Bird species richness is associated with phylogenetic relatedness, plant species richness, and altitudinal range in Inner Mongolia

Abstract Bird species richness is mediated by local, regional, and historical factors, for example, competition, environmental heterogeneity, contemporary, and historical climate. Here, we related bird species richness with phylogenetic relatedness of bird assemblages, plant species richness, topography, contemporary climate, and glacial‐interglacial climate change to investigate the relative importance of these factors. This study was conducted in Inner Mongolia, an arid and semiarid region with diverse vegetation types and strong species richness gradients. The following associated variables were included as follows: phylogenetic relatedness of bird assemblages (Net Relatedness Index, NRI), plant species richness, altitudinal range, contemporary climate (mean annual temperature and precipitation, MAT and MAP), and contemporary‐Last Glacial Maximum (LGM) change in climate (change in MAT and change in MAP). Ordinary least squares linear, simultaneous autoregressive linear, and Random Forest models were used to assess the associations between these variables and bird species richness across this region. We found that bird species richness was correlated negatively with NRI and positively with plant species richness and altitudinal range, with no significant correlations with contemporary climate and glacial–interglacial climate change. The six best combinations of variables ranked by Random Forest models consistently included NRI, plant species richness, and contemporary‐LGM change in MAT. Our results suggest important roles of local ecological factors in shaping the distribution of bird species richness across this semiarid region. Our findings highlight the potential importance of these local ecological factors, for example, environmental heterogeneity, habitat filtering, and biotic interactions, in biodiversity maintenance.

In addition to those local ecological drivers, it has also been reported that regional and historical factors play important roles in shaping the geographic distribution of species richness at macroecological scales, supporting the energy availability hypothesis, the water-energy dynamics hypothesis, the historical hypothesis, the refuge hypothesis, etc. (Davies et al., 2007;Fjeldså & Lovett, 1997;Li et al., 2013;Qu et al., 2015;Rahbek & Graves, 2001). Specifically, bird species richness is strongly associated with contemporary climate variables, such as precipitation, temperature, and actual evapotranspiration (Davies et al., 2007;Li et al., 2013;Rahbek & Graves, 2001). Glacial-interglacial refuge and geological events may also affect the distribution of species richness through their impact on species speciation and extinction (Fjeldså & Lovett, 1997;Qu et al., 2015).
With an area of 120 million ha (3.3 times the size of Germany and running 3,000 km from northwest to southeast), Inner Mongolia has a wide range of climate (e.g., mean annual temperature ranging from −2 to 6°C and mean annual precipitation ranging from 40 to 450 mm, Wu, Zhang, Li, & Liang, 2015). As a result, vegetation types, plant, and bird species are very diverse in Inner Mongolia, for example, there are forest, grassland, and desert, which are home to 2,447 known vascular plant species and 467 known bird species (Xu, 2007(Xu, , 2015Zhao, 2012), which provides an ideal system to investigate the geographic distribution of bird diversity. In this study, we assessed the associations between bird species richness and phylogenetic relatedness of bird assemblages, plant species richness, topography, contemporary climate, as well as glacial-interglacial climate change in Inner Mongolia (97°12′-126°04′E, 37°24′-53°23′N).

| Geographic data
Bird distribution data at the county scale were compiled from the third and fourth volumes of Fauna of Inner Mongolia (Xu, 2007(Xu, , 2015 , 1959-2004), provincial and regional floras, as well as herbarium specimens. Plant species richness is interpreted as food resources and habitat diversity (Zhang et al., 2013). Eighty-six counties were included (see Table S1 for more information), with areas ranging from 100 km 2 to 90,000 km 2 (area was not a factor significantly affecting bird species richness, Table 1). T A B L E 1 Results of single-variable analysis by ordinary least squares (OLS) and simultaneous autoregressive (SAR) models. MAT and MAP are mean annual temperature and precipitation. Change MAT and Change MAP are the contemporary-Last Glacial Maximum change in MAT and MAP. SR plant is species richness of plants. ALT range is altitudinal range. NRI is phylogenetic relatedness of bird assemblages. Coefficients (coef) and adjusted r 2 were given. All statistically significant p-values were less than .01 and are indicated as *

| Phylogeny
A distribution of 10,000 phylogenies was downloaded from the global phylogeny of birds (Jetz et al., 2014), including all 112 resident bird species in Inner Mongolia. Five thousand pseudoposterior distributions were sampled, and the maximum clade credibility tree was constructed using mean node heights by the software TreeAnnonator v1.8.2 of the BEAST package (Drummond & Rambaut, 2007;Ricklefs & Jønsson, 2014;Si et al., 2017). We used the resulting consensus phylogeny for all subsequent phylogenetic analyses. Phylogenetic relatedness of bird assemblages was represented by the Net Relatedness Index (NRI) (Webb, Ackerly, McPeek, & Donoghue, 2002). NRI is computed as where MPD obs is the observed mean phylogenetic distance (MPD) of birds in a county, meanMPD rnd is the mean MPD of the null models (shuffle distance matrix labels 999 times), and sdMPD rnd is the standard deviation of MPD of the null models. Positive NRI means birds in a county are more closed related than expected (clustered), while negative NRI means birds in a county are more distantly related than expected (overdispersed) (Webb et al., 2002). According to the phylogenetic niche conservatism hypothesis, a clustered phylogenetic structure indicates a dominant role of environmental filtering, and an overdispersed phylogenetic structure is driven by competition or facilitation (Webb et al., 2002).

| Statistical analyses
Bird species richness, plant species richness, and county area were log transformed to obtain normal distributed residuals. All variables were standardized (mean = 0 and standard deviation = 1) to make the regression coefficients comparable. Relationships between bird species richness and each associated variable were then estimated by ordinary least squares (OLS) models. To account for spatial autocorrelation of residuals, simultaneous autoregressive (SAR) models were also used for the single-variable analyses. Because Random Forest models could effectively capture interactions (e.g., in this study correlation between MAP and change in MAP is 0.76, and correlation between MAT and change in MAT is −0.70) and nonlinear relationships, and do not require the data to follow strict assumptions, for example, homoscedasticity and normality in errors (Breiman, 2001), they were implemented for the multiple-variable analyses, aiming to find the combination of variables most associated with bird species richness. For each combination, the Random Forest models were run 1,000 times on random splits of the data (50% training data and 50% evaluation data) and averaging the Pearson correlation between the predicted and the observed values (species richness of birds). To check which variables always occurred in the best combinations, the six combinations with highest Pearson correlations were chosen. SAR models were also conducted for the six combinations because of spatial autocorrelation of residuals. AIC w and r 2 of SAR models were listed for the six combinations. AIC w is interpreted as the probability of each model being the best model and indicates the relative merit of the competing models (Wagenmakers & Farrell, 2004). OLS, SAR, and Random Forest models, as well as data transformations, were performed in R 3.3.0 (R Core Team, 2016) using vegan (Oksanen et al., 2015), spdep (Bivand et al., 2015), and randomForest (Liaw & Wiener, 2002) R packages.

| RESULTS
The single-variable ordinary least squares models and simultaneous autoregressive models showed similar patterns about the relationships between bird species richness and the associated variables (Table 1)

| DISCUSSION
It has been reported that the geographic distributions of bird species richness are shaped by divergent factors, including local, regional, and continental variables (Heikkinen et al., 2007;Qu et al., 2015;Rahbek & Graves, 2001). However, few studies have simultaneously tested

| Net relatedness index and bird species richness
A study of tropical humming birds communities found overdispersed phylogenetic structure (more distantly related than expected) in wet lowlands and clustered phylogenetic structure (more closely related than expected) at high altitude, and interpreted this as evidence of the strong influence of competition and environmental filtering in these two habitats (Graham, Parra, Rahbek, & McGuire, 2009). In this study, our analyses showed an increasing overdispersion of phylogenetic structure with higher species richness, potentially indicating a role of competition in shaping the build-up of species-rich bird assemblages NRI = −1 × MPD obs − meanMPD rnd sdMPD rnd and a role of environmental filtering on phylogenetically conserved traits in species-poor assemblages (Figure 1). Being a fundamental process in ecology, competition is an important driver of both local community assembly and macro-ecological scales species distribution (Feng et al., 2016;Fine, 2015).
Consistent with these studies, we also found positive correlations between bird species richness and plant species richness (Figure 1, Table 1). It is possible that higher plant species richness could provide more diverse habitats and food supplies for birds, thereby supporting more bird species (Zhang et al., 2013). It is also possible that there could be other factors (biotic and abiotic) affecting the diversity of birds and plants in similar ways (Kissling et al., 2007).

| Environmental heterogeneity and bird species richness
Environmental heterogeneity is also an important driver of geographic distribution of species richness for different taxa, biomes, and spatial scales (Stein et al., 2014). The increase in environmental heterogeneity could provide more niches, refuges, and opportunities for speciation (Stein et al., 2014). Being an important and easily quantified proxy of environmental heterogeneity, altitudinal range has been widely related to bird species richness at macro-ecological scales (Davies et al., 2007;Jetz & Rahbek, 2002;Jiménez-Alfaro, Chytrý, Mucina, Grace, & Rejmanek, 2016).
Consistent with these studies, we also found that bird species richness in Inner Mongolia increases with altitudinal range (Figure 1). Two counties with high bird species richness in our study, which are Alxa Left Banner  F I G U R E 1 Relationships between bird species richness (log transformed) and the three most associated variables: phylogenetic relatedness of bird assemblages (Net Relatedness Index, NRI), plant species richness (SR plant, log transformed), and altitudinal range (ALT range). Singlevariable OLS linear fits (standardized) are shown, and their r 2 is given. All statistically significant p-values were less than .01 and are indicated as *

| Climate and bird species richness
The refuge hypothesis assumes that stable glacial-interglacial climate change may both facilitate speciation and restrict extinction, thus regions with stable climate may have more species due to the accumulation of both relict and new species (Fjeldså & Lovett, 1997). Correlations between glacial climate fluctuation and bird species richness have been tested in Africa, Australia, and the New World (Fjeldså & Lovett, 1997;Hawkins, Diniz-Filho, Jaramillo, & Soeller, 2006;Hawkins, Diniz-Filho, & Soeller, 2005). Our analyses showed no significant correlations between bird species richness and historical climate variables (Table 1) Numerous previous studies have also examined the correlations between contemporary climate and bird species richness to test the energy availability hypothesis and the water-energy dynamics hypothesis (Jetz & Rahbek, 2002;Rahbek & Graves, 2001). However, our analyses did not show significant correlations between contemporary climate variables and bird species richness. Contemporary climate may affect the distribution of bird species richness through their effects on the distribution of vegetation and plant diversity in Inner Mongolia (Wu et al., 2015).

| CONCLUSIONS
Through investigating the patterns and underlying drivers of the geographic distribution of bird species richness in Inner Mongolia, we found that local ecological factors, for example, environmental heterogeneity, habitat filtering, and biotic interactions, were associated with bird species richness, while regional and historical factors, that is, contemporary and historical climate, were not. Our findings highlight the importance of these local ecological factors in biodiversity maintenance.

ACKNOWLEDGMENTS
We thank the Starting Funding for Scientific Research from Inner Mongolia University (21400-5165111) for the support to GF. J.-C.S.
was supported by the European Research Council (ERC-2012-StG-310886-HISTFUNC). We would also like to thank the editor and three reviewers for their comments and suggestions, which have substantially improved this paper.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
CL collected the data and drafted the article. GF designed the study, analyzed the data, drafted, and revised the article. XS and LM provided data and revised the article. GY, J.-C.S., and JY revised the article.