Complexity responses of Rhododendron species to climate change in China reveal their urgent need for protection

Kun-Ji Li, Xio-Fei Liu, Jin-Hong Zhng, Xiong-Li Zhou, Liu Yng, Shi-Kng Shen,＊

a Yunnan Key Laboratory of Plant Reproductive Adaptation and Evolutionary Ecology,Ministry of Education Key Laboratory for Transboundary Ecosecurity of Southwest China, Institute of Biodiversity, School of Ecology and Environmental Science, Yunnan University, Kunming, 650504, Yunnan, China

b Institute of International River and Eco-security, Yunnan University, Kunming, 650504, Yunnan, China

Keywords:

ABSTRACT Global climate change has been widely recognized as important factors that threaten biodiversity.Rhododendron species are not only famous woody ornamental plants worldwide but are also indispensable components in alpine and subalpine vegetation in southwest China.However, the geographical distribution ranges response of this broad taxonomic group to future climate change remains not be fully understood.Herein, we studied the impact of climate change on the distribution of Rhododendron species in China by predicting the changes in their suitable habitats, centroid, and species richness under three climate change scenarios (SSP1-2.6, SSP2-4.5 and SSP5-8.5)in the 2090s.The species richness changes of Rhododendrons along altitude were also evaluated.In addition, we calculated the phylogenetic signals of distribution response to climate change.We found that the distribution responses of Rhododendron to climate change have weak phylogenetic signals.In the 2090s, the suitable habitats of about 87%of Rhododendron species will be reduced,77%of Rhododendron species are manifested as northward migration.The high species richness of Rhododendrons tends to migrate to transboundary areas with high altitudes in China.Some Rhododendron species with no concern previously should be taken seriously for their high risk of habitat loss under climate change.Thus,the urgent protection of Rhododendron species under climate change need to be paid more attention than previous acknowledged.We recommend carrying out the reintroduction of endangered species in future suitable habitat,strengthening the protection of transboundary areas with high species richness, and focusing on species with few concerns previously.

1.Introduction

Rapid climate change will alter the structure and function of global marine, freshwater, and terrestrial ecosystems (Parmesan, 2006; Ekl?f et al., 2012).It may also affect plant species diversity by affecting community composition, species distribution, and richness and will become one of the main threats to species diversity (Pereira et al., 2010).Although species can gradually adapt to new climate conditions through evolution, the evolution rate of species in niche change is much lower than that in climate change (Quintero and Wiens, 2013), which means that species will have difficulties in coping with the negative effects by climate change through their evolution.Climate change has been considered the main factor affecting the geographical distribution of plants as early as half a century ago(Parmesan,2006;Zhang et al.,2021).The patterns and rates of changes in the geographical distribution of different species under the influence of climate are different, which are determined by differences in the internal characteristics of species,tolerance to environmental stress, and external drivers (Pompe et al.,2008; Yu et al., 2019).Furthermore, plant groups with special geographical characteristics, such as plants in high altitudes, may be more sensitive to climate change.

High-altitude regions are more sensitive to climate change,and their warming rate is higher than that of low-altitude regions (Pepin et al.,2015).This rapid change in temperature will pose a great threat to alpine plant species and plant communities.The populations originally distributed in high-altitude areas or mountaintops may fall into a dilemma of having no suitable area to migrate under climate change,which will lead to the serious reduction of distribution range or even extinction (Hickling et al., 2005; Wilson et al., 2005).Moreover, the plants distributed in the low- and middle-altitudes of hillsides tend to migrate to high altitudes to cope with global warming (Penuelas and Boada, 2003; Lenoir et al., 2008), and their habitats will overlap with

plants at higher altitudes.Although high-altitude plants also can migrate upward,this migration may lead to interspecific competition due to the slow migration rate (Rumpf et al., 2018; Rana et al., 2021).Previous evidence has shown that the species composition of alpine plant communities is expected to change dramatically in the future(Dullinger et al.,2012; Hülber et al., 2016), which means that the community structure,species diversity, and interaction among species in these areas will correspondingly change.

The Rhododendron genus, which is widely distributed from the northern temperate region throughout southeast Asia to northeastern Australia (Shrestha et al., 2018; Zhang et al., 2021) and composed of more than 1,000 species globally.Rhododendron species are not only famous woody ornamental plants worldwide but are also indispensable components in the alpine and subalpine vegetation in southwest China that is one of global biodiversity hotspots (Liang et al., 2018; Shrestha et al.,2018).This genus has irreplaceable economic and ecological values in the biodiversity and the stability and sustainability of an ecosystem,especially for forest ecosystems (Zhang et al., 2021).Therefore, understanding the response of Rhododendron species to climate change will not only important to the germplasm resource conservation but also crucial for the stability maintains of alpine forest ecosystems.The response of Rhododendron distribution to climate change has attracted some attentions in recent years.These studies focus on the distribution changes of some Rhododendron species to formulate targeted protection measures(Giriraj et al., 2008; Yu et al., 2017, 2019, 2021; Veera et al., 2019; Lu et al., 2021; Zhang et al., 2021).However, most of the previous studies were based on representative concentration pathway (RCP) climate scenario with limited prediction period.The newest downscaled future climate projection dataset that predicts the climate in the 2090s(2080–2100)was provided by WorldClim(https://www.worldclim.org/))in 2020.The dataset is based on the scenario of shared social economic pathways(SSPs)and includes the impacts of land use and socioeconomic development on climate change in the forecast (Li et al., 2015; Meinshausen et al., 2020), which is more in line with the trend of future climate change.In addition, the future distribution of Rhododendron at high altitudes under the general trend of climate warming and the possibility of greater adverse effects on the distribution of high-altitude plant groups is necessary to evaluate.Therefore,the present study verified the large-scale response of Rhododendron species to climate change in China.We predicted the changes in their suitable habitats, centroid migration distance, and species richness under three climate change scenarios(SSP1-RCP2.6, SSP2-RCP4.5, and SSP5-RCP8.5) in the 2090s.In addition, we calculated the phylogenetic signals of distribution response to climate change.Combined with the trend of Rhododendron species richness along the altitude,we evaluated the survival status of Rhododendron species along the altitude under climate change.We hypothesized that Rhododendron species have complexity responses to climate change due to the wide distribution of this broad taxonomic group.Specially, we aimed to: (1) predict the suitable habitats, centroid migration distance,and species richness response of Rhododendron species along altitude to climate change; and (2) determine the influence of phylogenetic relationship on Rhododendron species’ response to climate change.The results will facilitate our understanding on the effect of climate change to Rhododendron species’ survival and provide insight into the appropriate strategies for their protection and ecological restoration under climate change scenarios.

2.Material and methods

2.1.Distribution data of Chinese Rhododendron species

We collected the distribution data of Rhododendron species from Chinese Virtual Herbarium (CVH, https://www.cvh.ac.cn/index.php,56,476 Rhododendron specimens with picture in total) and Global Biodiversity Information Facility (GBIF, https://www.gbif.org/, 35,685 Rhododendron occurrence site records in total).We removed duplicate records or records without coordinates.Then, we used Google Earth to verify the actual position of the coordinate of all distributed data on the satellite map and removed the data with remarkable errors(i.e.,sites on the herbarium).The distance of each distribution point with less than 2 km was thinned out, and finally the distribution points greater than 14 were selected for prediction to avoid spatial autocorrelation.A total of 142 Rhododendron species with 7,123 geo-referenced records were selected in the present study(Fig.S1).

2.2.Environmental variables

We used four types of data as environmental factors:climate variables,terrain factors, soil variables, and land use/cover pattern.The climate variables were 19 bioclimatic variables for current (averages for years 1970–2000) climatic conditions from the WorldClim 2.1 database (Fick and Hijmans, 2017) with a 2.5 arc-minute (～5 km) resolution.Future climate data under three scenarios including SSP1-RCP2.6 (SSP126),SSP2-RCP4.5 (SSP245), and SSP5-RCP8.5 (SSP585) in 2090s were also obtained from WorldClim.The topographic variables used the slope and aspect extracted from the SRTM 90 m DEM digital altitude database(https://srtm.csi.cgiar.org/)using ArcGIS 10.2.Soil data was downloaded from the Harmonized World Soil Database (Nachtergaele et al., 2010).Topsoil pH(t-pH),topsoil clay fraction(t-clay),and topsoil organic carbon(t-oc) were selected as the soil data in the current and future period(Merow et al.,2014).The current and future land use/cover change data was downloaded from the Geographical Simulation and Optimization System(GeoSOS).Altitude data were not used because it is a covariate for the calculation of global climate data,which has a certain correlation with climate data (Fick and Hijmans, 2017).We calculated the Pearson correlation coefficient among the climate variables to avoid their multicollinearity and repeated the MaxEnt model with default parameters for 5 times for each species to calculate the percentage contribution.The climate variables with a very small contribution rate(<1.0%)were removed,the correlation of the remaining climate variables was compared,and the one with the higher percentage contribution among the two variables with strong correlation(|r|> 0.8)was retained(Li et al.,2015).

2.3.Species distribution modeling

The parameters that most affect the prediction of MaxEnt are feature classes (FC) and regularization multiplier (RM).Six types of FC were available for selection,namely,linear features(L),quadratic features(Q),product features (P), hinge features (H), and threshold features (T)(Merow et al.,2013).The RM can increase the extrapolation ability of the model by controlling the feature type selected for establishing the model to prevent excessive complexity and overfitting (Phillips and Dudík,2008; Elith et al., 2010).The increase in this value will lead to the diffusion of the prediction region and seriously reduce the degree of model fitting;therefore, in this study, RM should not be more than 5.0.The “Wallace” package of R was used to pre-predict the species distribution model(SDM)under the free combination of five feature classes(L,LQ, h, LQH, and LQHP) and 10 regularization multipliers (0.5–5.0 with increments of 0.5),and the complexity of the model under this parameter combination was determined by comparing the Akaike information criterion correction(AICc)value of each model.

The training of the model was performed using the MaxEnt software.FC and RM selected the optimal parameters.The other parameters or output settings were as follows:1)the output file type is an ASCII file in cloglog format,2) the output randomly selected the background points,and 3)the k-fold cross validation method was used for validation.The k value was set to 10, that is, the distribution data was randomly divided into 10 samples,of which nine samples were used as the training subset,and the remaining one sample was used as the verification subset,which is repeated for 10 times.The other parameters were set to default,and the average value of 10 repetitions was used as the final result for subsequent analysis.

We evaluated model performance by following methods: 1) AUC(area under the curve) of the ROC (receiver operating characteristic)analysis(Phillips et al.,2006)and partial-ROC(Peterson et al.,2008).2)TSS (true skill statistics) (Allouche et al., 2006).3) BI (Boyce’s index)(Hirzel et al.,2006).AUC value has a range from 0.5 to 1.0,the models with AUC>0.7 indicates a good fit(Swets,1988).TSS varies between-1 and 1, in TSS value range, 0.6–0.8 is useful, and greater than 0.8 is excellent(Préau et al.,2018).Partial-ROC is a supplementary method for the inaccurate with AUC value causing by the pseudo absence data that the model used to fitting (Lobo et al., 2008; Peterson et al., 2008).The models with partial-ROC >1.5 can be considered as greater predictive power.We used the Niche Toolbox platform (http://shiny.con abio.gob.mx:3838/nichetoolb2/; Osorio-Olvera et al., 2020) for partial-ROC calculations and the parameters setting refer to the study by Sandoval et al.(2020).The Boyce’s index (BI) can measures the extent that model predictions differ from a random distribution of the observed presences across the prediction gradients(Boyce et al.,2002;Hirzel et al.,2006), it varies between -1 and 1.Positive values indicate a model which presents predictions that are consistent with the distribution of presences in the evaluation dataset.The BI was calculated by using ECOSPAT package of R (Di Cola et al.,2017).

2.4.Suitable habitats and centroid migration prediction

We divided the prediction results to a binary and projected them to the North Asia Albers Equal Area Conic projection coordinate system using maximum training sensitivity plus specificity as the existence threshold to calculate the suitable area of Rhododendron species in the current and future time periods and the future changes in the area.The response of the suitable area of Rhododendron species to climate change was evaluated by the change ratio of the suitable area and centroid migration distance in the 2090s.In addition, the migration ability of plants determines whether they can migrate to new suitable areas, but the migration ability is difficult to quantify.Therefore, the contraction ratio of current suitable area in the future was additionally calculated as an indicator of the reduction degree of original Rhododendron habitats.

2.5.Changes in suitable habitats and species richness at different altitudes

According to the altitude range of Rhododendron species distribution recorded in the Flora of China (1999), the Rhododendron species that have been predicted for suitable areas are sorted.The Rhododendron species only distributed at an altitude of 1,500 m and below were included in the low-altitude group (La), and the Rhododendron species only distributed at an altitude of 3,100 m and above were regarded as the high-altitude Rhododendron group (Ha) with 10 species in each group(Table 1).According to the characteristics of each sample data, t-test or nonparametric Mann Whitney U test were selected to test the significance of the suitable area and their future.

The species richness of Rhododendron can be obtained by calculating the number of unique species in a pixel and dividing them into five levels:rare(1–25),low(26–50),medium(51–75),high(76–100),and very high(>100).We extracted these values and fitted them with polynomials to obtain the change trend of Rhododendron species richness along the altitude.

2.6.Phylogenetic signal determination

The phylogenetic tree of Rhododendron species was constructed using the sequence composed of ITS, matK, rbcL, psbA, and trnH (Fu et al.,2022).A search of the National Center for Biotechnology Information yielded 93 Rhododendron species with the above four DNA sequences(Table S1), Then, MEGA11 was used to construct the maximum likelihood tree (Fig.S2).We calculated Bloomberg’s K as the phylogenetic signal using the“Picante”package(Kembel et al.,2010).We also tested its significance to verify the strength of phylogenetic signals.

3.Results

3.1.Model evaluation and factors that determine distribution

The MaxEnt models of Rhododendron species were all with AUC values above 0.9(Fig.S3)and significant(p<0.001)partial-ROCs above 1.8 (Table S2).The Boyce’s index were all above 0.6 except four Rhododendron species(Table S2).The TSS were between 0.62 and 0.99,and 70% of them (99) were above 0.8 (Table S2), indicating that these prediction results are credible.

Among all environmental variables, land use/cover pattern, isothermality (bio3), slope, temperature seasonality (bio4), t-pH, mean temperature of warmest quarter (bio10), t-clay, and precipitation seasonality (bio15) of more than 90 Rhododendron species affect the distribution of species to varying degrees (Fig.S3).Among the climatic variables,bio3,bio4,bio10,and bio15 affect the distribution of 122,105,99, and 91 Rhododendron species, respectively, with average percent contributions of 12%, 14.53%, 7.79%, and 3.72%, respectively.The average percent contribution of land use/cover types in all species is 12.09%.Among soil variables, t-pH and t-clay affect the distribution of most Rhododendron species (102 and 99, respectively), but the average percent contributions are 1.71% and 1.05%, respectively.Among the topographic variables, slope affected the distribution of 121 Rhododendron species with an average percent contribution of 7.11%(Fig.S3).

3.2.Changes in suitable habitats of Rhododendron species

The suitable areas of most Rhododendron species will be reduced in the 2090s(Fig.1a).The contraction ratio ranges of the original suitable habitats were 0%–53.090%, 0.003%–69.040%, and 0.003%–69.040%under the SSP126, SSP245, and SSP585, respectively (Fig.1b).The centroid migration distances were 1.00–283.87, 1.62–412.62, and 5.13–1149.62 km,in which 106,108,and 112 Rhododendron species will migrate to the northwest under the SSP126, SSP245, and SSP585,respectively (Fig.2).The change in the area of suitable habitats, thecontraction ratio and the migration distance increased significantly with the increase in greenhouse gas emissions in different climate change scenarios(Figs.1 and 2,P<0.05).

Table 1 List of Rhododendron species in high- and low-altitude groups.

Fig.1.Change ratios of suitable habitats (a) and original suitable habitats (b) of Rhododendron species at three future scenarios and their variance analysis (c).

3.3.Changes in species richness of Rhododendrons

The current high Rhododendron species richness distributed in Yunnan,Sichuan,Tibet,Guizhou,and Taiwan provinces.The Hengduan Mountain Range is a concentrated area with the highest species richness.Sporadic areas with high Rhododendron species richness also exist at the border between Tibet and neighboring countries.The species richness showed a whole decreasing trend from Southwest to East and North in China(Fig.S4a).

The area of each species richness category has the same change trend under the three climate change scenarios in the 2090s.Except the rare species richness, all other categories of Rhododendron species richness will face the decreasing suitable habitat(Table S3).The change extension increased with increasing greenhouse gas emissions.The areas with low Rhododendron species richness in South, East, and Central China almost completely disappeared under SSP585, leaving only the areas with rare species richness.New areas with rare species richness will arise in Inner Mongolia,Heilongjiang,Jilin,Qinghai and Xinjiang of China(Fig.S4b).The areas with a considerable increase in Rhododendron species richness exist in Northern Sichuan, Southern Tibet, and Western Tibet of China.The areas with a considerable reduction in richness were the Sichuan Basin, Yunnan, Guizhou, Chongqing, and other places, whereas the decline in species richness in Hubei and Southeast China is also serious(Fig.S4b).

3.4.Changes in Rhododendron species richness along the altitude

We used polynomial to fit the Rhododendron species richness at each altitude degree and obtained the variation curves of Rhododendron species richness along the altitude at the present and in the future.The fitting curve and scatter diagram of the observed values show that the Rhododendron species richness shows a single peak distribution along the altitude at the present and in the 2090s(Fig.3),that is,the species richness of Rhododendron initially increases and then decreases with the increase in altitude.The peak of the curve in each future climate scenario is lower than that of the current period, and the altitude reaching the peak of species richness is higher than that of the current period(Fig.3).A higher greenhouse gas emission is associated with a lower peak value of Rhododendron species richness and a higher altitude of reaching the highest Rhododendron species richness under the three climate change scenarios.In the altitude range below the peak,the species richness at the same altitude is lower in the 2090s under the three climate change scenarios than that of the current period.

3.5.Different responses of high-and low-altitude Rhododendron species to climate change

The prediction results of the suitable habitats of Rhododendron species distributed in different altitude showed obvious differences in geographical location, which were consistent with the geo-referenced records of Rhododendron species (Fig.S5).The current distribution areas of the HA group include the Hengduan Mountains, Eastern Himalayas, Northwest Yunnan, Western Sichuan, and other plateau areas.While,the current distribution areas of the LA group include coastal and plain areas, such as Guangdong, Guangxi, Fujian, Zhejiang, Hunan,Jiangxi,Taiwan,and Hong Kong.The distribution areas of Rhododendron species at low altitude were considerably larger than that of Rhododendron species at high altitude(Fig.4).

Fig.2.Centroid migration distance of Rhododendron species (a) and the direction and distance of centroid migration in three future climate scenarios: SSP126 (b),SSP245 (c), and SSP585 (d).The direction of the middle sector is the migration direction, and the size of the sector represents the number of Rhododendron species migrating in this direction.The proportion of different colors that make up the fan represents the number of species with the migration distance represented by that color.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Three Rhododendron species from the LA group have expanded their suitable areas under all climate scenarios in the 2090s: R.henryi (La1),R.mariae (La5), and R.seniavinii (La7).In the HA group, R.nitidulum(Ha3), R.sphaeroblastum (Ha6), and R.rufescens (Ha9) will reduce the area of the suitable area under SSP126 but will increase under the other two climate change scenarios with more warming.The Rhododendron species distributed at different altitudes do not show remarkable differences in the change ratio of suitable habitats, except SSP126.However,the contraction of the original suitable area of the HA group was remarkably greater than that of the LA group(Fig.5).

3.6.Phylogenetic signals of species’distribution response to climate change

The Bloomberg’s K values of all suitability change-related indexes were less than 1.The phylogenetic signals of these indexes were substantial,except the centroid migration of SSP585 and the change ratios of suitable areas in SSP245 and SSP585 (Table S4).The results imply that the distribution responses of Rhododendron to climate change are affected by external environmental factors.

4.Discussion

4.1.Factors affecting the distribution of Rhododendron species

Our results showed that bio3, bio4, bio15, and bio10 are the most important climatic factors affecting Rhododendron distribution.Besides climatic factors,land use/cover pattern and slope also play an important role in Rhododendron distribution.In addition to the direct impact of the overexploitation of the natural environment on the fragmentation of local biological habitats and the reduction of biodiversity,the change of land use pattern will also have a sustained impact on hydrology,climate,energy exchange,and other aspects at different regional scales and affect the distribution pattern of species(Marshall et al.,2021).Rhododendron,as a constructive species of vegetation in alpine, subalpine, and middle mountain zones,is mostly distributed in mountainous areas with rugged terrain.On the landscape scale, the slope will remarkably affect the climatic conditions, resource utilization, and the diversity of species composition on the community scale (Simonson et al., 2014).Soil properties have a great impact on plant distribution and have been used in the prediction of plant distribution (Morin and Lechowicz, 2013;Pannek et al.,2013).In this study,t-pH and t-clay had weak impacts on the distribution of most Rhododendron species (more than 99 species),and their percent contributions were small.This supports that the impacts of climate and land use change on species distribution are greater than that of soil properties at large-scale(Morin and Chuine,2006).The distribution of high- and low-altitude Rhododendron species are all affected by the precipitation of driest quarter (Fig.S6), which may be related to the generally poor drought resistance of Rhododendron species(Li et al.,2022).Drought stress may decrease the seed germination rate(Li et al., 2015) and flowering rate of Rhododendron species (Morimoto et al., 2011), thus affecting their population regeneration.The distribution of high-altitude Rhododendron species is mainly determined by four temperature related climate variables (isothermality temperature seasonality, mean temperature of warmest/coldest quarter) (Fig.S6), the result is consistent with the other studies of alpine plants which are sensitive to temperature changes (Simonson et al., 2014; Veera et al.,2019).

Among the nine indicators used to measure phylogenetic signals,six indicators could detect significant phylogenetic signals (P < 0.05),whereas all Blomberg’s K values were very small(<1).This result means that the distribution response of Rhododendron species to climate change is affected by external environmental factors rather than their evolutionary process(Blomberg et al.,2003).

Fig.3.Altitudinal variation of Rhododendron species richness under the current condition and three future scenarios in China(a).Comparison of species richness and altitudinal curves of Rhododendron species at the present and in the 2090s(b).The intersection of the dotted line with the Y axis represents the peak of richness,and the intersection with the X axis represents the altitude at the peak.Dotted lines in different colors represent different periods.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

4.2.Distribution and species richness patterns of Rhododendron species and their future changes

Our result showed that the suitable habitats of Rhododendron species in China are concentrated in Yunnan, Sichuan, Guizhou, Tibet, and the Hengduan Mountains, which is consistent with the previous study (Yu et al., 2017).These areas have a warm, humid, and slightly changed climate since the Cretaceous period,which is considered as the modern origin and diversity evolution center of Rhododendron(Xia et al.,2022).However,most Rhododendron species will migrate to the northwest(high latitude) under different climate change scenarios in the 2090s.The change pattern of species richness also shows that the areas where the Rhododendron species richness increases in the future are in the northwest (i.e., Northern Sichuan, Southern Tibet) of the areas with high species richness (i.e., Hengduan Mountains) in the current period,whereas the areas where the species richness decreases are in the southeast (i.e., Southern Sichuan, Yunnan, Guizhou).In the 2090s, the area of Rhododendron species with low richness level (>25) and above will decrease, and the area of rare richness level (<26) will increase naturally.The species richness of Rhododendron species in low- and medium-altitude areas below the peak altitude is also expected to decrease.These results are consistent with Yu et al.(2021), who predicted the distribution and species richness of endemic Rhododendron species to China in the 2070s.

Fig.4.Potential distribution patterns of Rhododendron species in the HA and LA groups (a) and the current suitable area (b).

The present study showed that, regardless of the greenhouse gas emission level, more than 120 Rhododendron species will lose their suitable area (nearly 87% for 142 Rhododendron species) by the 2090s.This is more pessimistic compared with the previous study (Yu et al.,2021).The difference may be caused by the difference in the future climate projection data set.Compared with RCP,the temperature of SSP is expected to rise to a greater extent.Our study also showed that the increase in the magnitude of climate change will lead to more drastic changes in the distribution of Rhododendron plants, and this change is often negative(i.e.,increase in the contraction ratio of suitable habitat).Thus, these pessimistic predictions suggest that we should be more careful about the response of Rhododendron species to climate change and formulate protection measures for species vulnerable to climate change in advance.

Fig.5.Change ratios of the suitable area (a), original habitats (b), and centroid migration (c) of Rhododendron species in the HA and LA groups and their variance analysis under the three future scenarios.

4.3.Dilemma of high-altitude Rhododendron species

Alpine regions are one of the most sensitive regions to climate change,and species distributed in high mountains may have narrow niches due to the strong local characteristics of highland ecosystems and their small population size (Beniston, 2006).In the present study, the suitable habitats of Rhododendron at high altitude were remarkably smaller than that at low altitude, which also supports this view.Although no remarkable differences were found between the high-altitude and low-altitude Rhododendron species in the change ratio of suitable habitats in the 2090s, the contraction ratio of the original suitable habitats of Rhododendron at high altitude will increase to a greater extent,whereas new suitable habitats will appear in farther areas, which means that migration to a new suitable area will be extremely hard.Rhododendron species, which are widely distributed along the altitude, may have greater tolerance to climate change,and plants distributed on the hillside will also have an upward migration to cope with climate change(Lenoir et al., 2008).However, for Rhododendron species group in the high-altitude areas where the distribution range includes mountaintops,the distribution range may still be greatly reduced because of the insufficient additional high-altitude areas for their migration in the future of climate warming.

In addition, our results showed that Rhododendron species richness may increase in high-altitude areas, which means that the suitability of these areas for Rhododendron may increase.The phenomenon that climate warming increases the suitability of high-altitude areas for plants was also found in previous study (Telwala et al., 2013).The Himalayas and Hengduan Mountains are the main distribution areas of Rhododendron.Recent studies have shown that the composition of plant communities in the Western Himalayas will undergo major replacement in the future,which will be more obvious in high-altitude areas(Rathore et al.,2022),and the main plant communities in Hengduan Mountains will also migrate to high-altitude areas (Liang et al., 2018).Therefore, the Rhododendron species distributed at high altitudes easily face high survival pressure under their narrow niche, high loss risk of original habitats,and competitive pressure.

4.4.Enlightenment on the protection of Rhododendron species

Among the Rhododendron species distributed in China, 194 Rhododendron species are classified as vulnerable, threatened, endangered, or critically endangered (including subspecies and varieties) (Ministry of Ecology and Environment of China, http://www.mee.gov.cn/gkml/h bb/bgg/201309/t20130912_260061.htm).For endangered plants, reintroduction is an important conservation strategy.Reintroduction requires that individual plants be transplanted to new habitats,and the selection of these new habitats needs to take into account many factors, such as ecological characteristics,human influence,and interspecific interaction(such as pollinators) (Hoegh-Guldberg et al., 2008).The historical distribution area of species for reintroduction will be the prioritized choice.However, global warming and land use change may reduce the habitat suitability of these areas.Therefore, areas with high suitability for Rhododendron can be used as the selection of reintroduction site for endangered species.In addition, for woody plants, especially arbors, reintroduction will take decades; therefore, future climate changes in the habitat of the reintroduction site also need to be taken into account.

In our study,the borderline regions with high species richness levels in the southwest of Tibet of China and other countries such as Bhutan,Nepal,and Sikkim have high Rhododendron species richness levels at the present and in the future, which means that their suitability for Rhododendron species will not be greatly affected by climate change.Moreover,their special geographical location (remote and border) will lead to insufficient attention and management, which will make the ecological environment of the region far from human factors (such as land use change)and become a natural refuge for endangered species and a good choice for reintroduction (Wang et al., 2016; Liu et al., 2020).In addition,almost all areas bordering the southwest of Tibet are mountainous areas or natural landscapes with complex terrain without obvious physical barriers; therefore, their ecological connectivity will not be greatly affected (Linnell et al., 2016; Pokorny et al., 2017; Liu et al., 2020).Therefore, we propose to strengthen the conservation of Rhododendron resources in transboundary areas with high species richness through international cooperation.

Some Rhododendron species with higher economic value or ecological values always receive more attention.Other species that are considered unimportant or difficult to study(e.g.,too small or distributed in remote mountainous areas)may be ignored.Whatever the reason for the lack of research on some groups,their neglect may lead to bias in the assessment of large-scale(such as global)biodiversity(Mouillot et al.,2013;Feeley et al.,2017).In our study,R.eclecteum and R.stewartianum are the only two species among the 142 Rhododendron species that will lose their suitable area and original suitable area by more than 50%in all climate scenarios, and research reports on these species are scarce.This finding may be due to the fact that R.eclecteum is categorized as a vulnerable species and R.stewartianum is a least concern species.Both species do not have a higher endangerment level at present.However, the suitable habitats of these Rhododendron species will be easily lost if these species will not gain enough attention in the future because of the worsening of climate change.50 Rhododendron species (including subspecies and varieties) were still classified as data deficient (Ministry of Ecology and Environment of China, http://www.mee.gov.cn/gkml/hbb/bgg/201309/t20130912_260061.htm); data-deficient species have been demonstrated with greater risk of endangerment in the future(Pelletier et al., 2018).Therefore, we recommend to pay more attention to Rhododendron species which received less concern previously.

5.Conclusion

The present study verified the geographical distribution ranges response of Rhododendron species to climate change in China, and predicted the changes in their suitable habitats,centroid migration distance,and species richness under the scenario of shared social economic pathways (SSPs), which is more in line with the trend of future climate change in the 2090s.We also first assessed the influence of the phylogenetic relationship on Rhododendron species’response to future climate change.We found that the distribution responses of Rhododendron to climate change have weak phylogenetic signals.In the 2090s, the suitable habitats of about 87%of Rhododendron species will be reduced,77% of Rhododendron species are manifested as northward migration.Some Rhododendron species with no concern previously should be taken seriously for their high risk of habitat loss under climate change.Owing to complex response of Rhododendron species to future climate change,we recommend paying more attention to species with few concerns previously or with low level of endangerment status, estimating their endangered risk using ecological models and developing conservation strategies in advance based on the estimating results.Some Rhododendron species with high risk of distribution ranges contraction even need urgent rescue through ex situ conservation in botanic garden.

Credit author statement

S–K S conceived the study design,conducted the analyses and revised the manuscript.K-J L conducted the analyses,and wrote the first draft of the manuscript.X–F L,J-H Z,X-L Z,and Y L contributed substantially to revisions.

Funding

This study was supported by the Science and Technology Development Fund of Guidance from the Central Government to Locals in Yunnan Province, China (No.202207AB110016), the Science and Technology Basic Resources Investigation Program of China (No.2022FY100205),National Natural Science Foundation of China (No.31870529), Major Program for Basic Research Project of Yunnan Province, China (No.202101BC070002), and the Program for Excellent Young Talents,Yunnan University,China.

Availability of data

Data are available on request from the authors.

Declaration of competing interest

The authors declare that they have no competing interests.

Appendix A.Supplementary data

Supplementary data to this article can be found online at https://doi.i.org/10.1016/j.fecs.2023.100124.