Predicting regional hotspots of phylogenetic diversity across multiple species groups

The protection of phylogenetic diversity has become a priority in conservation biology, but its achievement requires a detailed understanding of (a) hotspots of phylogenetic diversity on a management‐relevant scale and (b) the land use and climate factors determining local phylogenetic diversity. In this study, we identified spatial patterns of taxonomic and phylogenetic diversity and their environmental drivers.


| INTRODUC TI ON
Human population growth has led to an intensification of land use that often resulted in dramatic changes in species´ distributions and abundances (Habel, Samways, & Schmitt, 2019;Hooke, Martín-Duque, & Pedraza, 2012;Pimm & Raven, 2000). To slow down the rate of species loss, numerous conservation strategies have been developed (Gruber et al., 2012;Samways et al., 2020). Because range-size data for species are generally readily available, the focus of many of these strategies is the occurrence of enigmatic species or estimates of taxonomic diversity (Mammides, 2019;Miller, Jolley-Rogers, Mishler, & Thornhill, 2018;Zupan et al., 2014).
Phylogenetic diversity has recently gained increasing attention as a surrogate for the diversity of functional traits of species that support the maintenance of ecological processes (Forest et al., 2007;Srivastava, Cadotte, MacDonald, Marushia, & Mirotchnick, 2012). As such, it represents the ability of biological communities to respond to environmental changes (Winter, Devictor, & Schweiger, 2013). The recognition of phylogenetic diversity patterns is key to protecting both the unique features of biodiversity and the evolutionary history of local communities (Faith, 2013).
However, regions with comparatively high phylogenetic diversity can host biogeographically and phylogenetically unique species (Pouget et al., 2016). Phylogenetic diversity is also a surrogate for the historical factors underlying diversity patterns (Pinkert et al., 2018). By contrast, communities with disproportionally closely related species typically cluster in regions where environmental factors filter particular clades (Devictor et al., 2010;Faith, 1992;Tucker & Cadotte, 2013). Furthermore, patterns of phylogenetic diversity might differ considerably between taxonomic groups in the same area (Zupan et al., 2014), suggesting that single taxa are rather poor surrogates for the overall biodiversity. Nevertheless, the extent to which different species groups depict similar patterns of phylogenetic diversity in space remains largely unexplored, especially at scales relevant for conservation management.
Despite significant research effort on patterns and drivers of phylogenetic diversity during the last decade (Cadotte, Carscadden, & Mirotchnick, 2011;Cadotte & Davies, 2010;Cadotte, Dinnage, & Tilman, 2012), the results of those studies have seldom been integrated into conservation practices (Veron et al., 2015;Winter et al., 2013). This is in part due to the fact that decision-making in conservation practice is carried out by small and arbitrarily defined political entities (Schwartz, 1999), whereas most phylogenetic diversity studies focus on countries (Devictor et al., 2010;Graham, Parra, Rahbek, & McGuire, 2009) or continents (Thuiller et al., 2015;Zupan et al., 2014). Furthermore, these studies are based on large grain sizes, which makes it difficult to define local conservation priorities and can lead to strong biases in prioritizing conservation actions (Pouget et al., 2016). For example, Huang, Davies, and Gittleman (2012) found that estimates of the loss of phylogenetic diversity at a global scale underestimate the actual loss of local phylogenetic diversity. Here, we investigated the phylogenetic diversity of birds, bats, dragonflies, grasshoppers and butterflies, important species groups in conservation planning and decision-making, across Bavaria, a federal state in Germany. Our main objective was to identify regions in Bavaria where measures of taxonomic diversity fail to protect the phylogenetic diversity of the considered species groups. We therefore assessed the extent to which patterns in phylogenetic diversity are congruent with patterns of taxonomic diversity. Further, we assessed the environmental drivers of these patterns. In addition, we mapped hotspots of phylogenetic diversity of the five species groups.

| Study area
The study was conducted in the federal state of Bavaria (see Appendix S1a for major natural regions), located in south-eastern

| Biodiversity data
The dragonfly distribution data used in this study originated from "Libellen in Bayern" (see also Kuhn & Burbach, 1998), the grasshopper records from the "Mapping of Species Protection" (ASK database; Bavarian Environment Agency; Schlumprecht & Waeber, 2003), the bat data from "Fledermäuse in Bayern" (see also Meschede & Rudolph, 2004), data on breeding birds from "Atlas der Brutvögel in Bayern" (see also Rödl, Rudolph, Geiersberger, Weixler, & Görgen, 2012) and data on butterflies from "Tagfalter in Bayern" (see als Bräu et al., 2013). In addition, available data from the database of the Bavarian Environment Authority (Landesamt für Umwelt, 5 km × 5 km resolution) were used. The grid cells were adopted from the topographic map (TK) 1:25,000 of Bavaria and had an average area of ≈33.9 km 2 (TK ¼ grids) (see Appendix S6). Only grid cells whose area was entirely within Bavaria and which contained more than two species were included (Pfeifer, Müller, Stadler, & Brandl, 2009).

| Environmental data
The importance of land use in patterns of phylogenetic diversity was investigated using the percentage of cover of eight land use categories as extracted from CORINE (Coordinated Information of the European Environment). These data are based on a European-wide land cover mapping using LANDSAT-7 satellite images during the year 2000 (https://land.coper nicus.eu/pan-europ ean/corine-land-cover/ clc-2000).
This year was selected since most species data, including standardized surveys, were collected around the year 2000. Referring to the CORINE land cover classes we defined the following categories: conifer forest, broadleaf forest, mixed forest, sealed areas (including urbanization, industry, and traffic), meadows, arable land, streams and standing water.
The role of climate was tested using the annual mean temperature and annual precipitation data downloaded from WorldClim.org (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005). The data have a resolution of 30 arc seconds (0.93 km × 0.93 km = 0.86 km 2 ). All land use and climate variables were aggregated to our grid using R package raster (Hijmans et al,, 2018).

| Phylogenetic trees
The phylogeny of birds (see Appendix S2.1) was based on a Bayesian phylogenetic analysis (Hackett et al., 2008;Jetz, Thomas, Joy, Hartmann, & Mooers, 2012) available on www.birdt ree.org. One thousand samples (bias corrected according to Roff, 2006) were downloaded from the posterior distribution of topologies and branch lengths and applied to create a consensus tree using TreeAnnotator 1.8.2 in the software bundle BEAST (Drummond, Suchard, Xie, & Rambaut, 2012). The phylogeny of European dragonflies (Appendix S2.3) was taken from Pinkert et al. (2018).
Ultrametric branch lengths were estimated with a penalized likelihood approach based on a relaxed clock model of substitution rate variation among branches and a smoothing parameter λ = 1.0 using ape::chronos (Paradis, 2013), using a relaxed clock model of rate evolution and a smoothing parameter λ = 1.0.

| Data analysis
All analyses were performed using R version 3.5.1 (R Core Team, 2014). The species richness of assemblages within grids, as the sum of all co-occurring species, was calculated based on the distribution data of the species groups. To facilitate comparison with previous studies, the most common measure of phylogenetic diversity (i.e. Faiths PD; Faith, 1992) was calculated as the sum of branch lengths of the minimal spanning tree among co-occurring species using the pd function of the R package picante (Faith, 1992;Kembel et al., 2010). Because Faiths PD is inherently correlated with species richness (Pavoine & Bonsall, 2011), the observed phylogenetic diversity was controlled for species richness by calculating the residuals of a nonparametric regression of this relationship using the function loess (package stats with default settings) (Cleveland, Grosse, & Shyu, 1992). In addition, the 10 grids with the highest standardized phylogenetic diversity of all five studies species groups were mapped as hotspots. To visualize these common, regional hotspots, the residuals of each model for each species group were scaled by subtracting the mean from each value and dividing it by the standard deviation (function scale from the R package base).
The influence of land use and climate on phylogenetic diversities was assessed by fitting a spline-based smoothed regression in general additive models (GAMs), using the gam function of the R package mgcv to account for nonlinear trends. A distance-weighted auto-covariate (function autocov_dist from the R package spdep, (Bivand, Pebesma, & Gomez-Rubio, 2013) was included to account for potential spatial autocorrelation (Augustin, Mugglestone, & Buckland, 1996).
We correlated species richness and phylogenetic diversity (Appendix S4), and a principal component analysis (PCA) was conducted to visualize the spatial congruence of the phylogenetic diversity across the species groups. The PCA revealed congruence between the phylogenetic diversity of bats and grasshoppers and between that of birds and dragonflies (Appendix S5).
Finally, established protected areas (national parks, natural reserves and areas of the FFH Directive) were intersected with the 10 hotspots of cross-taxon phylogenetic diversity to calculate the extent of their potential overlap.

| Patterns of phylogenetic diversity
Regional hotspots of standardized phylogenetic diversity differed remarkably between the studied species groups (Figure 1 and Appendix S6).
For birds, the hotspots in central Bavaria were distributed between 48° and 49.5° (Figure 1a). For bats and grasshoppers, both large areas of disproportionately higher standardized phylogenetic diversity but only few hotspots clustered in southern Bavaria, in the Alps and in the Alpine foothills (Figure 1b,d). By contrast, phylogenetic diversity of birds, dragonflies and butterflies, there was lower in the Alpine region (Figure 1a,c,e), although one hotspot for butterflies was located here (Figure 1e). For dragonflies, grasshoppers and to a lower extent butterflies, regional hotspots of standardized phylogenetic diversity were in north-western Bavaria, in the area of Würzburg (Figure 1c-e).
The 10 hotspots of cross-taxon phylogenetic diversity were located in northern Bavaria, west of Würzburg; in the Alpine foothills and in eastern Bavaria (Figure 2). Established protected areas covered only approximately 9.6% of these hotspots.

F I G U R E 1 Standardized phylogenetic diversity for (a) birds, (b) bats, (c) dragonflies, (d) grasshoppers and (e)
butterflies. Colour intervals range from blue (low) to red (high). Yellow dots indicate those 10 grids hosting highest standardized phylogenetic diversity for each species group, and black dots larger Bavarian cities (Wue, Würzburg, Nu, Nuremberg, Re, Regensburg, Mu, Munich). Bold black lines inside the Bavarian border correspond to geomorphic units and thin lines to landscape units

| Influence of environmental factors
Land use significantly determined the phylogenetic diversity of bats, birds, grasshoppers and butterflies but not that of dragonflies

| D ISCUSS I ON
Our study showed that climate is a poor predictor of the standardized phylogenetic diversity of multiple species groups; instead, the standardized phylogenetic diversity of several species groups is influenced by land use. It is therefore the responsibility of land users to maintain the diversity of the tree of life while being aware that regional hotspots of standardized phylogenetic diversity differ considerably among taxonomic groups.
Our study contributes to closing this gap by considering the spatial variation of standardized phylogenetic diversity across five different species groups.
Previous studies determined that urban areas can contain high taxonomic diversity of, for example bats (Mehr et al., 2011) but the respective assemblages are composed of closely related species (Knapp, Kühn, Schweiger, & Klotz, 2008;La Sorte et al., 2018), resulting in reduced phylogenetic diversity (Riedinger, Müller, Stadler, Ulrich, & Brandl, 2013). Birds are an exception, in which a few species contribute to high levels of phylogenetic diversity in urban settings (Pfeifer et al., 2009;Sol, Bartomeus, González-Lagos, & Pavoine, 2017). The importance of such places for the protection of biodiversity therefore depends on the phylogenetic context of the resident species (Ibáñez-Álamo, Rubio, Benedetti, & Morelli, 2017).
We determined a similar pattern for agricultural landscapes, with their mixture of meadows, vineyards and orchards. These habitats hosted both common and rare bird species, such as Eurasian curlew (Numenius arquata), bee-eater (Merops apiaster), patridge (Perdix perdix), Common quail (Coturnix coturnix) and European lapwing (Vanellus vanellus), recruited from distantly related clades of the phylogenetic tree. One of these areas, where we found high standardized F I G U R E 2 Standardized phylogenetic diversity of all five studied species groups combined. Colour intervals range from blue (low) to red (high). Yellow dots indicate those 10 grids hosting highest standardized phylogenetic diversity of all five studies species groups and grey hachured areas the protected areas in Bavaria, including national parks, natural reserves and Flora-Fauna-Habitat Directive areas. Black dots mark larger Bavarian cities (Wue, Würzburg, Nu, Nuremberg, Re, Regensburg, Mu, Munich). Bold black lines inside the Bavarian border correspond to geomorphic units and thin lines to landscape units phylogenetic diversity, was around the city of Würzburg along the river Main (see also Appendix S1a). This area has a comparably high diversity of different land use types (e.g. riparian landscapes and vineyards), representing suitable habitats for dragonflies, grasshoppers and butterflies. For butterflies, the bell-shaped effect of arable land on diversity confirmed earlier findings of a strongly decreasing diversity in landscapes largely consisting of arable land (Habel, Ulrich, Biburger, Seibold, & Schmitt, 2019). There are several, intertwined reasons for this loss. For example, atmospheric nitrogen influxes from traffic, industry and agriculture, which reduces the quality of seminatural grasslands for butterflies (Wallis DeVries & van Swaay, 2006). Furthermore, land use intensification as well as abandonment of grasslands destroys formerly valuable breeding areas and habitats for butterflies (Van et al., 2019). A higher species F I G U R E 3 Significant results of generalized additive models of the standardized phylogenetic diversity of (a) birds, (b) bats, (c) grasshoppers and (d) butterflies as dependent variable and land use cover (%/100) as predictor variables. Dotted lines indicate 95% confidence intervals richness of bats, but lower standardized phylogenetic diversity was found with increasing amount of broadleaf forests (Figure 3b). This finding might be caused due to the co-occurring of closely related species from the genus Myotis, such as Natterer's bat (Myotis nattereri), greater mouse-eared bat (Myotis myotis), Bechstein's bat (Myotis bechsteinii). Those species typically occur together in large closed forests.
The Alps and the Alpine foothills host high numbers of butterfly species, but those species were most likely recruited from the same lineages. For example, of the 18 Erebia species in our dataset, 15 were restricted to the Alps (see also Pellissier et al., 2013). This results in high taxonomic diversity but low standardized phylogenetic diversity. In our study, certain land use variables influenced the standardized phylogenetic diversity of at least four of the five species groups, but most of the effects contrasted with those reported by studies at larger spatial scales (Safi, Armour-Marshall, Baillie, & Isaac, 2013;Voskamp et al., 2017). At a global scale, phylogenetic diversity has been explained by macroevolutionary processes such as biogeographic barriers as well as land use and climate (Fritz & Rahbek, 2012;Voskamp et al., 2017) or extinction and migration events (Davies & Buckley, 2011). At smaller scales, slope as topographic variable (González-Maya, Víquez-R, Arias-Alzate, Belant, & Ceballos, 2016) and elevation for birds (Dehling et al., 2014),ants (Machac, Janda, Dunn, & Sanders, 2011) and butterflies (Pellissier et al., 2013) influenced the phylogenetic diversity. A possible explanation is environmental filtering, and specifically the associated low temperatures on high ground. For invertebrates, Oliver et al. (2015) and Platts et al. (2019) studied the effects of climate change and habitat fragmentation/availability on a regional scale and stated that restoring seminatural areas and reducing fragmentation will mitigate species loss due to climate change. Mehr et al. (2011) concluded that land use is more important than climate for species richness of bats in Bavaria. Our result extends this finding to standardized phylogenetic diversity. Four climate-sensitive bat species, namely Myotis emarginatus, Pipistrellus kuhlii, Rhinolophus ferrumequinum and R. hipposideros, have been identified for Bavaria by Mehr et al. (2011). Temperature and precipitation have for example an influence on the food availability, development of juveniles or the spread of diseases (Sherwin, Montgomery, & Lundy, 2013).
However, the climate-sensitive species in our dataset are extremely rare in Bavaria. In comparison to annual mean temperature and precipitation, land use is more important for the standardized phylogenetic diversity of bats.
Most grasshoppers (Voith et al., 2016) and dragonflies (Winterholler et al., 2018) (Habel, Teucher, Ulrich, Bauer, & Rödder, 2016). Hence, climate sensitivity of butterflies might act on smaller spatial scales or might be related to extreme temperatures instead of mean temperatures than investigated in our study.
Nevertheless, recent studies also highlight the importance of land use on the preservation of butterfly diversity (Oliver et al., 2015;Thomas, 2016).

| CON CLUS IONS
We demonstrated that land use is of higher importance than macroclimate in determining local standardized phylogenetic diversity of different species groups. This finding highlights the responsibility of land users to protect the diversity of the tree of life. The hotspots of standardized phylogenetic diversity identified in our study can guide the prioritization of land areas for the conservation of the respective species groups. However, our results also demonstrate that, rather than using one species as a surrogate for others, information on different species groups should be combined to guide effective nature protection measures.

SF was funded by the Scholarship Programme of the German Federal
Environmental Foundation (DBU). All animal icons were taken from thenounproject.com. We thank everyone who collected data and the Bavarian Environment Authority for the data provision.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available on request from the Bavarian environment authority.