Functional diversity of avian communities increases with canopy height: From individual behavior to continental‐scale patterns

Abstract Vegetation complexity is an important predictor of animal species diversity. Specifically, taller vegetation should provide more potential ecological niches and thus harbor communities with higher species richness and functional diversity (FD). Resource use behavior is an especially important functional trait because it links species to their resource base with direct relevance to niche partitioning. However, it is unclear how exactly the diversity of resource use behavior changes with vegetation complexity. To address this question, we studied avian FD in relation to vegetation complexity along a continental‐scale vegetation gradient. We quantified foraging behavior of passerine birds in terms of foraging method and substrate use at 21 sites (63 transects) spanning 3,000 km of woodlands and forests in Australia. We also quantified vegetation structure on 630 sampling points at the same sites. Additionally, we measured morphological traits for all 111 observed species in museum collections. We calculated individual‐based, abundance‐weighted FD in morphology and foraging behavior and related it to species richness and vegetation complexity (indexed by canopy height) using structural equation modeling, rarefaction analyses, and distance‐based metrics. FD of morphology and foraging methods was best predicted by species richness. However, FD of substrate use was best predicted by canopy height (ranging 10–30 m), but only when substrates were categorized with fine resolution (17 categories), not when categorized coarsely (8 categories). These results suggest that, first, FD might increase with vegetation complexity independently of species richness, but whether it does so depends on the studied functional trait. Second, patterns found might be shaped by how finely we categorize functional traits. More complex vegetation provided larger "ecological space" with more resources, allowing the coexistence of more species with disproportionately more diverse foraging substrate use. We suggest that the latter pattern was driven by nonrandom accumulation of functionally distinct species with increasing canopy height.


Table S1
Overview of our 21 sites.

Figure S1
Map of eastern Australia with our 21 sites

Figure S2
Vegetation profiles of our 21 sites

Figure S3
Summary vegetation characteristics of our 21 sites

Figure S4
Results of the PCA analysis applied on the vegetation characteristics forage on vegetation and the ground. We constantly and systematically scanned all vegetation for birds. We thus aimed to minimize bias introduced by locating only singing or otherwise conspicuous individuals. We located most of the birds by sight without using auditory cues (n = 1959), while 188 birds were detected due to singing and 470 due to vocalizing (usually contact voices among members of a group). Once we located a bird, we counted to five before recording its behaviour to avoid bias towards recording conspicuous behaviours. If it did not forage within 1 min, we left it and continued searching for another bird. We were interested in patterns of resource partitioning, and thus did not record the process of searching for food, but only an event of actually procuring or attempting to procure food (prey attack). First records of prey attacks might sometimes differ from subsequent ones (Recher and Gebski 1990) and thus sampling more prey attacks from the same individual could be useful. On the other hand, this could lead to underestimating uncertainties in quantifying foraging behaviour (Hejl et al. 1990). We thus compromised and for each individual recorded at most three prey attacks (mean = 2.25 attacks per individual bird, n = 2624 individual birds). For each prey attack, we recorded bird species (or genus, if species identification was impossible), foraging method and substrate, foraging height, height of the plant the bird foraged on, distance from the plant stem, and foliage density around the foraging bird.
In terms of behaviour, we recognized eight types of foraging methods used by birds for attacking the prey that we adapted from previous studies of foraging in Australian birds (e.g. Recher et al. 1985,           Figure S2-7 Correlations of functional diversity indices calculated for our 21 study sites using foraging methods and main substrates separately, then averaged (n = 8 and 8).

Appendix S3 Combinations of foraging behaviors for calculation of functional diversity (FD)
We defined several combinations of behavioral traits we recorded in the field. These were: 1) Foraging method (acronym "Method" in figures below; n = 8) 2) Main foraging substrate ("SubMain"; n = 8) 3

Shannon diversity
Appendix S4 Geographic effects (latitude and distance)

Latitude
Australia has a peculiar east-west climatic gradient instead of the usual north-south gradient.

Geographic distance
There are many metrics that are available to quantify pairwise multivariate distances between sites, with appropriate distance metrics selected based on the type of data. We calculated distance matrices using: i) Euclidean distances for continuous measurements (vegetation characteristics) ii) Bray-Curtis distances for categorical traits (foraging behavior) iii) Shortest geographical distance in km (geographical location) -these were the same irrespective of the package we used; we finally used the fields package (Nychka et al. 2017 Pairwise distances were organized into distance matrices. We analyzed them using multiple regressions on distance matrices in the ecodist package .

Distance-based measures
Appendix S8 Within-species analyses of substrate use diversity Substrate use diversity increased with increasing canopy height across study sites. One possible explanation is that each species might use more substrates in more complex vegetation. We thus tested the within-species relationship between the diversity in substrate use and canopy height, controlled for the number of foraging records per species per site. We did this only for the Brown Thornbill (Acanthiza pusilla, n = 263 foraging records) and the Yellow-faced Honeyeater (Caligavis chrysops, n = 193), the two species with the highest number of foraging records available (Fig. A8-1  The number of species also increased with increasing canopy height across study sites (Figs. S9-2 and S9-3). However, there was not a set of species that would be present across all sites, due to biogeographic turnover of avifaunas. There is a set of species present at all sites with canopy height above ca. 20 m, but otherwise the pattern has more turnover than nestedness compared to foraging substrates. An interesting feature is the occurrence of specific sets of species on only certain sites, e.g. Paluma or Herberton Rg. (Fig. S9-3).
Nestedness and turnover components of total beta diversity of species composition (mean (SD)). Partitioning of substrate use and species composition into turnover and nestedness components was calculated in the betapart  and BAT  packages. As the authors of these packages do not agree on the best partitioning algorithm, we present results of both approaches. We used both presence-absence and abundance data. Abundance is coded from absent (black) to few records (dark blue) to many records (dark violet).

References in Appendix S9
Occurrence is codes as present (red) vs. absent (yellow      Bellthorpe NP. Sites are numbered with increasing canopy height as in Table S1.  Figure S1 and Table S1. "Foliage_VegStrata" is the sum of foliage coverage of all vegetation strata at each site. "Foliage_HeightStrata" is the same for height strata. For the definition of vegetation strata and height strata, please see Appendix A1. PC1 and PC2 are the first two PC axes from a PCA analysis on all vegetation characteristics (see Fig. S4).