The relationship between niche breadth and range size of beech (Fagus) species worldwide

This work explores whether the commonly observed positive range size–niche breadth relationship exists for Fagus, one of the most dominant and widespread broad‐leaved deciduous tree genera in temperate forests of the Northern Hemisphere. Additionally, we ask whether the 10 extant Fagus species’ niche breadths and climatic tolerances are under phylogenetic control.


| INTRODUC TI ON
Geographical range size is generally defined as the 2-dimensional extent of the spatial distribution of a species based on latitudinal and longitudinal extents (Gaston, 1991), whereas a species' realized niche is widely understood as the n-dimensional hypervolume (Hutchinson, 1957) defined by the multi-dimensional range of abiotic and biotic conditions under which it can sustain natural populations (Blonder, 2018). The quantification and comparison of niche hypervolumes (hereafter niche breadth) have long been of interest to ecologists (e.g. Blonder et al., 2014;Fridley et al., 2007;Hutchinson, 1957;Junker et al., 2016;Kambach et al., 2019;Sexton et al., 2017;Smith, 1982). Properties like species' niche breadths and niche overlaps are proposed to affect co-occurrence patterns at the community level (Bar-Massada, 2015), and realized niche breadths are often used to define the degree of species specialization (Devictor et al., 2010). Since specialists (species with a narrow niche) are thought to be more vulnerable to current and future climate change than generalists (Devictor et al., 2010; but see Colles et al., 2009), determining species' niche breadth can help identify priority species for conservation actions (Boulangeat et al., 2012).
A positive relationship between range size and niche breadth has been proposed by the "niche breadth hypothesis" (Sheth et al., 2020). A wide niche allows a species to cope with a wide range of environmental conditions (Brown, 1984). Mechanistically, a wide niche breadth can be brought about either by generalistic genotypes or by many different particularly adapted genotypes that replace each other in different parts of the geographical range (i.e. geographical vicariants) (Sheth et al., 2020). Both mechanisms would be reflected in a larger distribution range. Most studies have confirmed such a positive correlation (Boulangeat et al., 2012;Brown, 1984;Kambach et al., 2019;Slatyer et al., 2013;Sporbert et al., 2019;Zelený & Chytrý, 2019). For example, in a study on about 1200 plant species in the French Alps, specialist species were found to be more geographically restricted than generalist species (Boulangeat et al., 2012). The same pattern was found in the Czech flora (Zelený & Chytrý, 2019).
Similarly, Kambach et al. (2019) reported a positive relationship between niche breadth and geographical range size, both regionally (1255 plant species in the European Alps) and globally (180 plant species). Recently, a meta-analysis of 64 studies worldwide, found widespread convergence between geographical range size and niche breadth, even after taking into account differences in niche breadth measurements, taxonomic groups, spatial scales and sampling effects across studies (Slatyer et al., 2013). This has raised concerns for specialist species that might be disproportionally affected by habitat loss (Staude et al., 2020). However, contrasting patterns have also been observed (Kambach et al., 2019;Slatyer et al., 2013), potentially reflecting the multitude of factors affecting range size and realized niche breadth, such as dispersal abilities, regional availability of suitable niche space and historical events.
One reason why the positive niche breadth-range size relationship is not systematically observed in nature might stem from the many ways niche sizes are calculated. Numerous approaches have been proposed to estimate the realized niche breadth of species (Guisan & Zimmermann, 2000;Sexton et al., 2017). The resourcebased method (Smith, 1982) determines niche breadth as the range of favourable conditions along certain environmental gradients, such as temperature, soil moisture and nutrients, and light availability. It estimates the Grinnellian niche (Grinnell, 1917;Grinnell & Swarth, 1913) and it is the most widely applied method. However, as the resulting niche breadth depends on the selection of niche axes, it will only represent a part of the whole multi-dimensional niche as defined by Hutchinson (Devictor et al., 2010).
As large distribution ranges involve a higher probability to include a higher variation in climatic conditions, a positive relationship between range size and climatic niche breadth would not be surprising (Slatyer et al., 2013). However, it is much less clear whether range size, as a global distribution characteristic, can be predicted from niche breadth estimates derived from the local scale of populations and the communities in which the populations occur (Kambach et al., 2019). Here, we focus on the community approach, which relies on the assumption that local-scale environmental conditions are reflected in community composition. At this scale, local interactions between species come into play, which can either reinforce broad-scale climatically induced patterns or blur them (Sheth et al., 2020). In the first case, a wider range of communities in which a species occurs would not only indicate the existence of higher environmental variation within the species' range, but also provide evidence that the species is able to compete with many other cooccurring species under these conditions. Thus, in this case the niche breadth-range size relationship would become clearer because environmentally unsuitable habitats within the species' range would remain unconsidered. Alternatively, local interactions might weaken the niche breadth-range size relationship, as the presence and abundance of the co-resident species might depend on different site factors than those relevant for the species under consideration.
One method to estimate this niche breadth based on community composition uses community turnover rates across plots (the taxonomic β-diversity) as a measure of species' niche breadth (Fridley et al., 2007; successively modified by Zelený, 2009 andFridley, 2009). The fundamental assumption of the cooccurrence-based (biotic) niche concept is that widespread species are generalists that should occur with a broader range of community compositions (i.e. in a higher number of different communities) compared to specialists, given an equal drawn number of plots in which the species occurs (Fridley et al., 2007). Overall, this method characterizes both the Grinnellian niche and the Eltonian niche (Elton, 1927), as it quantifies species' response to multi-dimensional environment gradients and considers species interactions as well (Devictor et al., 2010;Fridley et al., 2007). In addition, it can be applied where environmental information is unavailable. Resourcebased and co-occurrence-based approaches differ in the dimensions measured and spatial scales, and can serve as complementary methods to estimate species' niches. While results based on the two approaches are not necessarily correlated (Emery et al., 2012;Pannek et al., 2016), a positive correlation is expected at broad spatial scales (Kambach et al., 2019).
Fagus is a key genus of the northern temperate forest flora.
While phylogenetic relationships (e.g. Denk, 2003;Renner et al., 2016;Shen, 1992), climatic limits (Fang & Lechowicz, 2006), biogeographical history (Denk & Grimm, 2009) and community composition (e.g. Hukusima et al., 2013;Kavgaci et al., 2012;Willner et al., 2017) within the genus have been extensively studied, much less attention has been given to the range size and niche properties of the component species. This knowledge is valuable as it may help understand how Fagus species might respond to climate change.
Increasing temperatures and more frequent extreme events, such as repeated heatwaves and summer droughts, have been projected for many mid-latitude regions (Booth et al., 2012;Geßler et al., 2007).
Fagus species such as F. sylvatica and F. grandifolia are sensitive to high temperature and repeated drought events (Booth et al., 2012;Clark et al., 2011;Silva et al., 2012), and the distribution of European beech (F. sylvatica) has been projected to shift northward in a future climate (Kramer et al., 2010). Studying the relationship between range size and niche breadth could provide valuable information for predicting the future distribution of Fagus species (Sheth et al., 2020).
Species' responses to past environment changes could have been influenced dramatically by evolutionary processes (Kramer et al., 2010;Parmesan, 2006). Species differ in their environmental niche space because their ancestor populations were adapted to different conditions within the ancestral range, or the ancestor was sufficiently plastic to cope with different conditions (Bromham et al., 2020). In the course of speciation, we would expect that the descendant species share either a similar adaptation to local environment or the generalistic genotypic constitution with their ancestors. In both cases, the niche characteristics of closely related species should be more similar than those of more distantly related species (Losos, 2008). Consequently, it is important to analyse niche properties in a phylogenetic framework (Graham et al., 2004;Kozak & Wiens, 2010). Information as to which niche characteristics are more strongly affected by phylogenetic control might help establish hypotheses on the underlying mechanisms responsible for range dynamics and speciation and provide insights into the evolution of different niche axes (Emery et al., 2012;Evans et al., 2009). Variation among different Fagus species may be expected because of regionally varying palaeoclimatic forces on range dynamics (Dynesius & Jansson, 2000;Magri et al., 2006). In the genus Fagus, the widespread and dominant North American F. grandifolia belongs to a more ancestral clade compared to the non-dominant, regionally rare and small-ranged Chinese species (Oh et al., 2016; but see Renner et al., 2016 andJiang et al., 2020). This phylogenetic pattern could suggest an evolutionary tendency towards ecological specialization.
As mentioned above, because species with larger ecological niches occur in a broader range of habitats, they tend to have larger distribution ranges and total population sizes. These conditions should provide higher chances for speciation events in quantitative and qualitative terms. A broader habitat range provides a higher probability for local adaptation processes. Such ecotypes have the chance to get isolated from the main population by ecological or spatial vicariance. The quantitative difference to smallranged species provides a stochastically higher chance for such events as well as for long-distance dispersal events. This opens another possibility for speciation processes due to geographical isolation. Testing whether interspecific similarities of niche characteristics and phylogenetic relatedness are positively correlated (Blomberg et al., 2003;Losos, 2008;Wiens et al., 2010), could provide hints for evolutionary tendencies towards ecological specialization. In addition, identifying the key climatic factors for evolution and speciation of the Fagus species might also enhance our understanding on their response to climate change. So far, although several studies have explored the roles of climatic (e.g, Evans et al., 2009;Graham et al., 2004;Kozak & Weins, 2006) or habitat niche evolution (e.g., Emery et al., 2012) in speciation, this approach has rarely been applied to species' co-occurrence-based assessment of niche breadth.

| Study species
The genus Fagus includes widespread as well as relatively rare species distributed across the Northern Hemisphere, from eastern North America, Europe and western Asia to eastern Asia ( Figure 1).
European beech (F. sylvatica) occurs in Europe, and F. orientalis in western Asia and southeast Europe (Willner et al., 2017). Fagus grandifolia is distributed in eastern North America, with F. grandifolia subsp. mexicana having a narrow, isolated range in the mountains of Mexico. It is noteworthy that the taxonomic status of the three geographically-isolated segregates (F. multinervis, F. orientalis and F. grandifolia subsp. mexicana) has been debated for a long time.
In some studies, F. multinervis and F. orientalis have been treated as subspecies (Renner et al., 2016), while there are recent phylogenetic hypotheses that recognize F. grandifolia subsp. mexicana as a distinct species (Jiang et al., 2020).

| Datasets for niche breadth estimation
The data used were mainly obtained from sPlot-the global vegetation-plot database-which has a worldwide coverage and a standardized taxonomic nomenclature . As data for Fagus in China were quite limited in sPlot, 219 additional Fagus plot data from China were added from our own unpublished field records. The taxonomic backbone of sPlot was used to harmonize the species nomenclature of these additional records. Species co-occurrences in the tree, shrub and herbaceous layer of each vegetation plot were used for co-occurrencerelated analyses. Additionally, the spatial coordinates (latitude and longitude) of the plots were used to extract climatic attributes. As we were interested in estimating niche breadth from the full range of biotic constellations, we did not apply any minimum threshold for the cover of Fagus species in the plot records.
Altogether, 107,758 vegetation plots in which at least one Fagus tree species occurs were obtained for the analyses ( Figure S1, Table 1).

| Datasets for range size estimation
The geographical distribution of the 10 Fagus species (Figure 1) was based not only on the occurrence records described above but also on the data collected by the Chorology Working Group at the University

| Climate data
For the localities of all 107,758 selected plots, we extracted raster cell values from grid layers of the 19 bioclimatic variables (BIO01 to BIO19) provided by the CHELSA database (https://chels a-clima te.org/; Karger et al., 2017). The spatial resolution was 30 arc seconds, corresponding to about 0.64 km × 0.93 km = 0.6 km 2 at the average latitude of the plots (46.31°N). For those plots along the coastlines resulting to be located in the sea due to coarse location accuracy, we extracted the climate information of the closest terrestrial grid cells with the recorded elevation using the snap function in ArcGIS.

| Estimation of biotic (co-occurrence-based) niches from plot data
To estimate co-occurrence-based niche breadth, we generally followed the approach proposed by Fridley et al. (2007), which uses the taxonomic β-diversity metric among the set of plant communities (here, vegetation plots) in which a focal plant species occurs.
To measure plot dissimilarity, we chose the multiple Simpson index (Baselga et al., 2007), which can disentangle changes in species composition caused by changes in species identities (species turnover) from those caused by species richness differences (nestedness effect), whilst being independent of the absolute species richness in the plots (Baselga et al., 2007;Manthey & Fridley, 2009). As the quantities of plots varied a lot among different Fagus species (Table 1), we conducted a resampling procedure to control for possible sample size effects. For each of the 10 Fagus species we calculated niche breadth 100 times by randomly drawing 20 plots with replacement out of the total number of plots that contain the focal species, and then taking the average dissimilarity value of the 100 iterations (Fridley et al., 2007;Pannek et al., 2016). We then calculated the variance of the dissimilarity values across the 100 runs.
Results based on other dissimilarity indices for turnover calculations, i.e. the Jaccard and Sørensen index (Manthey & Fridley, 2009), are presented in the Supporting Information ( Figure S2). The Sørensen index is a linear function of the multiplicative Whittaker's β-diversity metric (β w ), which eliminates possible effects of species pool sizes (Manthey & Fridley, 2009;Zelený, 2009).

| Estimation of climatic niches from plot data
We applied the method of dynamic range boxes (DRB) to estimate the climatic niche breadth since it has proven to be relatively robust to sampling effects and outliers, especially at high dimensionality (Junker et al., 2016). Accordingly, we calculated the realized climatic niche breadth based on the above-mentioned bioclimatic variables, using the R package 'dynRB' (Schreyer et al., 2018). To avoid bias related to inter-correlation, we excluded one of those variables with abso- To limit pseudo-replication at a given site, in cases where two or more plots had identical spatial coordinates, only one plot was

TA B L E 1
Geographical range sizes and plot information of the 10 extant Fagus species selected randomly. As a result, the sample size for the climatic niche estimation was reduced to 61,717 plots. We used two different aggregation functions ('mean' and 'gmean', both included in the package 'dynRB') to compare the coordinate-wise volumes. Note that the 'product' method was omitted as it was not recommended by the authors (Junker et al., 2016). Results based on the 'mean' method are shown in the Supporting Information ( Figure S3).

| Phylogenetic analysis
Before using phylogenetic methods to test whether specific macroclimatic traits are species-specific across genera, we ran a Random Forest (RF) analysis to select the key bioclimatic variables.
Specifically, we quantified the discriminability of each bioclimatic variable range between the sister species. We employed two widely used variable importance measures: the mean decrease accuracy and mean decrease Gini index (Breiman, 2001). Only the plot data with distinct coordinates (n = 61,717) were used here.
We calculated RF ensembles of recursive classification trees with the 'randomForest' package in R (Breiman, 2002). As settings, we used 10,000 iterations and the 'tuneRF' function to identify the 'optimal' number of input variables randomly chosen at each node. Since the number of plots per Fagus species varied considerably (n = 11 to 54,397), both a fixed sample size (n = 10) approach and a proportionate stratified sampling procedure (following the species sequence in respectively, with replacement) were adopted. For comparison, a conservative variant with the fixed sample size of n = 10 was calculated.
Out-of-bag (OOB) estimation of error rate was used to assess classification success and a permutation cross-validation test with 20% withhold test data was used to control for overfitting.
To assess whether niche breadth estimates as well as the least and most discriminating bioclimatic variables identified by the RF analysis (taking the 5th and 95th percentiles, respectively) are under phylogenetic control or show a random pattern of evolution, phylogenetic signals were tested across all 10 Fagus species. For each species, we calculated the biotic and climatic niche breadth estimates and the distinctive bioclimatic variables. Blomberg's K value is an indicator of the strength of the phylogenetic signal, which indicates strong phylogenetic control of characteristics when greater than one, and a random pattern of evolution when close to zero (Blomberg et al., 2003;Kembel, 2010

| Statistical analyses
We tested for differences in biotic niche breadth between the 10 Fagus species with a one-way analysis of variance (ANOVA) and Tukey's HSD post hoc tests, using an n = 100 random selection of plot records as replicates. To test hypothesis H1 (i.e. that range size is related to niche breadth) and H2 (i.e. that biotic and climatic niche breadths are correlated), we assessed the relationships between the different niche traits and range size by performing the phylogenetically corrected generalized least-squares (PGLS) regression using the R package 'caper' (Orme et al., 2018). Considering the relatively small sample sizes and the outlier species F. multinervis, we further assessed the relationships with a Spearman's correlation test. In addition, the PGLS regression and Spearman's correlation test were also performed when excluding F. multinervis. For hypothesis H3, which proposes that biotic and climatic niche breadths are non-randomly associated with phylogeny, we calculated the phylogenetic signal, using the 'picante' package in R (Kembel, 2010). Plotting of the phylogenetic signals was done with 'phylo4d' from the 'phytools' package (Revell & Revell, 2019). The RF analysis was done using the 'rf', 'rfUtilities', and 'ran-domForestExplainer' packages. All statistical analyses were conducted with R v3.5.1 (R Core Team, 2018), and further graphs were produced with the R package 'ggplot2' and ArcGIS 10.3.

| Overall patterns of range sizes and niche breadths
Geographical range sizes of the 10 Fagus species differed by five orders of magnitude, ranging from 37 km 2 (F. multinervis) to 2.1 × 10 6 km 2 (F.  (Figure 2). Similar patterns were observed based on the Jaccard and Sørensen index ( Figure S2).  (Figure 3). The aggregation method 'mean' resulted in a similar pattern ( Figure S3). A PCA of the 10 climatic variables also showed that plots of F. hayatae had a wide range along the first two axes ( Figure S4).

| Relationships between range sizes and niche breadths
According to the PGLS regression and Spearman's correlation test, neither the biotic nor the climatic niche breadth estimate was related to range size (p > 0.05; Figure 4a

| Phylogenetic analyses of climatic tolerances and niche properties
The RF analysis successfully classified the species by climate with an out-of-bag (OOB) error rate of 11.1% when using a fixed sample size (n = 10), but reaching an OOB error rate of 1.95% when using proportional sample sizes. Similar values were obtained with the crossvalidation approach when withholding 20% test datasets (9.86% and 2.03%, respectively). For both sampling strategies, the bioclimatic variable BIO03 (Isothermality, which is the mean diurnal range of temperature divided by the annual range of temperature, i.e. BIO02/ BIO07) had the strongest importance for discriminating the species, as it returned the largest decrease in accuracy and increase in Gini node impurity when excluded ( Figure S5). In contrast, the exclusion of the bioclimatic variable BIO06 (Minimum temperature of the coldest month) had the lowest importance ( Figure S5).
Blomberg's K values were smaller than one for all the biotic and climatic niche characteristics as well as the two aforementioned bioclimatic variables (BIO03 and BIO06). No biotic or climatic niche characteristics were significantly different from a random distribution (p > 0.1; Figure 5), suggesting that we cannot conclude that these characteristics are phylogenetically conserved within the genus Fagus. Rather, small and large niche breadths are distributed randomly among the clades within Fagus.

| DISCUSS ION
Contrary to our first hypothesis and the commonly reported positive relationship between range size and niche breadth (Boulangeat F I G U R E 2 Co-occurrence-based niche breadths of the 10 extant Fagus species using multiple Simpson index for turnover calculations. The colour spectrum follows the distribution range maps (Figure 1). Different letters indicate significant differences according to a Tukey's HSD post-hoc test (p < 0.05)

F I G U R E 3
Climatic niche breadths of the 10 extant Fagus species based on the resource-based method of dynamic range boxes (DRB). The aggregation method used was 'gmean' (Junker et al., 2016). The colour spectrum follows the distribution range maps (Figure 1) et Brown, 1984;Kambach et al., 2019;Sheth et al., 2020;Slatyer et al., 2013), we found no significant relationship between range size and the biotic and climatic niche breadth estimates for the species in the globally important tree genus Fagus.
Setting possible methodological issues aside (see Appendix 1 in the Supporting Information), there are many biological, geographical and historical factors that might affect the niche breadth-range size relationship for these closely related species, such as dispersal ability F I G U R E 4 Relationships between different niche breadths and range size (a-b) or log-transformed range size (c-d), and between biotic and climatic niche breadth (e-f). Panels a-f are based on the phylogenetically corrected generalized least-squares (PGLS) regression and panel g is based on the Spearman's correlation test (p < 0.05). Considering the outlier point of Fagus multinervis, fitted lines and coefficients of the PGLS regression of the 10 Fagus species are not shown in panels c-f. Multiple Simpson, Jaccard and Sørensen are different indices for biotic niche estimation. Mean and gmean represent different aggregation methods for climatic niche estimation and refer to the arithmetic and the geometric mean of the different niche dimensions, respectively. The acronyms BN and CN mean biotic niche and climatic niche, respectively or biotic interactions with other species, the regional availability of suitable niche space or differences in landscape heterogeneity, evolutionary processes or historical events, as well as human impacts (Boulangeat et al., 2012;Lambdon, 2008;Sillero, 2011;Slayter et al., 2013;Wandrag et al., 2019).
The North American species F. grandifolia probably represents a member of the oldest clade (Renner et al., 2016). However, its large distribution and its relatively low niche breadth may, to a large extent, represent lower spatial species turnover and environmental heterogeneity in eastern North America in comparison to Europe and East Asia. It is well-established that the diversity of angiosperms in eastern China is larger than in eastern North America (e.g. Qian et al., 2005). Fagus hayatae has a very disjunct distribution with two subranges in mainland China and Taiwan Island, and thus, covers a broad, yet disjunct range of climatic conditions (Hukusima et al., 2013;Shen, 1992;Shen et al., 2015).  (Gholizadeh et al., 2020;Kavgaci et al., 2012;Shen, 1992). For example, in the Hyrcanian area, F. orientalis grows between around 300 to 2700 m a.s.l., covering a broad range of climatic conditions (from warm and humid to cool and dry) (Gholizadeh et al., 2020). Similarly, F. sylvatica has a broad ecological amplitude and a wide range of habitats in Europe (Magri, 2008;Ujházyová et al., 2016;Willner et al., 2017). The Fagus species in Korea (F. multinervis) and Japan (F. crenata and F. japonica) have narrower biotic and climatic niche breadth, consistent with these species being limited to isolated islands, especially for F. multinervis, endemic to Ulleungdo Island (Hukusima et al., 2013).
The distribution of Fagus was severely influenced by the Quaternary historical events in the Northern Hemisphere (Huntley et al., 1989;Liu et al., 2003;Magri, 2008). Following the early Quaternary, the geographical range of Fagus dramatically shrunk and shifted southwards during the glacial periods, although the history of Fagus species during the interglacials before the Holocene remains unclear (Hukusima et al., 2013;Huntley et al., 1989;Liu et al., 2003;Magri, 2008;Magri et al., 2006). After the Last Glacial Maximum, northward or northwestward re-immigration of Fagus from the southern refugia happened in North America, Europe and Japan (Huntley et al., 1989;Liu et al., 2003;Magri, 2008), despite strong dispersal limitations (Saltré et al., 2013). In China, however, the northward postglacial migration of Fagus was limited by the monsoon climate with early-season aridity, which restricted the F I G U R E 5 Different characteristics and their phylogenetic signal for the 10 extant Fagus species, measured by Blomberg's K. If K is significantly larger than one, the characteristic is regarded as phylogenetically conserved (Blomberg et al., 2003). Blomberg's K was not statistically significant (p > 0.05) for any characteristic. The phylogenetic tree is based on results provided in Qian and Jin (2016) and Oh et al. (2016). The size/ colour of the circle represents the scaled and centred value of the corresponding characteristic. The acronym MTCM means minimum temperature of the coldest month distribution ranges of the Chinese species to the subtropical region only (Liu et al., 2003;Shen et al., 2015).
Human activities might have affected the distribution and community composition of Fagus forests to a certain degree. For example, F. sylvatica forests in Europe have been managed for a long time as beech is an important economic tree for wood production, especially in northern-central Europe (Magri, 2008). Activities such as livestock grazing and disturbance of the preceding forests by fire before Fagus became established have promoted the spread of Fagus in northern Europe (Bradshaw et al., 2010).
Overall, these idiosyncratic impacts on range dynamics differed from those factors affecting niche breadth, as witnessed by the absence of a significant correlation. Our study, with a sample size of 10 species, suffers from low statistical power and does not preclude finding such a relationship across species within other genera. It would be, therefore, interesting to benchmark our finding that different factors drive range size and biotic niche width independently on a larger dataset.
Additionally, the lack of relationship between range size and niche breadth might also suggest that the prevalent processes of community assembly differ across Fagus species. In most communities, both deterministic and stochastic processes are at work simultaneously (Stegen & Hurlbert, 2011). Thus, a low taxonomic turnover (β-diversity), such as in F. grandifolia stands in North America, might reflect a low impact of stochastic processes, but might also be brought about by deterministic processes such as strong environmental filtering under homogeneous climatic conditions. In contrast, high β-diversity, such as for the Fagus stands in China, might indicate that stochastic processes dominate (Daniel et al., 2019), resulting from the more pronounced geographical isolation and more glacial refuge areas in China. However, the same pattern could also be caused by deterministic processes with different environmental filtering regimes under heterogeneous climatic conditions.

| Correlations between niche concepts
The co-occurrence-based biotic niche breadth estimates were uncorrelated with the climatic niche breadth estimates, incongruent with our second hypothesis (H2). Although positive relationships have been observed in previous studies (Kambach et al., 2019;Pannek et al., 2016), it has to be admitted that the two adopted niche concepts differ greatly in dimensions and spatial scales and thus are not necessarily correlated (Emery et al., 2012;Pannek et al., 2016). The incongruence of results based on different indices of realized niche breadth suggests that these indices carry complementary information.

| No phylogenetic signal for niche properties in Fagus
Our results suggest dynamic development of the biotic and the climatic niches, which does not seem to be related to the rooting depth and phylogenetic distance of the respective clades (Losos, 2008;Wiens et al., 2010), in contrast with our hypothesis H3. Considering the lack of a clear phylogenetic signal towards ecological specialization, we conclude that the complex phylogeographical history of the genus Fagus does not allow us to find support for the stated "specialization hypothesis". In previous studies, both divergent (e.g. Evans et al., 2009;Graham et al., 2004) and conserved (e.g. Kozak & Wiens, 2006;Peterson et al., 1999)  and isolated mountain ranges (Oh et al., 2016). Having evolved from a common ancestor species, such geographical separation and differences in available niche space may result in the development of different biotic and climatic niches.
Cold tolerance has been regarded as a key trait for the geographical distribution patterns of trees (Hawkins et al., 2014;Wiens & Donoghue, 2004). The monthly mean of the minimum daily temperatures (BIO06) has been used to represent the cold tolerance of tree species for which large scale physiological data are usually hard to obtain (Hawkins et al., 2014). The RF classification analyses revealed that cold tolerance did not discriminate the species, suggesting it might be conserved across the whole Fagus-clade, possibly representing an ancestral adaptation. Uniform cold tolerance across the genus is consistent with Blomberg's K, revealing the absence of a phylogenetic signal within the genus. However, testing for conservatism at the genus level would require a broader taxonomic scope, for example, by including the whole Fagaceae family. Nevertheless, genus-level conservatism is suggested by the consistent association of Fagus with mesic temperate climates, relative to the broader distribution of the family.

| CON CLUS IONS
We estimated the biotic and climatic niche breadths of all 10 extant Fagus species and examined their relationships with range size.
Biotic and climatic niche breadth were uncorrelated with range size and phylogeny, and also incongruent with each other. Furthermore, there was no evidence for evolutionary tendencies towards ecological specialization in the younger Fagus clades occurring in East Asia.
We conclude that within widespread groups of related species such as in the Fagus genus, general macroecological patterns such as the range size-niche breadth relationship might be overridden by different regionally available niche space opportunities, differences in landscape heterogeneity and Quaternary histories.

ACK N OWLED G EM ENTS
We thank all the data contributors and those who helped in the fieldwork, as well as M. Sporbert, G. Seidler, Y. H. Feng and H. Zhang for