Introduction
Humans have fundamentally changed the face of the Earth, with negative side-effects for biodiversity across all major biomes. Tropical forests are among the biomes impacted most heavily given the large footprint of pervasive land use changes which have resulted in widespread habitat loss and fragmentation (Austin et al.
2017; Barlow et al.
2018). These land use changes have detrimentally affected the richness, abundance, and composition of many tropical taxa (Alroy
2017), leading to a pattern of extensive defaunation, with cascading effects on ecosystem functioning (Young et al.
2016).
Idiosyncratic responses of species to fragmentation are ubiquitous, rendering assemblage-level inferences regarding fragmentation effects generally difficult (Ewers and Didham
2006; Fahrig
2017). This is mainly because the treatment of species as equal entities by neglecting their unique evolutionary history, functional roles in the ecosystem, and their association with each other within the community (Pellens and Grandcolas
2016), paints an incomplete picture of the effects of habitat fragmentation. Therefore, recent studies assessing the effect of habitat fragmentation on tropical taxa have started to incorporate evolutionary information using phylogenetic diversity metrics in addition to species richness (Frishkoff et al.
2014; Santos et al.
2014; Cisneros et al.
2015,
2016; Aguirre et al.
2016; Frank et al.
2017). By doing so, these studies were able to uncover patterns previously undetected by studies with a sole focus on the taxonomic dimension of biodiversity. For example, the decrease of distantly-related plant species in a fragmented landscape (Santos et al.
2014) suggests that habitat fragmentation impoverished evolutionary history of the taxa in question. Similar trends were also found in more mobile taxa such as birds and bats, which also showed that closely-related species tend to co-occur more often than expected by chance in various types of disturbed habitats (Riedinger et al.
2013; Frishkoff et al.
2014; Frank et al.
2017). This pattern, often referred to as phylogenetic clustering, indicates a strong effect of habitat filtering (Vamosi et al.
2009).
Phylogenetic clustering in fragmented landscapes, however, is not consistently supported by empirical evidence. Several studies have documented the tendency of phylogenetic overdispersion at the edge of fragmented forests (Santos et al.
2010; Peralta et al.
2015), suggesting that different types of habitat within a fragmented landscape differentially affect the evolutionary dimension of biodiversity. Gaining better insights into the extent to which phylogenetic diversity of assemblages is eroded as a result of habitat fragmentation therefore is critical to improve our general understanding of biodiversity persistence in human-modified landscapes.
Despite their mobility, bats (Chiroptera) are among the many animal groups that are demonstrably affected by habitat loss and fragmentation (Meyer et al.
2016; Alroy
2017). Notwithstanding increased research effort devoted over recent years to better understand how bats respond to habitat fragmentation, studies at the assemblage level, typically comparing species richness, diversity, and assemblage composition between forest fragments and continuous forest, show inconsistent results and highlight the need for more research focusing on the functional and phylogenetic biodiversity dimensions (Meyer et al.
2016). Bats are a good model group to study the effect of habitat fragmentation on phylogenetic diversity given their high species richness, functional diversity, and key roles in ecosystem functioning (Kunz et al.
2011). Studies employing a phylogenetic approach to investigate bat responses towards habitat disturbance (Cisneros et al.
2015; Frank et al.
2017; Presley et al.
2018) have been made possible by the availability of phylogenetic trees of all extant bat species (Jones et al.
2002,
2005; Shi and Rabosky
2015) that can be used to calculate phylogenetic diversity metrics.
Of the few studies that have investigated bat phylogenetic diversity in fragmented landscapes, none has assessed responses across the entire gradient in habitat quality typically encountered, formed by the interiors (I) and edges (E) of continuous forest and forest fragments, as well as the intervening matrix (M), or the IEM gradient (Rocha et al.
2017a). Explicit consideration of the full IEM gradient, however, is important to better understand the extent of habitat filtering that usually is regarded as the cause of phylogenetic clustering in disturbed habitats (Riedinger et al.
2013; Frank et al.
2017; Presley et al.
2018) as species persistence in fragmented landscapes may be differentially affected by this gradient in habitat quality (Ferreira et al.
2017). The observed phylogenetic richness and structure in each habitat that comprises the IEM gradient can give an indication about the amount of evolutionary history retained by the constituent habitat elements of a fragmented landscape (Cisneros et al.
2015). Moreover, exploring which habitats share similar evolutionary history or harbour lineages that are more closely related compared to other habitats may give insights into the evolution of habitat preferences (Graham and Fine
2008).
To elucidate how habitat fragmentation affects the evolutionary dimension of bat diversity, we investigated the changes in phylogenetic alpha and beta diversity of Amazonian bat assemblages across two environmental gradients, one in habitat quality (IEM gradient) and one in habitat amount (forest size: continuous forest; fragments of 1, 10 and 100 ha), in the experimentally fragmented landscape of the Biological Dynamics of Forest Fragments Project (BDFFP), the world’s largest and longest-running experimental study of habitat fragmentation (Haddad et al.
2015), investigating a total of 12 habitat categories (the IEM gradient of the continuous forest and forest fragments). We expected that differences in phylogenetic diversity between forest fragments of different size will depend on the habitat quality therein so that the interaction between IEM gradient and forest size (habitat amount) will affect both phylogenetic alpha and beta diversity. The assemblages in the secondary forest matrix should retain the least total evolutionary history due to selection of bat lineages that are best adapted to different levels of habitat quality (Rocha et al.
2018), followed by edges, while the interiors of continuous forest should harbour the most evolutionary history due to greatest resource availability (Ries and Sisk
2004). Accordingly, phylogenetic clustering should be strongest in the matrix surrounding the smallest forest fragments as habitat filtering would result in each habitat category harbouring lineages that have already been adapted to a specific set of habitat conditions (Farneda et al.
2015; Ferreira et al.
2017). These effects would result in low phylogenetic beta diversity among similar habitat categories as phylogenetic turnover would be low in assemblages containing similar types of lineages.
Discussion
Using metrics of phylogenetic alpha and beta diversity, we showed that bat assemblages in the BDFFP landscape exhibit both a decrease of phylogenetic richness in the matrix and edges of forest fragments compared to the interior of continuous forest and a phylogenetic homogenization in these categories of the IEM gradient, as the sites in matrix and edges contained closely related bat lineages. The two environmental gradients investigated in this study explained quite well the observed variation in total phylogenetic richness (PD), but only the IEM gradient accounted for the phylogenetic information independent from species richness (SESPD). Although the mean phylogenetic distance of bat assemblages (MPD) in different fragment sizes apparently varies across the habitat categories of the IEM gradient even when quantified by its standardized effect size (SESMPD), the two environmental gradients here investigated were unable to explain the differences in phylogenetic structure among the 12 habitat categories. In agreement with this, the cluster analysis of phylogenetic beta diversity metrics did not clearly group assemblages from the same category of habitat quality and amount together as a result of high phylogenetic turnover between these assemblages.
The erosion of phylogenetic richness in edge and matrix of forest fragments, especially in the smallest fragments (1 ha), showed that habitat fragmentation at the BDFFP does not only negatively affect bat taxonomic and functional diversity in those habitat categories (Farneda et al.
2015; Rocha et al.
2017a) but also the total evolutionary history preserved. Phylogenetic beta diversity further showed marked changes in phylogenetic richness and structure of 1 ha fragment interiors along with its similarity to their edges in terms of preserved evolutionary history (Fig.
3), suggesting that patterns of phylogenetic diversity are fundamentally driven by the pervasive negative edge effects that commonly plague fragments of this size (Santos et al.
2010; Laurance et al.
2018).
In accordance with other studies comparing phylogenetic richness between various human-modified habitats of different levels of forest/vegetation complexity (Riedinger et al.
2013; Frishkoff et al.
2014; Edwards et al.
2015,
2017; Frank et al.
2017; Martins et al.
2017; Ribeiro et al.
2017; Pereira et al.
2018), the observed decrease in total evolutionary history in a structurally less complex habitat such as the secondary forest matrix is not surprising. Through the use of SES
MPD and SES
MNTD, we further showed that the edges and matrix of forest fragments also experienced phylogenetic clustering, whereby the matrix of 1 ha fragments was characterized by especially closely related lineages. Matrix sites surrounding 1 ha fragments are impoverished with respect to lineages with long evolutionary history, e.g.
Lampronycteris brachyotis and several
Micronycteris species (Fig. S3), and contain lineages that are phylogenetically clustered in the terminal branches. These absent lineages were also rarely present in disturbed habitat (Medellín et al.
2000; Frank et al.
2017), and their traits were documented to be highly associated with forest cover and tree height (Farneda et al.
2015). As these insectivorous lineages did not change their main feeding habit along their course of evolutionary history, they will be more vulnerable to habitat fragmentation compared to their frugivorous relatives which evolved independently multiple times within the Phyllostomidae in a relatively recent time (Rojas et al.
2011) and hence are better preadapted to disturbed habitat (Medellín et al.
2000; Farneda et al.
2015).
Although sites in the interior, edges, and matrix of 1 ha fragments experienced a considerable reduction in phylogenetic richness and exhibited phylogenetic clustering, we found that the change of phylogenetic diversity in lower quality habitat did not consistently depend on habitat amount. Phylogenetic beta diversity metrics further corroborate the weak effect of the two environmental gradients on between-assemblage differences in phylogenetic richness and lineage composition as there was no clear-cut pattern with regard to the relatedness and phylogenetic richness of bat assemblages in relation to IEM and forest size gradients. This is potentially due to the influence of landscape-level attributes which encompass wider environmental gradients (Rocha et al.
2017a; Tinoco et al.
2018). The low structural contrast between secondary regrowth in the matrix and the forest interiors during the sampling period could also attenuate any strong phylogenetic structure of bat assemblages along the interior-edge-matrix gradient; Patrick and Stevens (
2016) for instance found that environmental variables have a weaker effect upon phylogenetic structure when assemblages experience less harsh environmental conditions. Our measure of phylogenetic beta diversity also supports the unclear separation between different IEM gradient categories. Spatial autocorrelation in the case of total phylogenetic richness among habitat categories was unlikely an issue as the autocorrelation did not hold after we partitioned the phylogenetic beta diversity into its richness and replacement component.
The lack of a marked difference between the edge and matrix sites associated with the larger forest fragments and continuous forest possibly reflects a predominant effect of dispersal ability in the absence of contrasting environmental gradients (Moreno and Halffter
2001). Additionally, the abiotic filter that is responsible for phylogenetic clustering could possibly result from non-linear responses (Smith and Lundholm
2010; Stegen and Hurlbert
2011) that we were unable to detect using linear models. Stronger support for an effect of the IEM gradient in explaining phylogenetic structure may have been obtained if we had considered abundance-based metrics and quantitative variables such as seasonal temperature (Stevens and Gavilanez
2015) or elevation (Cisneros et al.
2014). Nonetheless, our findings still confirm our expectations that the low resource availability in the matrix selects for lineages that have traits favoured by the existing habitat filter (Farneda et al.
2015).
As our study system comprises a relatively modest gradient of forest structure (Rocha et al.
2017a), the presence of closely related bat lineages in the edge and matrix of small forest fragments could be attributed to selection towards traits determining competitiveness instead of the niche of species, particularly if the niche differences among bat lineages were not strongly related to their phylogeny (Mayfield and Levine
2010; Gerhold et al.
2015). Considering the strong sensitivity of several functional traits towards fragmentation (Farneda et al.
2015; Nuñez et al.
2019) we argue that phylogenetic clustering in habitat of low quality could lead to trait convergence to reduce competition asymmetry between closely-related species (Scheffer and van Nes
2006) and may further decrease the phylogenetic diversity of the assemblage. Thus, despite the ability of large forest fragments to retain amounts of total evolutionary history similar to the interior of continuous forest, the level of forest degradation at the edges and in the matrix still has a negative effect on the phylogenetic diversity of bats at the BDFFP.
We note that establishing the generality of our findings requires further studies as the experimentally controlled fragmented landscape of the BDFFP is a best-case scenario as levels of disturbance are reduced in relation to most real-world landscapes (Laurance et al.
2018). Additionally, forest fragments are surrounded by a soft matrix dominated by forest regrowth (Mesquita et al.
2015), and which is largely free from human disturbances which could interact additively or synergistically with fragmentation (Laurance et al.
2018). The matrix was dominated by tall secondary forest regrowth at the time of study (Rocha et al.
2017a), resulting in a rather homogenous makeup of the overall landscape which could promote random co-occurrence of bat lineages from our lineage pool (Morlon et al.
2011).
Landscape-scale studies are subject to the issue of pseudo-replication (Ramage et al.
2013) and our study at the BDFFP is no exception given that our sampling sites do not constitute true replicates of several independent fragmented landscapes. However, we believe our findings to be generalizable towards other landscapes with similar scale of habitat fragmentation. Our results regarding phylogenetic beta diversity showed that the two environmental gradients investigated were insufficient to detect meaningful trends in the pattern of phylogenetic diversity in our study system. Deciding which aspect of the environment actually matters for the taxa in question and on which ecological and evolutionary scale, however, remains a challenging part of such landscape-level studies.
Our study adds to a growing body of evidence suggesting biotic homogenization in human-modified landscapes is widespread not only with regard to the taxonomic facet of biodiversity but also its evolutionary dimension (La Sorte et al.
2018; Park and Razafindratsima
2019). Despite the aforementioned caveats, we showed fragmentation to result in the loss of species lineages and phylogenetic homogenization of assemblages in degraded edge and matrix habitats. This could further affect ecosystem stability as communities where species are evenly and distantly related to one another are more stable compared to communities where phylogenetic relationships are more clumped (Cadotte et al.
2012). The conservation of large forest fragments and the improvement of habitat quality thus should be prioritized in managing fragmented tropical forest landscapes to conserve species with longer evolutionary history.
Acknowledgements
We thank J. L. Camargo, R. Hipólito, A. J. Ferreira for logistic support and the following people for help with fieldwork: F. Farneda, J. Palmeirim, P. Bobrowiec, G. Fernandez, M. Groenenberg, R. Marciente, I. Silva, J. Treitler, J. Carvalho, S. Farias, U. Capaverde, A. Reis, L. Queiroz, J. Menezes, O. Silva, and J. Tenaçol. We also thank Dirk Metzler for his feedback on the statistical analysis. We are further grateful to the Associate Editor and two reviewers for helpful comments.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.