0

5

10

15

20

25

30

35

40

45

50

52

53

55

60

65

70

75

80

85

90

95

100

105

Early View (EV): 1-EV

and Tello (2014) found that a data set characterized by taxonomic, functional, phenetic, and phylogenetic diversity does not represent four independent dimensions of biodi-versity. Moreover, close correspondence among dimensions in general suggests that biodiversity may not be as dimen-sional as some current perspectives suggest (Stevens and Tello 2014). Investigating dimensionality (i.e. the number of orthogonal dimensions of biodiversity) across not just diver-sity gradients but also with respect to community assembly may provide stronger inferences regarding relationships among dimensions of biodiversity.

Spatial variation in structure of communities can be estimated using two broad sets of indices. Traditionally, diversity indices based on species identities and/or abundances (Maurer and McGill 2011) have been used to characterize spatial and temporal variation of communities. Complementary to indices of diversity, indices of commu-nity structure have also been used (Webb 2000, Vellend

Ecography 38: 001–015, 2015 doi: 10.1111/ecog.00847

© 2014 The Authors. Ecography © 2014 Nordic Society OikosSubject Editor: Douglas A. Kelt. Editor-in-Chief: Carsten Rahbek. Accepted 10 November 2014

Spatial variation in biodiversity has long fascinated ecolo-gists, evolutionary biologists, and biogeographers. Such variation typically has been characterized based on patterns of species richness (Hillebrand 2004). Nonetheless, more recently the multidimensionality of biodiversity has been better appreciated (Dornelas et al. 2012, Huang et al. 2012, Lyashevska and Farnsworth 2012). Indeed, functional, phylogenetic, phenetic (i.e. morphological), trait and genetic diversity all exhibit strong gradients that complement varia-tion in species richness (Longhi and Beisner 2010, Meynard et al. 2011, Adams and Hadley 2013, Stevens et al. 2013). However, we still lack a clear understanding of how these dimensions are related to each other through space or time (but see DeVictor et al. 2010). Recent work focusing on species geographic range maps has demonstrated fairly close correspondence among different dimensions of biodiversity in how they vary spatially in New World phyllostomid bats (Stevens et al. 2013, Stevens and Tello 2014). Thus, Stevens

Dimensionality of community structure: phylogenetic, morphological and functional perspectives along biodiversity and environmental gradients

Richard D. Stevens and Maria M. Gavilanez

R. D. Stevens (richard.stevens@ttu.edu), Dept of Natural Resources Management, Texas Tech Univ., Lubbock, TX 79409-2125, USA, and Natural Sciences Research Laboratory, Museum of Texas Tech Univ., Lubbock TX 79409-3191, USA. – M. M. Gavilanez, Escuela de Ingeniería Ambiental, Centro de Investigaciones, Univ. Estatal Amazónica, Paso lateral km 2 ½ vía a Napo, Pastaza, Puyo, Ecuador.

Biodiversity is a complex and multidimensional concept that characterizes variation of life on Earth. Nonetheless, most studies have examined only a few, if not just one, dimension in isolation. Herein we conduct analyses that explicitly incorporate correlations among multiple dimensions of biodiversity by characterizing morphological, phylogenetic, and functional structure of bat communities from Atlantic Forest of South America and examine degree of redundancy among these sets of descriptors. Second, we examine dimensionality (i.e. number of orthogonal dimensions) of community structure by quantitatively determining if these different sets of descriptors correspond to unique dimensions. We assess if dimensionality measured from empirical communities differs from that based on communities randomly assembled from a regional species pool. Finally, we examine whether different indices of community structure respond differently to environmental gradients spanning Atlantic Forest. We find that Atlantic Forest bat communities are highly variable in terms of morphological, phylogenetic, and functional structure. Different sets of community structure indices exhibited substantive correlations. Accordingly, dimensionality was lower than the set of six different descriptors or even the three different biological dimensions represented. Nonetheless, observed dimensionality was greater than that expected from a null model of assembly. Only abundance-based indices of phylogenetic structure exhibited significant environmental gradients. Temperature seasonality was the strongest predictor of phylogenetic structure, with overdispersed communi-ties characterizing more seasonal environments and underdispersed communities occurring in areas of lower variation in temperature. Dimensionality of community structure is low with phylogenetic structure exhibiting the strongest patterns, probably because phylogeny reflects many different ecological aspects of the phenotype that are not restricted to just one index of structure. Temperature seasonality is an important determinant of phylogenetic structure of bat communities in Atlantic Forest. This research helps us to better understand the factors underlying the distribution of biodiversity, which is increasingly important for endangered ecoregions such as Atlantic Forest.

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

2-EV

et al. 2011). Whereas diversity indices reflect variation in numbers of distinct entities and the equitability of their abundances within a local community, indices of structure estimate the distribution of species in a particular space defined by their ecological functions (Weiher et al. 1998, Petchey and Gaston 2002), phylogeny (Webb 2000), or phenotype (Findley 1973, Ricklefs and Travis 1980). Typically, functional, phylogenetic, or phenotypic spaces are characterized by measuring mean pairwise distances or mean nearest-neighbor distances, which estimate distribution of species within that particular space as well as their niche characteristics, and can potentially reflect resource partition-ing within communities.

Traditionally, community ecologists have used aspects of phenetic (i.e. morphological) variation to character-ize species’ niche space (Findley 1973, Ricklefs and Travis 1980, Willig and Moulton 1989, Stevens et al. 2006). More recently, phylogenetic approaches have been used to disentangle the influence of different mechanisms of assem-bly, such as competition or environmental filtering (Webb 2000). However, less appreciated is the fact that similar indices of structure can be estimated for multiple biologi-cal characteristics (Meynard et al. 2011, Swenson 2011). These indices may provide complementary information regarding differences among communities with respect to multiple dimensions of biodiversity. Estimating structure of communities based on multiple dimensions of biodiversity is uncommon but increasing in frequency (DeVictor et al. 2010, Strecker et al. 2011, Baraloto et al. 2012, Cisneros et al. 2014). Accordingly, how complementary measures of structure are related to each other, and whether struc-ture based on different dimensions of biodiversity indeed improves insights into how communities are organized, remains unclear. Moreover, the issue of how similarities and differences in structure are related to extensive environmen-tal gradients that have largely been proposed to explain the distribution of biodiversity (Currie 1991, Tello and Stevens 2010) remains unresolved (but see Estrada-Villegas et al. 2012).

Of particular interest is strength of correlation among dimensions of biodiversity and the indices of structure derived from them. Indeed, the amount of intercorrelation among parameters applied to explain biodiversity, in other words how dimensional a data set is in terms of biodiver-sity, has substantive implications. For example, if variation in all indices across a number of dimensions of biodiversity is highly correlated, this suggests low dimensionality that likely results from one or a few mechanisms that structure communities. In contrast, low correlations suggest high dimensionality that may reflect different processes control-ling spatial patterns of variation across different dimensions of biodiversity.

We focus on phyllostomid bat communities distrib-uted throughout Atlantic Forest of South America. The Phyllostomidae represents the most diverse family of bats in terms of phenotypic and functional diversity (Baker et al. 2003). They exhibit strong gradients of taxonomic diversity in the New World, particularly in species richness (Willig and Selcer 1989, Willig and Sandlin 1991, Villalobos and Arita 2010), that for the most part are associated with environmental variation (Tello and Stevens 2010).

Atlantic Forest, while being the second largest tropical forest, spans the longest latitudinal gradient of all tropical forests in the New World (Galindo-Leal and Câmara 2003, Ribeiro et al. 2009); as a result, communities are distributed along extensive environmental gradients. The bat fauna of Atlantic Forest is relatively well known and is represented by a large number of well-sampled bat communities (Table 1). As such, these represent an ideal focus for examining patterns of variation of indices of structure based on multiple dimen-sions of biodiversity. Finally Atlantic Forest is unique regard-ing human land use. The first humans colonized and began to modify this area more than 13 000 yr ago and Europeans entered South America in the Atlantic Forest approximately 500 yr ago (Dean 1995). Currently, more than 100 million inhabitants of more than 3000 cities, some of the largest on earth, were constructed on what was previously Atlantic Forest (SOS Mata Atlantica 1999, Morellato et al. 2000). To this end, understanding the ecology of organisms in this modified landscape is paramount for their conservation.

One important consequence resulting from the long latitudinal gradient that characterizes Atlantic Forest is a concomitant long gradient in environmental seasonality. This seasonality is more important to explaining patterns of diversity than the mean of environmental characteristics such as temperature, precipitation, and productivity (Martin et al. 2009). For bats, temperature seasonality in Atlantic Forest has been implicated as an important driver of both species composition (Mello 2009, Ortêncio Filho et al. 2010, Stevens and Amarilla-Stevens 2012). Furthermore, temperature represents a strong determinant of primary productivity (Churkina and Running 1998). Seasonal varia-tion in temperature is strongly related to plant phenology, in general (Tutin and Fernandez 1993) as well as in Atlantic Forest (Morellato et al. 2000, Marques et al. 2004), directly affecting resource levels for consumers such as bats. A recent study (Stevens 2013) indicated that temperature seasonality was the most important predictor of species richness of bat communities in Atlantic Forest.

Herein we address four questions. First, how are measures of morphological (phenetic), phylogenetic, and functional structure characterizing a large number of bat communi-ties from Atlantic Forest of South America related to each other? Do these three suites of descriptors represent unique dimensions from which to characterize the structure of com-munities? Is dimensionality of these three descriptors mea-sured from real communities different than that based on randomly assembled communities from the regional species pool? Finally, are different dimensions of structure related differently to variation in seasonality known to span Atlantic Forest and to be important in predicting other forms of biodiversity (Stevens 2013)?

Methods

Data on bat species composition

Data (Supplementary material Appendix 1) on phyllosto-mid community structure of 27 sites (Fig. 1) was gathered from published literature (Table 1). No sites were available between approximately 9° to 20°S latitude, and only two

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

[AQ1]

3-EV

Table 1. Results of analyses comparing empirical phylogenetic (Phylo), morphological (Morpho), and functional (Func) structure of Atlantic Forest phyllostomid communities to those constructed based on a null model. Two sets of null models were conducted; those that randomly selected species in the assembly process (Presence/absence-weighted) and those that randomly selected abundances (Abundance-weighted). A’s refer to analyses of mean pairwise distance (MPD) whereas B’s refer to those involving mean nearest neighbor distances (MNND). Lat, Long, and Richness correspond to latitudinal coordinates, longitudinal coordinates, and number of species occurring at each site, respectively. A plus () corresponds to communities that were significantly overdispersed when compared to the null model. A minus (–) corresponds to communities that were significantly underdispersed when compared to the null model. A zero (0) corresponds to communities exhibiting structure that is no different than that produced by the null model.

Presence/absence-weighted Abundance-weighted

Phylo Morpho Func Phylo Morpho Func

Site Lat Long Richness A B A B A B A B A B A B Reference

Fazenda Bituri 8.5 36.5 15 0 0 0 0 0 0 0 0 0 Menezes da Silva 2007Pernambuco Endemism Ctr. 8.6 35.1 20 0 0 0 0 0 0 0 0 0 0 0 Nobre de Oliveira 2010Serra do Caraca 20.1 43.5 9 0 0 0 0 0 0 0 0 0 0 0 Falcão et al. 2003Talhadinho 20.7 49.3 13 0 0 0 – 0 0 0 0 0 0 0 0 Breviglieri 2008Serra da Bodoquena 21.0 56.8 9 0 0 0 0 0 0 0 0 0 Camargo et al. 2009Ibitipoca 21.7 43.9 7 0 0 Nobre et al. 2009Paraiso 22.3 42.9 22 0 0 0 0 0 0 0 – 0 Esbérard 2007Caetetus 22.4 49.6 8 0 0 0 0 0 0 0 0 0 0 0 0 Pedro et al. 2001Sao Marcelo 22.4 47.0 15 0 0 0 0 0 0 0 0 Silveira et al. 2011Poco das Antas 22.5 42.3 10 0 0 0 0 0 0 0 0 0 0 0 Baptista and Mello 2001Morro de Diabo 22.5 52.3 12 0 0 0 0 – – 0 – Reis et al. 1996Tingua 22.6 43.4 19 0 0 0 0 0 0 0 0 0 0 0 0 Dias and Peracchi 2008Porto Rico 22.8 53.2 13 0 – 0 0 0 0 0 0 0 0 0 Ortêncio Filho et al. 2010Trapicheiros 22.9 43.2 13 0 0 – – – – – – – 0 – Esbérard 2003Ilha da Gipoia 23.0 44.4 19 0 0 0 – 0 – – – – 0 – – Esbérard 2009Rio das Pedras 23.0 44.1 19 0 0 0 0 0 0 0 0 0 0 0 0 Esbérard 2009RDJ Botanical Garden 23.0 43.2 22 – – 0 0 – 0 0 0 0 0 0 0 Esbérard 2003Ilha Grande 23.2 44.2 26 – – – – 0 0 0 0 0 0 0 0 Esbérard et al. 2006Arthur Thomas 23.4 51.2 12 0 0 0 0 0 0 0 0 0 0 0 0 Félix et al. 2001Yaguarete 23.8 56.0 14 0 0 0 0 0 0 – 0 0 – 0 Stevens et al. 2004Mbaracayu 24.2 55.5 9 0 0 0 0 0 0 0 0 0 0 0 0 Stevens et al. 2004Intervales 24.3 48.4 16 0 0 0 0 0 0 0 0 0 0 0 0 Passos et al. 2003Salto Morato 25.2 48.3 19 0 – 0 0 – 0 0 0 0 – 0 0 Munster 2011Iguacu 25.4 54.2 12 0 0 0 0 – 0 – 0 – 0 Sekiama et al. 2001Morro da Mina 25.4 48.8 18 – 0 – – – 0 0 – 0 0 0 0 Scultori da Silva 2009Nascentes do Garcia 27.1 49.1 15 – – 0 0 – 0 0 – 0 0 0 Althoff 2007Fredrico Westphalen 27.4 53.4 6 0 0 0 0 0 0 – 0 0 0 – – Bernardi 2011

sites were available that were north of 9°S. Nonetheless, sites are well distributed across the majority of Atlantic Forest. As Atlantic Forest roughly interdigitates into Cerrado in the north and Chaco or Pampas in the south, a number of sites are represented by relictual islands of Atlantic Forest embedded in a matrix of other habitat types. At all sites, bats were sampled using mist nets, an effective means of deter-mining the presence and relative abundance of phyllostomid bats (Kunz et al. 2009). To be included into our analyses, data were required to come from a single site such as a for-est reserve or national park. Thus, data are comparable in that they represent structure of spatially delimited local com-munities. Because studies varied in terms of effort employed to estimate community structure, we applied rarefaction (Gotelli and Colwell 2001), to exclude all sites from which species richness did not remain stable during addition of the final 50 individuals. These communities are a subset of those presented elsewhere (Stevens 2013).

Measures of structure of bat communities

We used three perspectives (phylogenetic, morphological, and functional) to characterize structure of phyllostomid

communities. Distances used represented mean pairwise- distance (MPD) and mean nearest-neighbor distance (MNND, Webb 2000) among species and individuals. Mean pairwise-distance measures the volume of all species in a community within a particular space (phenetic, phyloge-netic, or functional) whereas mean nearest-neighbor distance measures how clumped species are in a particular space. We conducted two main suites of analyses, one based on distances calculated from presence/absence of species at a site (i.e. distances among species), and the other based on distance calculated from relative abundances of constituent species (i.e. distances among individuals). While distance based on presence/absence reflects among-species varia-tion within communities, abundance-weighted distances reflect both among-species and within-species variation. Accordingly, such analyses provide comprehensive informa-tion regarding species-level structure and structure of indi-viduals within communities, respectively (Cadotte et al. 2010).

To characterize phylogenetic structure, we used the supertree provided by Bininda-Emonds et al. (2007) to calculate MPD and MNND. Measures of phylogenetic structure can be based solely on topology or on topology plus branch lengths between nodes; the latter represent measures

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

4-EV

Figure 1. Map depicting spatial location of Atlantic Forest bat communities used in analyses. Star indicates one extra-limital site character-ized by Atlantic Forest vegetation.

in millions of years of divergence from a common ances-tor. These indices assess relatedness among co-occurring taxa across the entire tree (MPD) as well as their terminal related-ness (MNND). Phylogenetic indices were calculated using functions provided in package Picante (Kembel et al. 2010) in the R software environment for statistical computing (R Development Core Team).

Phenetic distances, measuring the distribution of species in morphological space, were based on a Mahalanobis dis-tance (Marcus 1990) calculated using seven log-transformed morphological characteristics (Stevens and Willig 2000): forearm length, greatest length of skull, condylobasal length, length of maxillary toothrow, width of post-orbital constric-tion, breadth across upper molars, and breadth of braincase. Mahalanobis distances are similar to Euclidean distances but account for correlations among morphological characteristics.

Measures were based on the mean of at least 4 males and 4 females for most species. These measures estimate not only body size but also key structures associated with forag-ing and diet selection (i.e. the cranium; Stevens and Willig 2000, Stevens et al. 2006). MPD and MNND based on Mahalanobis distance characterized morphological structure of species within each community.

We used functional groups modified from Kalko (1998) as a categorical variable. Specifically, each bat was classified to one of the following groups: 1) highly cluttered glean-ing insectivore, 2) highly cluttered gleaning omnivore, 3) highly cluttered gleaning frugivore – mostly Ficus spp., 4) highly cluttered gleaning frugivore – mostly Piper spp., 5) highly cluttered gleaning frugivore – other, 6) highly cluttered gleaning nectarivore, 7) highly cluttered gleaning sanguinivore and 8) highly cluttered gleaning carnivore.

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

5-EV

dimensionality yield a very uneven distribution of eigenval-ues whereby most variation is accounted for by the first few principal components and the remainder account for little variation. In contrast, data that is highly dimensional will present a number of uncorrelated variables. Eigenvalues gen-erated from such a dataset are expected to be more uniform, each accounting for similar amounts of variation. We used Camargo’s evenness index (Camargo 1993) to characterize evenness of eigenvalues. A low evenness index indicates low dimensionality whereas a high evenness index indicates high dimensionality. Low dimensionality indicates that all indi-ces are highly correlated or redundant whereas high dimen-sionality indicates that indices exhibit low correlations or are complimentary.

Null models

We used a null model to generate expected values of com-munity structure given the observed number of species/indi-viduals present at each site. In the null model, both number of species per cell and number of cells occupied by species was maintained (e.g. both row and column sums fixed, occurrences randomly swapped; i.e. the independent-swap algorithm of Gotelli [2000]). Null model simulations were conducted in R using functions from the package Picante (Kembel et al. 2010). A measure of structure for a particular community was considered significant if it fell outside the extreme 2.5% of the distribution of random structures gen-erated from the null model.

To evaluate if dimensionality of community structure was different than null model expectations, we compared even-ness of eigenvalues from empirical data to a distribution of bootstrap evenness indices characterizing expected values of community structure based on the null model. More specifi-cally, expected values for all sites generated from null models (i.e. mean of distribution generated from the null model) based on the six measures of structure were combined into a dataset. This dataset was bootstrapped. We then conducted a PCA on the bootstrapped dataset and calculated Camargo’s evenness index on eigenvalues from this PCA. This evenness index characterized dimensionality of the dataset generated from the null model analysis. This was repeated 10 000 times to generate a distribution of bootstrapped evenness indices. If the evenness index for the empirical data fell outside the extreme 5% of null distributions of indices, we concluded that dimensionality was different from that expected given communities assembled by a null model.

Environmental gradients

Bat distributions are affected by a number of characteristics, but precipitation (Grindal et al. 1992, Erickson and West 2002) and temperature (Park et al. 2000) are important both directly and indirectly, as these characteristics affect the amount and distribution of available resources. Prior work in Atlantic Forest has suggested that seasonality is an important determinant of bat species composition. We wanted to test the hypothesis that aspects of temperature variability within years were important predictors of bat community structure. To estimate within-year temperature

A second variable categorized each species within one of three trophic levels (e.g. primarily animal eaters, generalist omni-vores, or primarily plant eaters). We also used mean relative abundance (calculated as the relative abundance of a species averaged across all sites), incidence, and mass as three other important functional traits representing the importance of species across Atlantic Forest. Functional distances among species within each community were based on Gower’s distance (Gower 1971) using the package FD (Laliberté and Legendre 2010) written for the statistical software R. This index determines a distance in functional space using a combination of categorical, ordinal, and continuous data and was used to calculate functional MPD and MNND.

Relationships among measures of structure

We characterized relationships among similar measures (i.e. functional-, morphological-, phylogenetic-MPD or MNND) of structure from different biodiversity dimen-sions based on Pearson product-moment correlation coeffi-cients (Sokal and Rohlf 1995). Significance was determined using geographically effective degrees of freedom (Dutilleul 1993) to account for inflation of type-1 error due to spatial autocorrelation. To explored unique and shared associations among the three dimensions we used variance decomposi-tion based on redundancy analyses (Borcard et al. 1992). Using this approach we decomposed variation in a particular dimension of structure into four components: 1) that unique to a particular dimension (variation unaccounted for by the other two dimensions), 2–3) that uniquely accounted for by either of the other two dimensions, and 4) that accounted for jointly by the other two dimensions. This partitioning procedure characterizes associations among different dimen-sions (i.e. redundancy) as well as magnitude of independent variation (i.e. complementarity) represented by any particu-lar dimension.

Quantifying covariation among dimensions of community structure

We examined multivariate dimensionality of community structure by calculating eigenvalues in two complementary ways to assess dimensionality. Eigenvalues were extracted from a correlation matrix characterizing relationships among the six indices of structure using a principal components analysis (PCA). We did this for each of the two data sets (presence/absence-wieghted and abundance-weighted) sepa-rately. Each eigenvalue represented an orthogonal derived axis of community structure. We used the Kaiser-Guttman stopping rule (Jackson 1993) to distinguish eigenvalues that were larger than expected and thus could be considered important. Each important eigenvalue represents a conser-vative measure of an orthogonal dimension of community structure.

Second, to go beyond the examination of number of important eigenvalues, we used the distribution of magnitudes of eigenvalues across all principal compo-nents as an additional overall measure of dimensionality. Multivariate datasets containing variables that are highly correlated exhibit low dimensionality. Such data sets of low

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

6-EV

phenetic, and functional space exhibited a mean correla-tion of r 0.74 (Fig. 2). Associations among nearest neigh-bor distances were weaker, exhibiting a mean correlation of r 0.56 (Fig. 3).

Little unique variation (i.e. unexplained by the other two dimensions) existed among dimensions of structure (Fig. 4). Functional structure exhibited the least unique variation (presence absence-weighted – 0.19, abundance weighted – 0.18) whereas phenetic structure was the most independent form of variation based on presence absence-weighted data (residual 0.41) and phylogenetic structure was the most independent form of structure based on abun-dance-weighted data (residual 0.53). However, variation was not completely redundant. For example, both morpho-logical and functional structure accounted for substantive variation in phylogenetic structure, but only functional structure accounted for unique variation. In most cases, the vast majority of variation accounted for was redundant (i.e. shared between the two other dimensions), but typically dif-ferent dimensions of structure accounted for some unique variation, often significantly.

For both presence/absence-weighted and abundance-weighted indices, two ‘important’ axes of structure were identified by PCA (Fig. 5). For presence/absence-weighted data (Fig. 5A, B), all measures of structure were positively related to the first dimension that accounted for 65% of the variation among sites regarding structure. This axis reflected overall increases in distances among species based on phy-logenetic, phenetic, and functional dimensions. The second axis based on presence/absence-weighted data accounted for an additional 22% of variation among sites regarding struc-ture. This axis was positively associated with mean pairwise distances and negatively associated with nearest neighbor distances, and effectively separates these two kinds of varia-tion (Fig. 5A)

The second PCA summarized variation in abundance-weighted indices (Fig. 5B). The first two PCA’s accounted for 45 and 36 percent of the variation among sites, respec-tively. The first axis was strongly and positively associated with mean pairwise distances based on the three dimensions of structure. This axis was weakly positively and negatively related to measures based on nearest neighbors. The second axis presented opposite patterns, being strongly positively related to nearest neighbor distances and weakly related both positively and negatively to structure based on overall dis-tances.

The empirical evenness index calculated on eigenvalues generated from the presence absence-weighted indices was 0.35 whereas that generated from the abundance data was 0.46. Bootstrapped distributions of evenness indices based on the null communities for presence/absence-weighted and abundance-weighted data, respectively, had modes of 0.23 and 0.32 respectively (Fig. 5C and D respectively). In both cases the empirical evenness index fell outside the upper tail of the distribution of indices based on bootstrapped data from the null model, indicating that empirical structure was significantly more dimensional than expected based on the null model data.

Environmental gradients were not significantly related to structure when based on presence/absence-weighted indices (Table 2). In contrast, when differences in abundance among

variability we obtained for each site an estimate of tempera-ture seasonality and annual range of temperature based on data from WorldClim (Hijmans et al. 2005) climate data layers. Temperature seasonality (bioclimatic variable 4) is defined as the standard deviation of temperature across months multiplied by 100. Annual range of temperature (bioclimatic variable 7) is defined as the differences between the coldest and hottest temperature. We then conducted two omnibus tests, for presence/absence-weighted and abun-dance-weighted structure indices separately, using a vari-ance decomposition approach based on redundancy analysis (Borcard et al. 1992). An omnibus test begins with an overall test of significance. Only upon rejection of this more gen-eral null hypothesis are lower subordinate level hypotheses tested, thus reducing type-I errors because lower order tests are protected because of the significance of the overall test (Keselman et al. 1979). As applied here we were interested in whether structure (all phylogenetic, phenetic, and functional indices included as independent variables in the same analy-sis) was significantly related to temperature seasonality and annual range of temperature. To control for possible effects of spatial autocorrelation (Peres-Neto and Legendre 2010) in the response variables we used spatial principal coordinates of neighbor matrices (PCNMs, Borcard et al. 2004) to repre-sent spatial gradients, and then conducted a variance decom-position analysis (Borcard et al. 1992) where the six measures of structure represented the dependent matrix. To determine pure environmental effects (after controlling for spatial autocorrelation) the two temperature measures were treated as independent variables and spatial PCNM’s were covari-ates. Upon a significant result, we then tested each dimen-sion (based on MPD and MNND together) of biodiversity separately. Upon a significant result for each dimension, we tested each of the two measures (MPD and MNNT) sepa-rately. Such a suite of analyses controls for type-I error while also allowing us to pinpoint exactly which aspects of struc-ture could be predicted by aspects of within year variation in temperature. These analyses also allowed determination of whether spatial autocorrelation was the driving determi-nant of significance; if the pure environmental effect was still significant then spatial autocorrelation cannot account for the environmental gradient. Finally, we were also interested in whether temperature seasonality or annual temperature range was a better predictor of structure. We conducted a similar omnibus test, this time decomposing variation into components of pure seasonality, pure range of temperature, and a correlated component.

Results

Indices of structure were quite variable across sites as well as across different dimensions of biodiversity. In many cases, structure was nonrandom (Table 1) exhibiting both over- and underdispersion in phylogenetic, phenetic, and func-tional space. Often times, structure exhibited a particular form of nonrandomness along one dimension and either was not different from random structure or exhibited a very different form on nonrandom structure on other dimensions. Nonetheless, relationships among measures of structure were moderate to strong. Mean distances in phylogenetic,

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

7-EV

Figure 2. Scatterplots indicating degree of association among indices of structure based on mean pairwise distance (MPD). The left column provides indices using presence/absence-weighted data while the right column provides indices using abundance data.

taxa were accounted for (abundance-weighted indices) aspects of within-year variation in temperature were strong predictors of structure. Pure environmental effects accounted for approximately 24% of the variation among sites in terms of structure. The environmental decomposition of structure indicated that temperature seasonality accounted for larger (also 24%) and more significant amounts of variation than

did temperature range (a non-significant predictor). Unlike morphological and functional structure, only phylogenetic structure was significantly predicted by environmental vari-ables, with seasonality of temperature being the only sig-nificant predictor. A similar pattern emerged when both variables comprising abundance-based phylogenetic struc-ture were examined separately. Both MPD and MNND

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

8-EV

Figure 3. Scatterplots indicating degree of association among indices of structure based on mean nearest neighbor distance (MNND). The left column illustrates indices using presence/absence data while the right column illustrates indices using abundance data.

exhibited significant environmental gradients and seasonal-ity of temperature was a better predictor of structure than range of temperature (Fig. 6).

Discussion

Organization of phyllostomid bat communities in Atlantic Forest of South America is spatially variable in terms of

species richness (Stevens 2013) as well as phylogenetic, phenetic, and functional dimensions of structure. Abundance-based descriptors of structure typically were more variable than those based on presence/absence-weighted data and could better resolve underlying environmental gradients of community organization. Different measures of structure, however, are substantively correlated. Nonetheless, empirical communities exhibit greater dimensionality than do those based on assembly from a null model. Phylogenetic structure

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

9-EV

Figure 4. Venn diagrams depicting results of variation partitioning analyses. Rectangles represent total variation in a particular dimension of biodiversity. Ovals represent amount of variation in a particular dimension accounted for by another predictor. Solid black, solid white, and hatched ovals correspond to functional, morphological, and phylogenetic dimensions, respectively. Areas of overlap between two dimensions refer to variation accounted for jointly by the two dimensions of biodiversity. Areas of an oval that do not overlap with the other represent unique variation accounted for by a particular dimension. Notation below figures indicates significance of particular partitions. For example, in the upper left panel, the partition of unique morphological (Morpho) variation after accounting for what is shared with functional variation ([Morpho l Fun]) is nonsignificant whereas the partition of the unique functional (Fun) variation after accounting for what is shared with morphological variation ([Fun l Morpho]) is significant.

appears to exhibit the most unique variation, especially when measured using abundance-based descriptors. Moreover, this dimension is significantly related to temperature seasonality whereby sites with the greatest seasonality present commu-nities that are the most phylogenetically clustered and those that experience the least seasonality exhibit more phyloge-netic overdispersion.

Actual dimensionality of measures of structure was less than the 6 measures used and even less than the three dimensions of phylogenetic, morphological, and functional structure. Indeed, strong positive correlations exist among all measures of structure, which is not uncommon (DeVictor et al. 2010, Davies and Cadotte 2011). Such substantive correlation suggests that the effective dimensionality of

community structure is comparatively low and can be decomposed into just a few dimensions. Further highlight-ing the lack of dimensionality of phylogenetic, morphologi-cal, and functional structure for bat communities in Atlantic Forest is the observation that orthogonal dimensions of variation identified by PCA are organized around statisti-cal but not biological perspectives. Indeed, mean pairwise differences and mean nearest neighbor distances tended to form separate groups of variation, being negatively related for presence/absence-weighted measures and fairly orthogo-nal for abundance-based measures. These results suggest that these two types of measures (i.e. mean pairwise distances or mean nearest neighbor distances) represent distinct ways in which biodiversity changes, either filling overall spaces

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

10-EV

Table 2. Omnibus spatial analysis and decomposition of environmental gradients of structure of Atlantic Forest bat communities. In such an analysis once a result is non-significant, no further subordinate analyses are conducted. The ‘All-presence/absence’ analysis indicated that together spatial and environmental variable did not account for significant variation in indices of biodiversity. In contrast, for the ‘All-abundance’ analysis they did. To determine which dimensions of biodiversity were related to these variables, analyses were then conducted for three dimensions separately. When only phylogenetic structure exhibited significant relationships, it two indices (MDP and MNND) were then analyzed in the same fashion. Spatial analyses use spatial principal coordinates of neighbor matrices (PCNM’s) to maintain type I error at a 0.05 while testing the hypothesis that the two seasonality variables together account for significant variation in a particular structure index. Upon significance of the two seasonality variables based on the spatial analysis, relative contributions of temperature seasonality and annual temperature range to a particular component of structure were determined (Environmental decomposition). Tests that are significant or approach significance (denoted by ampersand) are in bold.

Spatial analyses Environmental decomposition

VariableCombined

statistic pSeasonality

statistic pSpatial statistic p

Combined statistic p

T-seasonality statistic p

T-range statistic p

All-presence/absence

0.01 0.525 0.00 0.413 0.00 0.739

All-abundance 0.60 0.001 0.24 0.001 0.34 0.005 0.26 0.001 0.24 0.003 0.08 0.073@

Functional structure

0.31 0.080@ 0.05 0.211 0.22 0.128

Morphologi-cal structure

0.03 0.481 0.00 0.466 0.05 0.408

Phylogenetic structure

0.62 0.002 0.25 0.003 0.35 0.005 0.28 0.001 0.23 0.003 0.05 0.075@

Phyloge-netic MPD

0.42 0.067@ 0.19 0.057@ 0.19 0.195 0.23 0.008 0.11 0.050 0.00 0.735

Phyloge-netic MNND

0.79 0.001 0.29 0.003 0.48 0.005 0.31 0.006 0.32 0.003 0.12 0.034

Figure 5. Analysis of dimensionality of biodiversity of Atlantic Forest phyllostomid communities. Upper panels describe principal compo-nents analyses on 3 dimensions of biodiversity based on 6 different indices. (A) Left panels correspond to the analysis based on presence/absence-weighted data whereas the right panels (B) corresponds to the analysis based on abundance data. Above histograms display scree plots describing the magnitude of each eigenvalue generated from a particular analysis. Black bars represent the empirical data whereas grey bars represent data generated from the null model. Horizontal black lines correspond to an eigenvalue of 1; principal components with eigenvalues greater than this threshold account for more variation than any of the original variables on average and represent an important dimension of biodiversity. Insets describe correlations of original variables with the first two principal components of each analysis. PMPD, MMPD and FMPD are phylogenetic, morphological, and functional mean pairwise differences, respectively. PMNND, MMDDN and FMNND are phylogenetic morphological and functional mean nearest neighbor distances respectively. In figures below, dimensionality of biodiversity is estimated. Histograms represent the distribution of evenness values based on bootstrapped null-communities whereas arrows represent evenness indices calculated for the real communities.

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

11-EV

Figure 6. Gradients of phylogenetic diversity related to temperature seasonality.

(i.e. phenetic, phylogenetic) as reflected by mean pairwise distances, or increasing similarity of closely related species as reflected by nearest neighbor distances.

Despite low dimensionality, local communities exhib-ited greater dimensionality than those generated from a null model. Aspects of how different indices are calculated can affect magnitude of values, patterns of variation, and even degree of correlation among dimensions. For example, differences in tree topology can substantively affect the strength of the relationship between phylogenetic structure and species richness (Rodriguez et al. 2005). Nonetheless, since the same data and indices were used to calculate mea-sures of biodiversity across multiple dimensions, this cannot account for differences between null models and empirical dimensionality. Such differences likely are more related to assembly of and coexistence within local communities. For example, substantive beta diversity characterizes regional variation among communities of most organisms (Koleff et al. 2003), including bats (Stevens and Willig 2002, Rodriguez and Arita 2004, Larrea-Alcázar et al. 2011). When differences in composition also reflect different morpholo-gies, phylogenetic relatedness, or ecological functions, then such differences will affect correspondence of these three dif-ferent types of biodiversity measures. Moreover, when spe-cies are sorted during the assembly process it is likely that different dimensions are important in the selection process depending on the community. This may explain cases of significant overdispersion and clustering that were incon-gruent across dimensions of biodiversity (Table 1). Such

incongruence decouples dimensions by decreasing corre-lated variation and thereby increasing dimensionality. A lack of correspondence in nonrandom structure across multiple dimensions likely is enhanced when different processes struc-ture different communities. Such a pattern suggests context-dependent dynamics of the assembly process whereby the importance of a particular facet of biological variation in determining coexistence of species is not uniform across communities. Despite low dimensionality, empirical com-munities may still be more dimensional than those randomly assembled from the regional species pool because drivers of beta diversity serve to decouple variation across dimensions during the assembly process.

Phylogenetic structure exhibited the greatest amount of unique variation. At least two phenomena could contribute to this. The first is that phylogenetic information is a very powerful surrogate for many forms of ecological variation (Cadotte et al. 2010) and when determined from branch lengths can even capture variation in genetic diversity above the species level (Faith 1992, Bininda-Emonds et al. 2007). We used a relatively small number of descriptors to charac-terize both phenetic and functional structure and any limited suite has the potential to incompletely characterize a larger set. Thus, one reason why phylogenetic structure exhibits more unique variation and exhibits stronger environmental gradients may be that it captures more ecological variation in terms of phenetic and functional structure than do these very indices, a phenomenon that has also been reported in studies of diversity-ecosystem function relationships (Cadotte et al. 2009). Secondly, it is likely that functional and phe-netic dimensions of biodiversity together remain incomplete descriptors of biodiversity as a whole. Nonetheless, because of niche conservatism (Wiens et al. 2010), phylogeny likely reflects not only phenetic and functional variation among species, but also other forms of variation important in the assembly process, and thus exhibits more unique variation and stronger environmental gradients. Indeed, as our results and others (Cadotte et al. 2008, Huang et al. 2012) indicate, when only one type of index is to be chosen, phylogenetic descriptors should be preferred, as they incorporate much vari-ation not necessarily explained by other indices of structure.

Abundance-weighted phylogenetic structure was the only metric to exhibit significant environmental gradients. Abundance-weighted phylogenetic structure characterizes the phylogenetic tree of individuals within a community. Thus, how individuals are allocated to a set of species with a particular phylogenetic structure defines the difference between presence/absence-weighted and abundance-weighted measures. Phylogenetic structure of communities that incor-porates abundance reflects not only the distribution of traits but also the distribution of abundances of species within communities as well as the differential effects of environmen-tal gradients, in this case temperature seasonality, on individ-uals. Indeed, abundance-weighted phylogenetic structure of communities offers additional explanatory power regarding how biological diversity is distributed along environmen-tal gradients. In many communities, rarity of most species makes estimation of presence/absence more difficult. One effect of this is that comparisons of incidence suffers from substantial stochasticity in detection of rare taxa. The shapes of abundance distributions, however, are much more easily

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

12-EV

principal ecological and evolutionary strategies (i.e. special-ists on Solanum, Cecropia, Piper, or Ficus – Andrade et al. 2013) characterizing the highly diverse Phyllostomidae. More simplified phylogenetic trees characterizing commu-nities in more seasonal areas are represented by those spe-cies with high global abundance and large geographic ranges (Arita 1993). Under this scenario, abundant species that are evenly distributed across the phyllostomid tree give rise to clumping based on abundance-weighted phylogenetic indi-ces. Moreover, species that are added to the tree as species richness increases are those from the same genera as pre-existing species but are not as geographically widespread or abundant; such additions serve to increase distances in phylogenetic space.

Patterns of phylogenetic overdispersion and clustering typically are interpreted in a very different way than that described above. Typically, overdispersion is interpreted to be the result of strong interspecific interactions that lead to competitive exclusion (Webb 2000). Theoretically, limiting similarity sets a minimum on how similar species can be and still coexist in systems of strong competition, and Violle et al. (2011) coined the term ‘phylogenetic limiting similar-ity’ to refer to the fact that closely related species often share similar niches and are therefore highly competitive. Inferior competitors that are too similar are locally extirpated and this causes phylogenetic overdispersion. In this case, over-dispersion is represented by few species that are far apart in phylogenetic space. However, this is not the only phyloge-netic pattern that can result from competitive interactions; the exact opposite can occur as well (Helmus et al. 2007). For example, just as a lack of niche differences can lead to competitive exclusion and thus phylogenetic overdispersion, a lack of differences in species competitive abilities can also lead to competitive exclusion but resulting in phylogenetic clustering (Mayfield and Levine 2010). In addition, envi-ronmental filtering can give rise to phylogenetic structure as well (Webb 2000). Typically, phylogenetic clustering is interpreted as a product of environmental filtering to all but a particular group of closely related species that can tolerate a given set of environmental conditions. Nonetheless, our results for bats in Atlantic Forest suggest that environmental filtering, in this case related to temperature seasonality, can also give rise to a pattern of clumping of individuals across phylogenetically distant species. Use of abundance-weighted measures of structure is not common in the literature but is growing. Little theoretical basis exists to predict variation in these metrics. Theoretical expectations of how phylogenetic descriptors respond to structuring mechanisms likely are not the same for abundance-weighted and presence/absence-weighted indices. Indeed, such plural phylogenetic patterns resulting from similar mechanisms acting on different types of indices admonish blind interpretation of phylogenetic patterns.

In conclusion, dimensionality of biodiversity of phyllos-tomid bats in Atlantic Forests of South America is low given the three dimensions and six indices used to characterize it. However, despite this low dimensionality, real communities present higher dimensionality than random communities. We suggest that this may be related to differences in distri-butional responses of bat species to temperature seasonality. Community structure increases along seasonality gradients

estimated (Chao et al. 2010), and are expected to exhibit less stochastic noise, thus providing more robust resolution of environmental gradients. Accordingly, abundance-weighted phylogenetic structure was positively associated with tem-perature seasonality indicating more overdispersed commu-nities in less seasonal environments. Abundance-weighted phylogenetic MPD or MNND (dependent variable) can be regressed onto corresponding presence/absence-weighted indices (independent variable) to partition variation into intra- and interspecific components. Phylogenetic MNND based on presence/absence accounted for 42% of variation in abundance-weighted MNND. Also, phylogenetic MPD based on presence/absence accounted for approximately 9% of the variation in abundance-weighted MPD. Thus, much of the variation (i.e. more than 50% in both cases) underly-ing the relationship between phylogenetic structure and tem-perature seasonality has to do with intraspecific clustering. Such a gradient, whereby intraspecific phylogenetic clump-ing increases with increased seasonality, could be expected based on prior theory. For example, in high stress areas there are fewer resources, which may lead to more temporal varia-tion in population size and thus a higher probability of local extinction of rare taxa (Rosenzweig 1971, Rosenzweig and Abramsky 1993). Such a process would increase evenness in harsh environments. Conversely, in more favorable sites, increases in competitively dominant individuals will cause greater differences between rare and abundant taxa (Drobner et al. 1998), thereby increasing indices of dominance in more favorable sites. Indeed, temperature seasonality affects species evenness more than species richness in terms of the seasonal structure of bat communities in Paraguayan Atlantic Forest (Stevens and Amarilla-Stevens 2012).

Seasonality is an important determinant of structure of Atlantic Forest bat communities (Mello 2009, Ortêncio Filho et al. 2010, Stevens and Amarilla-Stevens 2012). More seasonal areas present fewer species of phyllostomid bats (Stevens 2011). The observed pattern of change in phylogenetic structure illuminates how such reductions in species richness occur. Sites characterized by low abundance-weighted phylogenetic distance (e.g. Yaguarete, Iguacu, Fredrico Westphalen) have fewer species of widespread (e.g. throughout the Atlantic Forest) generalist taxa. For example, all sites, even those with the lowest species richness, possess individuals from taxa such as Artibeus lituratus and Sturnira lilium. Moreover, Carollia perspicillata, Desmodus rotundus, and Platyrrhinus lineatus occur across more than 75% of sites sampled. These taxa all have large geographic range size and commonly high local abundance (Arita 1993). In addition, Artibeus fimbriatus and Pygoderma bilabiatum are almost entirely restricted to Atlantic Forest but are very common within this ecoregion; each occurs over ∼78% of sites. These common/abundant bat species are fairly evenly distributed across the phyllostomid phylogeny and form the core of Atlantic Forest bat communities. Nonetheless, many com-munities contain many more than these six species, espe-cially in areas of low temperature seasonality. Seasonality thus may act as a filter that prunes back the phylogenetic tree of individuals to a few main branches represented primarily by the core species that are fairly evenly spread across the phyllostomid tree. Branches that remain in the most seasonal environments are represented by species that represent the

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

13-EV

Cisneros, L. M. et al. 2014. Multiple dimensions of biodiversity along an extensive tropical elevational gradient. – J. Anim. Ecol. 83: 1124–1136.

Currie, D. J. 1991. Energy and large scale patterns of animal-species and plant species richness. – Am. Nat. 137: 27–49.

Davies, T. J. and Cadotte, M. W. 2011. Quantifying biodiversity: does it matter what we measure? – In: Zachos, F. E. and Habel, J. C. (eds), Biodiversity hotspots. Springer, pp. 43–59.

Dean, W. B. 1995. With broadax and firebrand: the destruction of the Brazilian Atlantic Forest. – Univ. of California Press.

DeVictor, V. et al. 2010. Spatial mismatch and congruence between taxonomic, phylogenetic and functional diversity: the need for integrative conservation strategies in a changing world. – Ecol. Lett. 13: 1030–1040.

Dornelas, M. et al. 2012. Quantifying temporal change in biodi-versity: challenges and opportunities. – Proc. R. Soc. B 279: 4997–5003.

Drobner, U. et al. 1998. The relation between community biomass and evenness: what does community theory predict, and can these predictions be tested? – Oikos 82: 295–302.

Dutilleul, P. 1993. Modifying the t-test for assessing the correlation between 2 spatial processes. – Biometrics 49: 305–314.

Erickson, J. L. and West, S. D. 2002 The influence of regional climate and nightly weather conditions on activity patterns of insectivorous bats. – Acta Chiropt. 4: 17–24.

Estrada-Villegas, S. et al. 2012. Climate, habitat, and species interactions at different scales determine the structure of a Neotropical bat community. – Ecology 93: 1183–1193.

Faith, D. P. 1992. Conservation evaluation and phylogenetic diversity. – Biol. Conserv. 61: 1–10.

Findley, J. S. 1973. Phenetic packing as a measure of faunal diver-sity. – Am. Nat. 107: 580–584.

Galindo-Leal, C. and Câmara, I. G. 2003. The Atlantic Forest of South America: biodiversity Status. Threats and outlook. – Island Press.

Gotelli, N. J. 2000. Null model analysis of species co-occurrence patterns. – Ecology 81: 2606–2621.

Gotelli, N. J. and Colwell, R. K. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. – Ecol. Lett. 4: 379–391.

Gower, J. C. 1971. A general coefficient of similarity and some of its properties. – Biometrics 27: 857–871.

Grindal, S. D. et al. 1992. The influence of precipitation on repro-duction by Myotis bats in British Colombia. – Am. Midl. Nat. 128: 339–344.

Helmus, M. R. et al. 2007. Separating the determinants of phylo-genetic community structure. – Ecol. Lett. 10: 917–925.

Hijmans, R. J. et al. 2005. Very high resolution interpolated climate surfaces for global land areas. – Int. J. Clim. 25: 1965–1978.

Hillebrand, H. 2004. On the generality of the latitudinal diversity gradient. – Am. Nat. 163: 192–211.

Huang, S. et al. 2012. Traits, trees and taxa: global dimensions of biodiversity in mammals. – Proc. R. Soc. B 279: 4997–5003.

Jackson, D. A. 1993. Stopping rules in principal components analysis: a comparison of heuristical and statistical approaches. – Ecology 74: 2204–2214.

Kalko, E. K. V. 1998. Organisation and diversity of tropical bat communities through space and time. – Zoology 101: 281–297.

Kembel, S. W. et al. 2010. Picante: R tools for integrating phylog-enies and ecology. – Bioinformatics 26: 1463–1464.

Keselman, H. J. et al. 1979. Protecting the overall rate of type-1 errors for pairwise comparisons with an omnibus test statistic. – Psychol. Bull. 88: 884–888.

Koleff, P. et al. 2003. Are there latitudinal gradients in species turnover? – Global Ecol. Biogeogr. 12: 483–498.

by adding more closely related individuals (and species as well) to the preexisting range of variation, thereby increas-ing phylogenetic clustering of individuals. Seasonality is an important predictor of distribution of phyllostomids in Atlantic Forest as well as the entire New World (Stevens 2011). Future work should examine the generality of this finding across other Neotropical ecoregions as well as other parts of the World. Understanding the predictors of patterns of species coexistence and community assembly across the multiple facets of biodiversity constitutes a crucial step to implementing conservation strategies, particularly in highly threatened ecosystems such as Atlantic Forest. Furthermore, insights such as those presented here regarding how seasonal-ity can be used to predict community structure and diversity represent valuable information for developing strategies to combat climate change, which has emerged as one of the major threats to biodiversity we are currently experiencing (Whittaker et al. 2005).

Acknowledgements – RDS was supported by the NSF (DEB-1020890) while producing this manuscript. Sergio Estrada-Villegas provided exceptionally helpful review. This is manuscript number T-9-1246, College of Agricultural Sciences and Natural Resources, Texas Tech Univ.

References

Adams, R. I. and Hadley, E. A. 2013. Genetic diversity within vertebrate species is greater at lower latitudes. – Evol. Ecol. 27: 133–143.

Andrade, T. Y. et al. 2013. Hierarchical fruit selection by Neotropical leaf-nosed bats (Chiroptera: Phyllostomidae). – J. Mammal. 94: 1094–1101.

Arita, H. T. 1993. Rarity in Neotropical bats: correlations with phylogeny, diet, and body mass. – Ecol. Appl. 3: 506–517.

Baker, R. J. et al. 2003. Diversification among New World leaf-nosed bats: an evolutionary hypothesis and classification inferred from digenomic congruence of DNA sequence. – Occas. Pap. Mus. Texas Tech Univ. 230: 1–32.

Baraloto, C. et al. 2012. Contrasting taxonomic and functional responses of a tropical tree community to selective logging. – J. Appl. Ecol. 49: 861–870.

Bininda-Emonds, O. R. P. et al. 2007. The delayed rise of present-day mammals. – Nature 446: 507–512.

Borcard, D. et al. 1992. Partialling out the spatial component of ecological variation. – Ecology 73: 1045–1055.

Borcard, D. et al. 2004. Dissecting the spatial structure of ecological data at multiple scales. – Ecology 85: 1826–1832.

Cadotte, M. W. et al. 2008. Evolutionary history and the effect of biodiversity on plant productivity. – Proc. Natl Acad. Sci. USA 105: 17012–17017.

Cadotte, M. W. et al. 2009. Using phylogenetic, functional and trait diversity to understand patterns of plant community productivity. – PLoS One 4: e5695.

Cadotte, M. W. et al. 2010. Phylogenetic diversity metrics for ecological communities: integrating species richness, abun-dance and evolutionary history. – Ecol. Lett. 13: 96–105.

Camargo, J. A. 1993. Must dominance increase with the number of subordinate species in competitive interactions. – J. Theor. Biol. 161: 537–542.

Chao, A. et al. 2010. Phylogenetic diversity based on Hill numbers. – Phil. Trans. R. Soc. B 365: 3599–3609.

Churkina G. and Running, S. W. 1998. Contrasting climatic con-trols on the estimated productivity of global terrestrial biomes. – Ecosystems 1: 206–215.

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

14-EV

Rosenzweig, M. L. 1971. Paradox of enrichment: destabilization of exploitation ecosystems in ecological time. – Science 171: 385–387.

Rosenzweig, M. L. and Abramsky, Z. 1993. How are diversity and productivity related? – In: Ricklefs, R. E. and Schluter, D. (eds), Species diversity in ecological communities. Univ. of Chicago Press, pp. 52–65.

Sokal, R. R. and Rohlf, F.J. 1995. Biometry, 3rd ed. – W. H. Free-man.

SOS Mata Atlantica 1999. – Boletim Informativo, Junho/Julho, 1(8).

Stevens, R. D. 2011. Relative effects of time for speciation and tropical niche conservatism on the latitudinal diversity gradient of phyllostomid bats. – Proc. R. Soc. B 278: 2528–2536.

Stevens, R. D. 2013. Gradients of bat diversity in Atlantic Forest of South America: effects of environmental seasonality, sampling effort and spatial autocorrelation. – Biotropica 45: 764–770.

Stevens, R. D. and Willig, M. R. 2000. Community structure, abundance, and morphology. – Oikos 88: 48–56.

Stevens, R. D. and Willig, M. R. 2002. Geographical ecology at the community level: perspectives on the diversity of New World bats. – Ecology 83: 545–560.

Stevens, R. D. and Amarilla-Stevens, H. N. 2012. Seasonal envi-ronments, episodic density compensation and dynamics of structure of chiropteran frugivore guilds in Paraguayan Atlantic Forest. – Biodivers. Conserv. 21: 267–279.

Stevens, R. D. and Tello, J. S. 2014. On the measurement of dimensionality of biodiversity. – Global Ecol. Biogeogr. 23: 1115–1125.

Stevens, R. D. et al. 2006. Latitudinal gradients in the phenetic diversity of New World bat communities. – Oikos 112: 41–50.

Stevens, R. D. et al. 2013. Stronger tests of mechanisms underlying geographic gradients of biodiversity: insights from the dimensionality of biodiversity. – PLoS One 8: e56853.

Strecker, A. L. et al. 2011. Defining conservation priorities for freshwater fishes according to taxonomic, functional, and phylogenetic diversity. – Ecol. Appl. 21: 3002–3013.

Swenson, N. G. 2011. The role of evolutionary processes in producing biodiversity patterns, and the interrelationships between taxonomic, functional and phyllogenetic diversity. – Am. J. Bot. 98: 472–480.

Tello, J. S. and Stevens, R. D. 2010. Multiple environmental determinants of species richness: energy, heterogeneity, and seasonality. – Ecography 33: 796–808.

Tutin, C. E. G. and Fernandez, M. 1993. Relationships between minimum temperature and fruit production in some tropical forest trees in Gabon. – J. Trop. Ecol. 9: 241–248.

Vellend, M. et al. 2011. Measuring phylogenetic biodiversity. – In: Magurran, A. E. and McGill, B. J. (eds), Biological diversity: frontiers in measurement and assessment. Oxford Univ Press, pp. 194–207.

Villalobos, F. and Arita, H. T. 2010. The diversity field of New World leaf-nosed bats (Phyllostomidae). – Global Ecol. Biogeogr. 19: 200–211.

Violle, C. et al. 2011. Phylogenetic limiting similarity and competitive exclusion. – Ecol. Lett. 14: 782–787.

Webb, C.O. 2000. Exploring the phylogenetic structure of eco-logical communities: an example for rainforest trees. – Am. Nat. 156: 145–155.

Weiher, E. et al. 1998. Community assembly rules, morphological dispersion, and the coexistence of plant species. – Oikos 81: 309–322.

Whittaker, R. J. et al. 2005. Conservation biogeography: assessment and prospect. – Divers. Distrib. 11: 3–23.

Wiens, J. J. et al. 2010. Niche conservatism as an emerging prin-ciple in ecology and conservation biology. – Ecol. Lett. 13: 1310–1324.

Kunz, T. H. et al. 2009. Methods for capturing and handling bats. – In: Kunz, T. H. and Parsons, S. (eds), Ecological and behavioral methods for the study of bats. Univ. of Chicago Press, pp. 3–35.

Laliberté, E. and Legendre, P. 2010. A distance-based frame-work for measuring functional diversity from multiple traits. – Ecology 91: 299–305.

Larrea-Alcázar, D. M. et al. 2011. Spatial patterns of biological diversity in a Neotropical lowland savanna of northeastern Bolivia. – Biodivers. Conserv. 20: 1167–1182.

Longhi, M. L. and Beisner, B. E. 2010. Patterns in taxonomic and functional diversity of lake phytoplankton. – Freshw. Biol. 55: 1349–1366.

Lyashevska, O. and Farnsworth, K. D. 2012. How many dimensions of biodiversity do we need? – Ecol. Indicators 18: 485–492.

Marcus, L. 1990. Traditional morphometrics. – In: Rohlf, F. J and Bookstein, F. L. (eds), Proceedings of the Michigan morpho-metrics workshop. Univ. of Michigan Museum of Zoology, Ann Arbor, MI, pp. 77–122.

Marques, M. C. M. et al. 2004. Phenological patterns among plant life-forms in a subtropical forest in southern Brazil. – Plant Ecol. 173: 203–213.

Martin, P. R. et al. 2009. Latitudinal variation in the asynchrony of seasons: implications for higher rates of population differ-entiation and speciation in the tropics. – Ideas Ecol. Evol. Biol. 2: 9–17.

Maurer, B. A. and McGill, B. J. 2011. Measurement of species diversity. – In: Magurran, A. E. and McGill, B. J. (eds), Bio-logical diversity: frontiers in measurement and assessment. Oxford Univ. Press, pp. 55–65.

Mayfield, M. M. and Levine, J. M. 2010. Opposing effects of competitive exclusion on the phylogenetic structure of communities. – Ecol. Lett. 13: 1085–1093.

Mello, M. A. R. 2009. Temporal variation in the organization of a Neotropical assemblage of leaf-nosed bats (Chiroptera:

Phyllostomidae). – Acta Oecol. 35: 280–286.Meynard, C. N. et al. 2011. Beyond taxonomic diversity patterns:

how do a, b and g components of bird functional and phylo-genetic diversity respond to environmental gradients across France? – Global Ecol. Biogeogr. 20: 893–903.

Morellato, L. P. C. et al. 2000. Phenology of Atlantic Forest trees: a comparative study. – Biotropica 32: 811–823.

Ortêncio Filho, H. et al. 2010. Time and seasonal patterns of activ-ity of phyllostomid bats in fragments of a stational semide-cidual forest from the upper Parana River. – Southern Brazil. Braz. J. Biol. 70: 937–945.

Park, K. J. et al. 2000. Torpor, arousal and activity of hibernating greater horseshoe bats Rhinolophus ferrumequinum. – Funct. Ecol. 14: 580–588.

Peres-Neto, P. R. and Legendre, P. 2010. Estimating and controlling for spatial structure in the study of ecological communities. – Global Ecol. Biogeogr. 19: 174–184.

Petchey, O. L. and Gaston, K. J. 2002. Functional diversity (FD), species richness and community composition. – Ecol. Lett. 5: 402–411.

Ribeiro, M. C. et al. 2009. Brazilian Atlantic Forest: how much is left and how is the remaining forest distributed? Implications for conservation. – Biol. Conserv. 142: 1141–1153.

Ricklefs, R. E. and Travis, J. 1980. A morphological approach to the study of avian community organization. – Auk 97: 321–328.

Rodriguez, A. S. L. et al. 2005 Integrating phylogenetic diversity in the selection of priority areas for conservation: does it make a difference. – In: Purvis, A. et al. (eds), Phylogeny and con-servation. Cambridge Univ. Press, pp. 101–140.

Rodriguez, P. and Arita, H. T. 2004. Beta diversity and latitude in North American mammals: testing the hypothesis of covaria-tion. – Ecography 27: 547–556.

0

5

10

15

20

25

30

35

40

45

50

55

60

61

65

70

75

80

85

90

95

100

105

110

115

121

15-EV

Willig, M. R. and Sandlin, E. A. 1991. Gradients of species density and species turnover in New World bats: a comparison of quadrat and band methodologies. – In: Mares, M. A. and Schmidly, D. J. (eds), Latin American mammalogy: history, biodiversity, and conservation. Univ. of Oklahoma Press, pp. 81–96.

Willig, M. R. and Moulton, M. A. 1989. The role of stochastic and deterministic processes in the structure of Neotropical bat communities. – J. Mammal. 70: 323–329.

Willig, M. R. and Selcer, K. W. 1989. Bat species density gradients in the New World: a statistical assessment. – J. Biogeogr. 16: 189–195.

Supplementary material (Appendix ECOG-00847 at www.ecography.org/readers/appendix). Appendix 1.