Present and Future Climate ‐ Related Distribution of Narrow ‐ Versus Wide ‐ Ranged Ostrya Species in China

: The niche breadth–range size hypothesis states that geographic range size of a species is positively correlated with its environmental niche breadth. We test this hypothesis and examine whether the correlation varies with climate change and among taxa through modeling (processing Maximum entropy (Maxent)) potential distributions in present and future climate scenario of four sympatric Ostrya species in China and with different geographic range sizes, including extremely rare O. rehderiana . Potential geographical distributions of narrow ‐ versus wide ‐ ranged Ostrya spe ‐ cies were predicted based on their niche breadths. Niche equivalency and similarity tests were per ‐ formed to examine niche overlap between species pairs. Potential distribution areas of wide niche breadth species ( O. japonica and O. trichocarpa ) were significantly wider than those of narrow niche breadth species ( O. multinervis and O. rehderiana ) although niche divergence was hardly observed among them. In the future scenarios of global climate change, wide ‐ ranged O. japonica would have wider potential distribution than in the current scenario, even expanding their geographic range. Conversely, suitable habitats of narrow ‐ ranged O. multinervis and O. rehderiana would be reduced strikingly in future scenarios compared to in the current scenario, and they might be subjected to a high risk of extinction. Potential distribution range sizes of the Ostrya species would positively cor ‐ relate with their niche breadths in future scenarios, and their niche breadths would determine their distribution variation with climate change. The Ostrya species having broader niche currently would be further widespread in future scenarios while narrowly distributed Ostrya species having narrower niche currently would further reduce their distribution range under changed climate and might be subjected to a high risk of extinction in future scenarios. Our results support the range size–niche breadth hypothesis both at present and future climate scenarios, and they provide useful reference for conservation of rare species like O. rehderiana .


Introduction
Geographic range size, the area in which a species occurred, is determined by intrinsic (e.g., propagule or body size, dispersal ability, population density) and extrinsic (e.g., local environmental conditions, topographical features) factors [1,2]. Among those complex closely related factors, the dominant factors are usually changing and may vary among organisms and species [3][4][5][6][7]. Numerous hypotheses have been proposed to explain the diversification in geographical range size among species and across space, such as geometric constraint [8], habitat area [9], climatic variability [3], dispersal ability [6], niche breadth [10].
Among the mentioned above, niche breadth-range size hypothesis states that geographic range size is positively associated with environmental niche breadth and has been supported in some species (e.g., Gaston and Spicer [11], Boulangeat et al. [12]; Botts et al. [13], Carrillo-Angeles et al. [14]; Yu et al. [15]; Vincent et al. [16]). Species with broad niche may have better adaptation to various environmental types than those with narrow niche [17][18][19][20]. If this hypothesis holds true, it can be inferred that species having wide distribution range currently will further widen their distribution ranges, while species having narrow distribution range currently will further narrow their distribution ranges, under changed climate in the future [21]. However, the positive relationship is still uncertain in different taxonomic groups; additional evidence is required to support this hypothesis [22][23][24].
Moreover, geographic range size is usually used to predict the extinction vulnerability or invasion risk of a given species [25,26]. Predicting the potential geographic distribution range size and identifying the dominant factors would underlie biodiversity conservation and management [27][28][29]. Species distribution models (SDMs) are practical to predict the potential geographic distribution under different climate scenarios [30,31], which have been widely used for the conservation of rare species [14,[32][33][34]. Maximum entropy (Maxent) modeling is one of the most commonly used predictive methods in SDMs [35], which can achieve greater predictive accuracy than other methods with presence-only and biased sampling data [33,[36][37][38]. Benefiting from the statistical mechanics, the Maxent approach is powerful modeling for geographic distributions of rare species with narrow ranges and presence-only data [39][40][41][42]. The Maxent approach also performed well for distribution prediction of widespread species [33,43,44].
Here, SDMs are applied to four sympatric Ostrya species in China and with different geographic range sizes, including extremely rare O. rehderiana, to test the range size-niche breadth hypothesis. We expect (a) the widespread Ostrya species to have wider niche breadths than the narrowly distributed rare ones, and (b) Ostrya species having wide and narrow distribution range currently will further expand and shrink their distribution ranges under changed climate in future scenarios, respectively.
All of the Ostrya species are deciduous trees with scaly and rough barks. The male inflorescences would be formed from April to July, blooming in the spring of next year [49]. O. japonica is widespread from southwest to northeast China, while the other three  [49,51]. O. rehderiana and O. multinervis had similar wild geographic distribution in the last glacial maximum (LGM), but then the habitat of the former was reduced dramatically while the latter maintained a stable population size [52]. Furthermore, O. rehderiana is on the International Union for Conservation of Nature (IUCN)'s Red List of Threatened Species as critically endangered [50]. It is also among the first-class state protection wild plants in China based on the latest report [53]. Ostrya species have different geographic range sizes, providing an opportunity to test the niche breadth-range size hypothesis through expectations of the models using narrowly distributed rare versus widespread congeneric species.

Species Occurrence Records
We collected occurrence records of four Ostrya species in China from the Global Biodiversity Information Facility (GBIF, https://www.gbif.org (accessed on 15 June 2020)) [54][55][56], the Chinese Virtual Herbarium (CVH, http://www.cvh.ac.cn (accessed on 15 June 2020)), the Plant Photo Bank of China (PPBC, http://ppbc.iplant.cn (accessed on 15 June 2020)), and the National Specimen Information Infrastructure (NSII, http://www.nsii.org.cn (accessed on 15 June 2020)). We also record the geographic coordinate using a handheld global positioning system (GPS) device according to the field investigation, as well as information in the literature [51,52,57]. Records with unclear or incorrect information were excluded, and the data collected before 1970 were also excluded because they are out of present climatic periods [58]. Each species' distribution points without duplication in a 2.5 arc-minute resolution grid cell were selected to avoid sample bias and reduce spatial autocorrelation [58][59][60][61]

Climate Variables
Current climate variables (BIO1-19) were obtained from WorldClim Version 2.1 [62] at 2.5 arc-minute resolution layers of China ( Table 1). The 19 bioclimatic layers are the average temperature and precipitation data from 1970 to 2000. Considering that the spatial heterogeneity of climate variability may cause the variation of potential distribution in regional scales, it is necessary to adopt an appropriate climate system model [63,64]. We employed the future climate projections derived from the climate system model in the middle resolution of Beijing Climate Center (BCC-CSM2-MR), demonstrating the bestperforming in China with Coupled Model Intercomparison Projects 6 (CIMP6), to forecast the Ostrya distribution in the future [65][66][67]. Additionally, we obtained bioclimatic projections from the model processed for shared socio-economic pathways (SSPs) 1-2.6 and 5-8.5 in the period 2081-2100. SSPs were the new climate projections updated from the representative concentration pathways (RCPs) based on the Sixth Assessment Report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) [65,66]. SSP1-2.6 and SSP5-8.5 represent the low and high radiative forcing by the end of the 21st century, respectively [68,69]. Under SSP1-2.6, the radiative forcing would stabilize at 2.6 W m −2 and global warming would not be increased more than 1.5 °C in 2100, while under SSP5-8.5, emissions would be high enough to produce a radiative forcing of 8.5 W m −2 in 2100 [70].

Maxent Modeling
We employed Maxent V3.4.1 [40] to predict the potential distributions of Ostrya in current and future scenarios. According to the tutorial of Maxent software [71], environmental layers and occurrence records were tested with each Ostrya species to process the species distribution models. The recommended default values were accepted, including maximum iterations (N = 500), regularization multiplier (N = 1), convergence thresholds (1 × 10 −5 ), and the maximum number of background points (N = 1 × 10 5 ). The background points were taken from the entire study area, which is a priori equally likely to contain the species [72][73][74]. The presence/absence is treated as unknown at these locations, and Maxent contrasts the features at the presence locations to those in the background sample [33]. To reduce the number of correlated variables, the Spearman correlation coefficient and jackknife method were applied to inspect the importance of the climate variables [75,76]. The variable with a higher contribution was rejected when the absolute value of the correlation coefficient of a climate variable pair was higher than 0.8 (p < 0.05). To avoid the over-estimation of models by the area under the receiver operating characteristic curve (AUC), we also evaluated the partial AUC (pAUC) [31,77,78] with R package ntbox [79]. In addition, models running replicated 10 times were required respectively; replicated run type was cross-validation due to the small occurrence data [31,40]. Based on potential distributions in the present scenario, we predicted the range variations of Ostrya in the future. Finally, we analyzed and visualized the occupied areas of Ostrya species in ArcGIS 10.2 (Esri, Redlands, CA, USA). The potential distribution of species was quantified from 0 to 1; we reclassified it into four groups following the method used by Yang et al. [75] and others [80][81][82], namely, least potential (<0.2), moderate potential (0.2-0.4), good potential (0.4-0.6), and high potential (>0.6). Moreover, we calculated the geographic ranges for each of four Ostrya species in the high potential region under present and future climate scenarios.

Niche Breadth Comparison
The niche breadths of four Ostrya species were estimated by R package ENMTools [83]. B1 and B2, representing the inverse concentration and uncertainty niche breadth respectively, were calculated to quantify the niche breadth of each species, and easily for general use [84]. This index estimated the niche breadth were weighted by the number of different environment variables. The value of 1 indicates generalist taxa that are equally abundant across environments, and 0 indicates specialist taxa that favor specific environment [85]. Each species was calculated with 10 repetitions, the results were compared using one-way ANOVA followed by Tukey's honestly significant difference (HSD) test. These analyses were performed using R version 4.0.2 [86].

Niche Overlap Test
To test whether the niches diverge among four Ostrya species, we employed niche equivalency and similarity tests to compare our observed niches overlap against null distributions by R package ecospat [87]. Kernel smoother density was used to estimate the densities of the species occurrence in environmental spaces, while the environmental variables were reduced into axes using principal components analysis to quantify the metrics of niche overlap tests [88].
Schoener's D and Hellinger's I were calculated to examine if there were niche overlap between species pairs; they are more robust and direct than other metrics [88]. Metric D reflects microhabitat environments, while I focus on the probability distributions between species pairs [89,90]. To ensure that the null hypothesis can be rejected with high confidence, we performed 1000 randomized pseudo-replicates for each pair of four Ostrya species tested respectively. Both values ranged from 0 (no overlap) to 1 (complete overlap) [87,91]. These analyses were performed using R version 4.0.2 [86].

Model Performance and Contributions of Climate Variables
The AUC and pAUC values of all models were higher than 0.8 and 0.9, respectively ( Table 2). The distributions of Ostrya species were influenced by different climate variables (Table 1 and Figure A1). The minimum temperature of the coldest month (BIO6) was the principal temperature factor impacting the distributions of four Ostrya species, achieving 59.87% and 70.12% in the contributions of O. rehderiana and O. trichocarpa, respectively.
The lowest temperature in winter around ±5 °C was suitable for the growth of O. multinervis and O. rehderiana, while the other two species could adopt a wider range of BIO6 ( Figure 2). Besides, Ostrya species were affected by different precipitation variables except for O. trichocarpa which remained mostly unaffected. The distribution of O. multinervis was largely determined by the precipitation of the driest month (BIO14); precipitation in this month lower than 23 mm was unsuitable for O. multinervis ( Figure A1l). Precipitation seasonality (BIO15) has an important impact on O. rehderiana, the adaptive range was approximately 62.1% ( Figure A1m).

Potential Distributions of Ostrya under Present and Future Climate Scenarios
Four Ostrya species showed diverse potential distributions under the present climate scenario (Figure 1). O. japonica, the widest distributed species, was adapted to the climate in the northward and medium-high altitude region. The Loess Plateau was the core distribution region of O. japonica (Figure 1a). O. trichocarpa mainly occurred in the south of China, including Sichuan Basin, southern coastal, Hainan and Taiwan Islands (Figure 1d). O. multinervis exhibited a scattered distribution in the eastern and midwestern China, as well as high-altitude mountains in Taiwan Islands (Figure 1b). Central-eastern China, especially in the Hunan and Zhejiang provinces, was suitable for the growth of O. rehderiana; there was a partial overlapping with O. multinervis in the east (Figure 1b,c).
The climate in the future was more suitable than the present for O. japonica and O. trichocarpa, but the potential distribution range of O. multinervis and O. rehderiana would be reduced significantly in 2081-2100 (Figures 3 and A2). With proceeding climate change, O. japonica and O. trichocarpa were predicted to broaden their potential distribution regions in China. O. japonica expanded the range to higher altitude, including the south Tibetan Plateau and the Hengduan Mountains in the southwest China, the Changbai and Daxinganling Mountains in the northeast China (Figures 3a and A2a). However, under SSP5-8.5, the climate in the northeast China would no longer be suitable for its growth in the end of 21th century ( Figure A2a). The high potential distribution ratio of O. japonica would be shifted to 6.07% in 2081-2100 under SSP5-8.5 climate scenario, which was the largest area among all species under SSP5-8.5 (Figure 4b). O. trichocarpa consolidated the core areas of high potential habitat, including Sichuan Basin, southern China, Hainan and Taiwan Islands (Figures 3b and A2b). However, both of O. multinervis and O. rehderiana were confronted with habitat reduction significantly, along with climate change (Figures 3c,d and A2c,d). Although the current distribution was broader, climate change caused the potential habitat of O. multinervis retracting and fragmenting over time (Figures 1b, 3b and A2b). Eastern China would no longer be a suitable shelter for O. rehderiana under SSP5-8.5, where exists the distribution location of this species, as well as with high potential distribution region under the current climate scenario (Figures 1c, 3c and A2c).

Niche Breadth and Overlap
The niche breadth of O. multinervis and O. rehderiana were significantly narrower than O. japonica and O. trichocarpa (Table 3) [87], ranging from 0 (no overlap) to 1 (complete overlap). The black dotted line represents the observed niche overlap between species pair (i.e., the empirical values of D), whereas the histograms are those expected under the null hypotheses, and * represents that the niches are significantly (p < 0.05) more different than expected by random.

Niche Breadth and Distribution Range Size
The niche breadth-range size hypothesis was supported by the results of the four Ostrya species in this study. The models processed by Maxent were creditable in the predicted distributions of Ostrya species in China (Table 2) [77,78]. Two species with broader niche breadth, O. japonica and O. trichocarpa, acquired larger potential geographic occurrence areas and would be further widespread in the future (Table 3 and Figures 3a,d and A2a,d). In contrast, O. multinervis and O. rehderiana with narrower niche breadth had smaller potential geographic occurrence areas and would further narrow the area and have greater extinct possibilities under changed climate in the future (Table 3 and Figures 3b,c and A2b,c). To our knowledge, Ostrya species have not been reported as invasive species, nor as those whose ecological niche shifts appeared fast [90,92], suggesting that niches of the species we studied here would not be able to shift sharply at least in a relatively short period like centuries [93]. In this study with the duration of one hundred years, therefore, the niches of Ostrya species were assumed retained while the geographic ranges are variable in response to climate change [94]. This is also in accordance with previous studies; broad niche species would cross a wider range than species with specialized niche [13,15,16]. Furthermore, no phylogenetic relationship was found in the niche breadth of Ostrya species [51], which was consistent with the study in Fagus species [24].
Our results suggest that climatic variabilities strongly affected the distributions of genus Ostrya in China. Although the niche divergence existed only between O. japonica and O. multinervis, the low niche overlap indicated the variation responded to climatic variables in different species (Figure 5 and A3). Based on the contributions of climate variables (Table 1), the different curves of response to the minimum temperature in the coldest month (BIO6) might be one of the crucial reasons for the divergence of narrow-and wide-ranged Ostrya species. Suitable ranges of the low temperature in winter differed between the narrow-and wide-ranged species (Figure 2), which was the principal bioclimatic variable influencing the distribution of the four Ostrya species. Species with narrow geographic ranges, due to their lower level of genetic variation and plasticity in response to novel selection pressures, tend to be more sensitive to climate change, as compared with those with wide geographic ranges [7,95], just as indicated by Aspinwall et al. [96] that tree species with narrow distribution range are more vulnerable to a warmer climate than widespread ones. The endangerment for O. multinervis and O. rehderiana may be due to their high sensitivity to temperature in winter. O. japonica and O. trichocarpa both having broader potential distribution could withstand lower temperatures ( Figure 2). Low temperature range and freezing risk are the major environmental factors determining the distributions of tree species [97,98]. Further research is needed to examine morphological, physiological or molecular responses of Ostrya species to chilling stress. In addition, the temperature in winter would have a significant impact on the naked male inflorescence of Ostrya species [47]. Male inflorescences of O. rehderiana formed from May to June of the previous year and bloomed in April after overwintering [99], while the germination of seeds in O. rehderiana needs prolonging (>100d) cold stratification to break dormancy [100]. The narrow winter temperature range requirement of O. rehderiana appears to be an adaptability compromise between these factors (Figure 2). On the other hand, O. multinervis and O. rehderiana were more sensitive to precipitation than the other two species, especially in the driest month and seasonality during one year (Table 1 and Figure A1l,m). The increasing unbalanced and extreme precipitation under future climate scenarios might be another issue that determines the divergence of range sizes with broader niche species [101].
Notably, the niche breadth of least-record species O. trichocarpa was wider than other narrow-ranged species (O. multinervis and O. rehderiana), and it also had larger potential habitats in the present and future. The ecological niche of species contains two dimensions of both fundamental and realized one [102,103]. The former involves the relationship between species and environmental factors, while the latter considers biological interactions [103][104][105][106]. Our results modeled the fundamental niche of Ostrya species only, while the effects of intra-or inter-specific relationships were not considered. Moreover, the existence of plants also depends on various environmental conditions, the climate niche that we focus on is one type of fundamental niche [107]. This may be one of the reasons for the difference between O. trichocarpa and the other two narrowly distributed species. More studies on those species are needed in the future. Besides, the discrepancy might be attributed to the accuracy of modeling. It is undeniable that the model's accuracy was positively related to the sample size [108,109], even though Maxent could achieve excellent performance on predicting the distribution of narrow ranged species [110]. The extremely small sample size of O. trichocarpa and presence-only data used in models might increase the uncertainty of Maxent model outputs [39,111]. More algorithms should be considered to improve the accuracy of modeling [112]; reciprocal transplantation experiments also should be needed to test the performance of models [113][114][115].

Conservation Implications
Most of the narrow-ranged species are rare species and at high risk of extinction [26,116,117]. The positive relationships between range size and niche breadth could provide a reliable reference for conservation planning and policy making. Our findings of potential Ostrya suitable habitats in current and future climate scenarios are suggestive for their conservation.
The SDMs projected a dramatic reduction in the distribution areas of O. multinervis and O. rehderiana, where the survival probability in east China would be reduced. O. rehderiana would have a high risk of extinction despite its potential distribution region in Zhejiang, Hunan and Jiangxi Provinces. O. trichocarpa would expand enormously in the future; its potential habitat would range from the southern coastal areas to the inland areas of southwestern China as well as Taiwan and Hainan Islands (Figures 3 and A2). Climate change would intensify habitat fragmentation of O. multinervis and further human activities would aggravate the process, if possible, which would increase the difficulty of their conservation (Figures 3b and A2b) [26,[118][119][120]. Considering of the distribution region shift of O. rehderiana, ex situ conservation or national nature reserves in other places such as Hunan Province should be implemented (Figures 1c and 3c) [115]. Considering that deleterious mutations accumulated so far are one of the causes for the rarity of O. rehderiana, the genetic diversity should be maintained through conserving Ostrya [52].

Limitations
In this study, the niche breathes of Ostrya were estimated by a model whose performance would be limited by the dimensionality of predictive factors [113,121]. For example, the species with samara (e.g., Ostrya species) rely on wind dispersal [47,122], while the seedlings of O. rehderiana are weak and easy damage, for instance by strong wind [99]. The topographic heterogeneity in soil nutrients and microclimate would have an impact on the species diversity [123,124]. The opposite responses to human influence might aggravate the differences between rare and widespread species [26,44]. Thus, future studies should consider more possible influence factors on the model construction. Unfortunately, there have been very few detailed demonstrations on other abiotic and biotic environmental factors which may vary in the future, causing our results to deviate from the actual situation [125,126]. Furthermore, the model's accuracy is usually positively related to the sample size as mentioned in 4.1 [108,109]. Limited by data availability, the occurrence records included in our study were not large enough. Thus, more field surveys would be needed to extend the knowledge about the distributions of Ostrya species in China [58,76]. Field studies, including common garden and reciprocal transplantation experiments, should also be taken to further compare widespread and rare species under environmental tolerance [113,114].
Moreover, genetic variation has an impact on the distribution of species, and the accumulation of deleterious mutations might explain why O. rehderiana is rare [52]. Widespread species O. japonica has higher genetic diversity than others [57]. The species diversification rate is also driven by the climate change [127]. Thus, more attention should be paid to the genetic diversity impacting on species niche studies in the future.

Conclusions
Potential distribution range sizes of the four Ostrya species would positively correlate with their niche breadth in the present and future scenarios; niche breadth would determine the distribution variations with climate change. The lowest temperature in winter would be the major climatic factor influencing the establishments of the four Ostrya species. The wide niche breadth species (O. japonica and O. trichocarpa) would be further widespread in the future scenarios while narrowly distributed species (O. multinervis and O. rehderiana) having narrower niche currently would further narrow their distribution range under changing climate and might be subjected to a high risk of extinction in future scenarios. Our findings support the niche breadth-range size hypothesis and provide a valuable reference for the conservation of endangered and rare species like O. rehderiana. In the future, research topics related to range sizes of more taxa on broad biogeographic scales and their dynamics in a changing climate are expected.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Acknowledgments:
We thank Wei-Jun Zhang for his help in data collection. We also thank the anonymous reviewers for their constructive comments.  Table 2 for the meaning of each climatic variable. The black dotted line represents that the probability of species existing is 0.6, species would archive high potential adaptation above this line.  O. trichocarpa (l) using similarity test. Pink and blue represent equivalency and similarity tests respectively. The Matric I assumed the niche overlap in distribution probability (Warren et al., 2008), ranging from 0 (no overlap) to 1 (complete overlap). The black dotted line represents the observed niche overlap between species pair (i.e., the empirical values of I), whereas the histograms are those expected under the null hypotheses, and * represents the niches are significantly (p < 0.05) more different than expected by random.