Understanding the patterns of bird diversity and its driving force is necessary for bird strike prevention. In this study, we investigated the effects of landscape attributes on phylogenetic and functional diversity of bird communities at Nanjing Lukou International Airport (NLIA). Bird identifications and counting of individuals were carried out from November 2017 to October 2019. Based on the land cover data, the landscape was divided into four main types, including farmlands, woodlands, wetlands and urban areas. Bird phylogenetic and functional diversity were strongly affected by landscape matrix types. Bird species, phylogenetic and functional richness were highest in woodlands, while phylogenetic and functional structure were most complex in wetlands. We found that related carnivorous birds occupied most niches at the NLIA. Moreover, bird assemblages exhibited phylogenetic and functional clustering in all landscape matrix types, which meant environmental filtering governed community assembly. Landscape attributes of different matrix types influenced phylogenetic and functional diversity of bird communities. Landscape diversity affected bird functional structure positively, and the mean of patch area in the landscape had the strongest explanatory effects. To prevent bird strikes, we suggest that more small-area woodlands with low levels of aggregation can be considered and farmlands should be patchier and more regular. At landscape level, the number of patches can be adjusted to reduce the frequency of bird activities.

Wavelet analysis, a spectral decomposition technique, was used to determine the key spatial scale that was important to quantify the relative influence of matrix types on species' distributions. We generated 1, 000 random points within 8 km radius extent from the NLIA and extracted their land-cover types. Then we ran the wavelet analysis with MATLAB 9.9 (MathWorks, Natick, Massachusetts; https://ww2.mathworks.cn/products/matlab.html). The result showed that a 288-m radius from the center of each point was appropriate. We then randomly selected 9 sampling points in farmlands, 7 in wetlands, 5 in woodlands and 4 in urban areas (25 points in total) to make sure that sampling effort on each matrix type was roughly proportional to its area. Sampling points were at least 576m apart. From November 2017 to October 2019, we carried out point-count bird surveys monthly. Each observation was in 20 minutes. We usually surveyed birds on sunny and windless days. We recorded all birds seen or heard and excluded fly-overs to obtain the birds data. Considering that the combination of functional groups and landscape matrices enables a comprehensive assessment of bird diversity, we chose three main functional groups based on the traits data collected by Wang et al. 2021.

We cut the global phylogenetic tree of birds by subsampling 5,000 'Hackett All species: a set of 10,000 trees with 9993 OTUs each' trees from BirdTree (http://birdtree.org). Then we constructed a new maximum clade credibility tree with 0.5 posterior probability limit by the software TreeAnnonator v1.10.4 in the BEAST package v1.10.4. We focused on 4 kinds of functional traits relevant ecologically to bird strikes: morphological characteristics (body length, bill length, wing length, tarsus length and body mass), nest sites (ground, water, shrub, canopy and wall), feeding behaviors (carnivores, insectivores, omnivores, granivores and piscivores) and foraging strata (below water surface, ground, understory and midhigh). Morphological characteristics were continuous values and others were treated as binary traits. 

Based on the new phylogenetic tree constructed, we calculated Faith's phylogenetic distance (Faith's PD) to describe the total sum of phylogenetic history. We also calculated mean pairwise distance (MPD) and mean nearest taxon distance (MNTD) to represent phylogenetic structure. We computed functional richness (FRic) and functional dispersion (FDis) to represent functional diversity and structure. All metrics were calculated using R package 'picante' and 'FD'. Species richness (SR), MPD, MNTD and FDis of functional groups were also calculated to evaluate phylogenetic and functional structure of different functional groups.

We calculated landscape metrics of the class level and landscape level using R package 'landscapemetrics'. At the class level, number of patches, edge density, mean of patch area, mean perimeter-area ratio, mean fractal dimension index and aggregation index of each matrix type were calculated. At the landscape level, number of patches, mean of patch area, effective mesh size, mean fractal dimension index, mean perimeter-area ratio and Shannon's diversity index were calculated.

The data files included with our submission can be opened with R or RStudio for further analysis.