnidirectional microphone (Wildlife Acoustics, Maynard, Massachusetts) positioned 1.5 m above ground at the centre of each plot. Each recorder was programmed to operate from dusk to dawn (17:30h to 06:30h) for four consecutive nights, covering the activity period of all bat species. Thus, each plot was sampled for 52h for a total of 1,768 recorder-hours over 136 sampling nights. Each recorder was configured to passively record bat activity in real time. To improve the range and quality of the ultrasound pulse recordings, we set the sample rate to 384 kHz with 32-bit resolution and only sounds with 18 dB exceeding ambient noise were considered. Ultrasounds were saved in WAC format with 30-min, producing 26 files every 13 hours of recordings. Using the Kaleidoscope 4.0.3 program (Wildlife Acoustics, Maynard, Massachusetts), these files were converted into WAV format and subdivided into 5-sec files. Spectrograms with at least two echolocation pulses of a single species or phonic complexes (species with similar calls that cannot be confidently distinguished) in each 5-sec file were defined as a ‘bat pass’, and this is considered as the unit of measure for bat activity. This study was undertaken under licenses for scientific purposes from Instituto Chico Mendes de Conservação da Biodiversidade - ICMBio (#53340-1). As this study did not involve the bat capture, it did not require approval from an ethics committee.

All recordings were inspected manually using the Kaleidoscope 4.0.3 program, and all species/phonic complexes were identified using the Amazonian bat ultrasound key sourced from López-Baucells et al. (2016). We separated Pteronotus parnellii into two sonotypes according to Pavan et al. (2018): Pteronotus 55 kHz = P. rubiginosus and Pteronotus 60 kHz = P. alitonus. Eptesicus brasiliensis and E. chirichinus were classified as a single phonic complex (Eptesicus brasiliensis/chirichinus) because of overlapping ultrasound characteristics (López-Baucells et al., 2016). We excluded all Phyllostomidae and Molossidae species from the analysis because these families cannot be reliably identified.

Species traits

We carefully selected seven functional traits that are widely used to assess the functional composition of bat assemblages across tropical forest landscapes, including two ultrasound characteristics (frequency of maximum energy and call structure), three body measurements (body mass, relative wing loading and aspect ratio), and vertical stratification of flight activity as a measure of forest habitat use.

Environmental variables

We used five local predictor variables to represent the floristic structure and vegetation clutter of each sampling site based on vegetation surveys within each of our forest plots. In particular, we assessed the (i) number of adult tree species (i.e., diameter at breast height ≥10 cm), (ii) number of adult tree stems (tree abundance), (iii) forest basal area, (iv) fire severity and (v) vegetation clutter. Variables (i) to (iv) were obtained from a previous study within the same forest plots (see Benchimol & Peres, 2015 for further details). Fire severity was a categorical variable derived from the extent to which each island had been affected by fires during the El Niño drought of late 1997 to early 1998 (Benchimol & Peres, 2015) and the number of charred trees and height of char marks on each tree; burnt islands have been regenerating and therefore we assume that fire severity can be a robust proxy for secondary vegetation. We further quantified vegetation clutter by placing 3 m × 3 m white panel sheets on each of the four sides of the recorder, which were photographed from 8 m (see Marciente et al., 2015). Using the ImageJ 1.8.0 software, we then thresholded these photos into black & white images, with black pixels representing the vegetation, and then estimated the mean proportion of vegetation from the four panels per plot.

We also obtained four patch and landscape variables: (i) forest area (in hectares), (ii) the shortest distance from each survey island to the nearest continuous forest (in meters), (iii) shortest distance from the recorder to the edge of the islands and the mainland (in meters), and (iv) the amount of landscape-scale forest cover. We used RapidEye© high-resolution (5-m pixel) imagery of the Balbina archipelago to quantify the patch and landscape variables. Forest area was measured for islands and continuous forest. To assess forest area in continuous forests, we assigned a value ten-fold larger than our largest island of 1,690.04 ha, a widely used procedure in fragmented landscape studies to include continuous forests in the analyses (Benchimol & Peres, 2015). Forest cover around each island (i.e., old-growth and/or secondary vegetation in different stages of regeneration) was measured considering four different spatial scales (within a buffer radius of 500 m, 1000 m, 1500 m and 2000 m from each recorder), in order to cover the home range sizes of different aerial insectivorous bat species. We then obtained the proportion of forest cover within each landscape using the QGIS 2.18.14 software.

References

Benchimol, M., & Peres, C. A. (2015). Predicting local extinctions of Amazonian vertebrates in forest islands created by a mega dam. Biological Conservation, 187, 61–72. http://dx.doi.org/10.1016/j.biocon.2015.04.005

López-Baucells A., Rocha R., Bobrowiec P.E.D., Bernard E., Palmeirim J. & Meyer C. (2016). Field guide to Amazonian bats (1st ed). Editora INPA.

Marciente, R., Bobrowiec, P. E. D., & Magnusson, W. E. (2015). Ground-vegetation clutter affects phyllostomid bat assemblage structure in lowland Amazonian forest. PLoS ONE, 10(6), e0129560. https://doi.org/10.1371/journal.pone.0129560

Pavan, A. C., Bobrowiec, P. E. D., & Percequillo, A. R. (2018). Geographic variation in a South American clade of mormoopid bats, Pteronotus (Phyllodia), with description of a new species. Journal of Mammalogy, 99(3), 624–645. https://DOI:10.1093/jmammal/gyy048