Rhododendron diversity patterns provide new insights for conserving China's montane flora

ABSTRACT

Under anthropogenic global change, mountain plants are among the most threatened flora, making it essential to address conservation gaps to achieve the “30 × 30” target outlined in the Convention on Biological Diversity. Focusing on Rhododendron, the emblematic genus of montane ecosystems in Asia, we compiled an updated checklist of 603 Rhododendron species in China and integrated high‐resolution distribution, phylogenetic, functional trait, and environmental data to examine spatial diversity patterns, driving factors, and priority conservation areas. Our results indicated that the taxonomic, phylogenetic, and functional dimensions of Rhododendron diversity were highly consistent. The Three‐Parallel‐Rivers Region in the Hengduan Mountains emerged as the overarching hotspot for richness, endemism, and uniqueness. These diversity patterns were collectively influenced by temperature seasonality and elevation range, with temperature‐related energy factors further contributing to the overall phylogenetic and functional structures. The mechanisms underlying uniqueness exhibited regional divergence: In southwestern China, high uniqueness was attributed to the accumulation of intermediate‐aged and younger distinctive lineages, coupled with rapid community differentiation; in southern China, it was mainly associated with strong spatial compositional differentiation among communities; whereas in northeastern China, it largely resulted from the preservation of relict lineages. Under the 10% conservation target, we delineated nine priority conservation units across the southwestern, southern, and northeastern mountainous regions of China. However, when the conservation target was increased to 30%, these priority areas expanded to encompass broader mountain regions, highlighting a greater conservation gap. This study provides innovative insights for incorporating multidimensional diversity into effective conservation efforts for Rhododendron species in China.

Integrating distribution, phylogenetic, and functional trait data for 603 Rhododendron species in China identified significant conservation hotspots of multidimensional diversity, particularly in the Hengduan Mountains. Climate seasonality and topographic heterogeneity jointly influenced these patterns; however, notable conservation gaps remained, especially under the 30 × 30 conservation target.

INTRODUCTION

Since the onset of the Industrial Revolution, human activities have altered the Earth's ecosystems on an unprecedented scale and intensity (Dirzo et al., 2014), resulting in profound and multifaceted negative impacts on global biodiversity and accelerating species extinction rates far beyond the natural background levels (De Vos et al., 2015; Ledford, 2019). In response to this critical challenge, the Kunming‐Montreal Global Biodiversity Framework has established ambitious conservation targets, aiming to protect at least 30% of the world's terrestrial, inland water, and coastal and marine areas by 2030 (CBD, 2022). Achieving this goal requires the urgent identification of priority conservation areas and implementation of efficient, coordinated conservation strategies (Kling et al., 2019; Lu et al., 2023).

Biodiversity, which encompasses all forms of life on Earth, is inherently multidimensional and scale‐dependent (Purvis and Hector, 2000). The multidimensionality manifests in various aspects, including taxonomy, phylogeny, genetics, and functional traits (Liao et al., 2025). Specifically, taxonomic diversity (species richness), phylogenetic diversity, and functional diversity correspond to the total number of species, the accumulation of evolutionary history (Faith, 1992), and trait variability within a specific region, respectively (Petchey and Gaston, 2006). Endemism highlights the degree to which phylogenetic or functional diversity is confined to specific areas (Rosauer et al., 2009), while threat assessment focuses on the diversity patterns of endangered species (Callmander et al., 2007). Due to ecological and evolutionary constraints, the spatial distribution patterns of these diversity dimensions may be incongruent (Guo et al., 2022; Montaño‐Centellas et al., 2023). For example, global centers of plant functional diversity, particularly for herbaceous plants, showed markedly distinct spatial patterns compared to taxonomic and phylogenetic diversity (Luo et al., 2024). Lauraceae showed discrepancies among functional diversity, evolutionary distinctiveness and global endangerment (EDGE), and other diversity indices (Zhou et al., 2023). Therefore, relying solely on taxonomic diversity metrics is insufficient for effectively conserving evolutionary history and variability of functional traits (Mazel et al., 2014; Brum et al., 2017; Luo et al., 2024). Moreover, prevailing conservation strategies often prioritize regions of high diversity while overlooking the distinctiveness of regional species assemblages. Community uniqueness (Baselga et al., 2025), based on beta diversity (Luukkonen et al., 2024), is typically assessed by calculating the mean dissimilarity between a focal region and others (Mokany et al., 2022; Tsang et al., 2023), reflecting a region's irreplaceability in environmental space. Thus, it constitutes a critical dimension in the planning and execution of biodiversity conservation initiatives. In recent years, researchers have developed various biodiversity indices, such as μ‐diversity (Tordoni et al., 2024), functionally rare, evolutionarily distinct, globally endangered species (FREDGE) (Cui et al., 2024), and the conservation gap index (Zhang et al., 2023b), each with a different research focus. While these indices have greatly enhanced conservation planning, the growing number of indices has also complicated the identification of priority areas for conservation. Selecting the most appropriate diversity index for specific conservation objectives has thus emerged as an urgent and complex challenge.

Rhododendron is the largest genus of woody plants in the Northern Hemisphere, comprising over 1,000 species worldwide, primarily distributed in alpine and subalpine regions of the North Temperate Zone (Chamberlain et al., 1996; Shrestha et al., 2018b). China hosts the world's highest species diversity of the genus, with approximately 600 recognized species. The Himalaya‐Hengduan Mountains serve as a major center for the contemporary distribution, species diversity, and diversification of Rhododendron (Shrestha et al., 2018b). Furthermore, the number of species continues to increase as new taxa are consistently discovered (e.g., Liu et al., 2022; Chang et al., 2023). However, climate change and intensified human activities are progressively degrading the habitats of Rhododendron species, putting some at risk of survival (Ma et al., 2014). According to the China Biodiversity Red List—Higher Plants (2020) (Ministry of Ecology and Environment, Chinese Academy of Sciences, 2023), 18% of Rhododendron species in China are categorized as threatened, while 20% remain unevaluated due to data deficiency. Previous conservation assessments of Rhododendron in China have predominantly emphasized taxonomic aspects, with limited integration of phylogenetic and functional dimensions (Yu et al., 2017; Shrestha and Wang, 2018; Zhao et al., 2023b). For instance, Shrestha and Wang (2018) utilized a complementary algorithm to identify priority conservation areas for 556 species, and highlighted the importance of regions in southern and southeastern China, beyond the Hengduan Mountains. Zhao et al. (2023b) identified hotspots for 546 wild Chinese Rhododendron species based on species richness and endemism. However, unidimensional approaches may not adequately capture the full scope of biodiversity, potentially leading to insufficient or ineffective conservation efforts for Rhododendron species.

Understanding the patterns of diversity and the key drivers influencing Rhododendron species is essential for formulating effective, scientifically sound, and forward‐looking conservation strategies. The diversity observed within Rhododendron arises from the interconnection and synergistic effects of multiple factors (Shrestha et al., 2018a; Guan et al., 2024). Among the various hypotheses proposed to elucidate biodiversity patterns, environmental factors are often regarded as key determinants (Hawkins et al., 2003), particularly at local scales and over short time periods. These factors include energy availability, climatic seasonality, habitat heterogeneity, and water availability. Specifically, the energy hypothesis posits that energy availability is crucial for sustaining diversity gradients (Currie, 1991). The habitat heterogeneity hypothesis proposes that greater habitat heterogeneity fosters species coexistence (Stein et al., 2014; Stein et al., 2015). In addition, the climatic stability hypothesis highlights the role of both contemporary and historical climate stability in facilitating species accumulation (Stevens, 1989). Furthermore, current patterns of diversity are closely tied to evolutionary history. High species richness is associated with regions rich in ancient lineages that may have undergone prolonged periods of effective diversification, and regions with elevated diversification rates driven by accelerated speciation processes (Rohde, 1992; Chown and Gaston, 2000).

Previous studies have made substantial advancements in investigating the relationships between species richness and environmental factors in Rhododendron (Shrestha et al., 2018a; Xia et al., 2022; Zhao et al., 2023b; Guan et al., 2024). However, differing research perspectives have resulted in different conclusions. For instance, Guan et al. (2024) found that mountain area and elevation range were the strongest factors explaining Rhododendron diversity patterns, while Xia et al. (2022) identified elevation and annual precipitation as the key ecological drivers of global species richness. In China, Shrestha et al. (2018a) demonstrated that habitat heterogeneity and climatic seasonality were the most significant predictors of species richness.

Despite these advancements, current research on Chinese Rhododendron faces several limitations. The baseline resources for Rhododendron are incomplete, as most studies have not systematically incorporated recently described species. Additionally, phylogenetic relationships remain inadequately resolved, with existing studies covering less than half of the known Chinese Rhododendron species and often neglecting functional traits. These gaps impede a mechanistic understanding of diversity patterns from both phylogenetic and functional perspectives. Furthermore, the influence of human activities on Rhododendron distribution remains poorly understood.

In this study, we compiled a comprehensive checklist of Rhododendron species in China. By integrating multidimensional datasets, including distribution data, phylogenetic relationships, key functional traits, and environmental and anthropogenic variables, we mapped diversity patterns, identified biodiversity hotspots and priority conservation areas, and elucidated the key drivers underlying these patterns for 603 species in the genus. Specifically, our study aims to address the following questions: (1) What are the spatial patterns of taxonomic, phylogenetic, and functional diversity of Rhododendron in China? (2) What factors predominantly shape these patterns? (3) Which regions should be prioritized for conservation, and what corresponding strategies are recommended? By focusing on Rhododendron as a representative group for mountain ecosystems, our findings provide a scientific basis for understanding the maintenance of mountain plant diversity and offer both theoretical and practical guidance for developing regional and national conservation strategies.

RESULTS

Spatial patterns of Rhododendron diversity in China

The diversity patterns of Rhododendron species exhibited broad consistency across taxonomic, phylogenetic, and functional dimensions. The mountainous regions of southwestern China served as the core areas of diversity, with the Hengduan Mountains, particularly the Three‐Parallel‐Rivers Region (approximately south of 29°N), harboring the highest levels of diversity, endemism, and uniqueness (Figure 1). Correlations among metrics within the same category were stronger than those between different categories. Specifically, diversity metrics (SR: species richness, PD: phylogenetic diversity, FD: functional diversity) displayed highly significant positive correlations (r = 0.86–0.96), as did endemism metrics (WE: weighted endemism, PE: phylogenetic endemism, FE: functional endemism) (r = 0.80–0.95) and uniqueness metrics (PU: phylogenetic uniqueness and FU: functional uniqueness) (r = 0.91) (Table S1). However, the phylogenetic dimension displayed partially distinct patterns. For instance, high PE values were found not only in the southern Hengduan Mountains but also in the Changbai Mountains of northeastern China (e.g., R. redowskianum, representing a basal lineage of Rhododendron with localized distribution), whereas both WE and FE were low in the latter region. For the 268 threatened species, the Hengduan Mountains exhibited the highest diversity, followed by southeastern Yunnan (Figure S1). A highly positive correlation was also observed among the six diversity metrics (SR, PD, FD, WE, PE, FE; r = 0.61–0.94) (Table S2).

Figure 1.

Spatial patterns of Rhododendron diversity metrics in China

(A) Species richness (SR). (B) Phylogenetic diversity (PD). (C) Functional diversity (FD). (D) Weighted endemism (WE). (E) Phylogenetic endemism (PE). (F) Functional endemism (FE). (G) Evolutionary distinctiveness and global endangerment (EDGE). (H) Phylogenetic uniqueness (PU). (I) Functional uniqueness (FU).

Both sesPD (standardized effect size of phylogenetic diversity) and sesMPD (standardized effect size of mean pairwise phylogenetic distance) indicated significant phylogenetic clustering (negative values) in the eastern Himalayas and Hengduan Mountains, whereas phylogenetic overdispersion (positive values) was observed in northeastern, central, and southern China (Figure S2A, B). Regions such as the Changbai Mountains showed pronounced phylogenetic overdispersion. Although functional structure in the Hengduan Mountains showed a tendency toward clustering, the extent of clustering was not as widespread as that at the phylogenetic level (Figure S2C, D). This discrepancy was likely related to the lack of significant phylogenetic signals in certain traits, such as leaf shape (Table S3). Significant functional overdispersion was detected in Yunnan and Guangxi.

The CANAPE analysis identified 246 grid cells as centers of endemism for Rhododendron species in China (Figure 2). Paleo‐endemism centers were primarily located in northeastern China, including northern Nei mongol, and eastern Jilin and Liaoning. Mixed and super endemism centers were largely concentrated in the eastern Himalayas, the southern Hengduan Mountains (especially the Three‐Parallel‐Rivers Region), southeastern Yunnan, and Taiwan. Neo‐endemism centers were scattered across northwestern Guizhou, southern Qinghai, and southern Shaanxi, with the fewest grid cells identified. Further analysis of the net diversification rates and divergence times across different types of endemism centers revealed that paleo‐endemism centers were characterized by significantly lower diversification rates and significantly older divergence times (Figure 2D, E). In contrast, neo‐endemism centers showed significantly higher diversification rates and more recent divergence times. Mixed and super‐endemism centers displayed significantly higher diversification rates, coupled with intermediate divergence times.

Figure 2.

Endemism centers and diversification patterns of Rhododendron in China

(A) Centers of phylogenetic endemism identified using CANAPE (categorical analysis of neo‐ and paleo‐endemism). (B, C) Spatial patterns of mean net diversification rate (B) and mean divergence time (C) across different grid cells. (D, E) Statistical comparisons of mean net diversification rate (D) and mean divergence time (E) among different types of endemism centers using the Kruskal–Wallis test. In (D, E), the center line indicates the median, the box indicates the interquartile range (IQR), and the whiskers extend to 1.5 × IQR.

Both PU and FU metrics identified southwestern China—encompassing the Yunnan‐Guizhou Plateau, the Hengduan Mountains, and the eastern Himalayas (Figure S3B)—as the regions with the highest uniqueness values (Figures 1H, I, 3A, B). This was followed by southern China (south of the Qinling–Huaihe line, the western edge of the Sichuan Basin, and the eastern part of the Yunnan–Guizhou Plateau), whereas the lowest levels of uniqueness were observed in northeastern China, with the exception of the Changbai Mountains. The phylogenetic and functional uniqueness‐distance increase model further revealed distinct biogeographic mechanisms underlying these patterns: The high uniqueness in southwestern China was associated with both steep slopes and high intercepts (Figure 3C, D). In southern China, uniqueness was primarily driven by steep slopes, whereas in northeastern China, it was influenced by intercepts.

Figure 3.

Spatial patterns of the uniqueness‐distance model for Rhododendron in China

(A, B) Bivariate maps showing the spatial distribution of the slope and intercept in the uniqueness‐increase model for phylogenetic level (A) and functional level (B). (C, D) Plots illustrating the increase in Ruggiero's phylogenetic (C) and functional (D) dissimilarity, respectively, with distance from the focal cells in three regions (southwestern China, southern China, and northeastern China), along with the corresponding fits of the distance‐increase models.

Factors influencing Rhododendron diversity patterns

The GAM models identified climatic seasonality (temperature seasonality, TSN) as the primary explanatory variable for PU and FU, ranking second for SR and WE, and third for sesMPD and sesMFD (Figure 4). The heterogeneity variable (elevation range, ELER) was the foremost variable for SR and WE, and also played a critical role in influencing PU and FU. Energy variables, MTWQ (mean temperature of warmest quarter) and MAT (mean annual temperature), were the leading predictors of sesMPD and sesMFD, respectively.

Figure 4.

Effects of environmental factors on multiple dimensions of Rhododendron diversity in China

(A–F) Relative importance of explanatory variables in the GAM models and standardized coefficients in the SAR_err models for species richness (A), weighted endemism (B), standardized effect size of mean pairwise phylogenetic distance (C), standardized effect size of mean pairwise functional distance (D), phylogenetic uniqueness (E), and functional uniqueness (F). For the SAR_err models, statistical significance was assessed using z tests. ***P < 0.001; **P < 0.01; *P < 0.05. Abbreviations of explanatory variables are defined in Table 1; the main abbreviations used in this figure are as follows: AET, actual evapo‐transpiration; AI, arid index; ART, annual air temperature range; ELER, elevation range; MAT, mean annual temperature; MTCQ, mean temperature of coldest quarter; MTWQ, mean temperature of warmest quarter; PET, potential evapotranspiration; PSN, precipitation seasonality; SDI, Shannon diversity of enhanced vegetation index; TCL, temperature anomaly; TSN, temperature seasonality.

The SAR_err analyses further indicated that TSN and ELER were the significant explanatory variables of the overall patterns (Figure 4). Specifically, TSN had negative effects on SR, WE, PU, and FU, whereas ELER demonstrated significant positive effects on these metrics. MTWQ and MAT were positively correlated with sesMPD and sesMFD, respectively. Temperature anomaly had no significant effect on any diversity metrics. For threatened species, ELER and TSN likewise emerged as the primary variables influencing species richness and endemism (Figure S4).

Priority conservation areas for Rhododendron species in China

We focused on the indicators with the fastest cumulative rates at both the phylogenetic and functional levels, integrating these with the uniqueness‐distance increase model. This approach identified two types of regions, “islands” and “hills” (Baselga et al., 2025), at both levels. Simultaneously, the priority conservation areas we delineated encompassed the minimum area required to protect all threatened species.

The accumulation rates of PD and FD varied across diversity metrics. PE facilitated the fastest accumulation of PD for both total and threatened species, while FE and WE resulted in the steepest increases in FD (Figure 5A, D). This phenomenon was associated with the strong correlation between WE and FE (r = 0.95). When the top 10% of priority conservation areas were targeted for planning, at the phylogenetic level, evolutionary islands were primarily located in the Hengduan Mountains and the Changbai Mountains, whereas evolutionary hills were mainly located in Guizhou, southern Hunan, and northern Guangxi (Figure S5A). Under the 17%, 25%, and 30% thresholds, the spatial overlap between evolutionary islands and evolutionary hills in the Hengduan Mountains increased further (Figure S5C, E, G). Functional analyses based on WE and FE yielded highly consistent results (Figure S6), with functional islands and hills overlapping extensively in the Hengduan Mountains and the Nanling Mountains, and additional islands scattered in southeastern Yunnan and Taiwan (Figure S5B).

Figure 5.

Diversity accumulation curves and minimum area analysis for the conservation of Rhododendron species in China

(A, B) Accumulation of phylogenetic diversity for all species (A) and threatened species (B) under different prioritization criteria (EDGE, evolutionary distinctiveness and global endangerment; FD, functional diversity; FE, functional endemism; FU, functional uniqueness; PD, phylogenetic diversity; PE, phylogenetic endemism; PU, phylogenetic uniqueness; SR, species richness; WE, weighted endemism). (C, D) Accumulation of functional diversity for all species (C) and threatened species (D). (E, F) The minimum area required to protect all species (E) and threatened species (F) based on the complementarity algorithm. Legends for (E, F) show grid cells containing the top 2.5%, 2.5%–5%, 5%–10%, 10%–17%, and 17%–30% of species or threatened species.

The complementarity analysis identified 94 grid cells that encompassed all Rhododendron species, accounting for 4.43% of their total distribution range. The grid cells with the highest species coverage were predominantly located in the Hengduan Mountains (Figure 5E, F). A total of 72 grid cells covered all threatened species, with the highest concentrations occurring in the Hengduan Mountains and southeastern Yunnan.

Overall, we integrated multidimensional diversity hotspots, scale‐dependent uniqueness (hills vs. islands), endemism centers, and conservation risk (threatened species) to delineate four types of priority conservation areas with different emphases (Figure 6A): Type I (composite hotspots), represented by the eastern Himalayas and Hengduan Mountains (region ① in Figure 6A) and the Daliang Mountain–Wumeng Mountain ③, particularly the Three‐Parallel‐Rivers Region of the former, was characterized by remarkable biodiversity, endemism, and distinctiveness. These areas supported a high proportion of threatened species and were key regions for rapid evolutionary diversification. Type II, centered on the Nanling, Miaoling, and Luoxiao Mountains ⑤, displayed pronounced slope‐driven uniqueness and provided critical habitats for 41 threatened species. Type III, represented by the Changbai Mountains ⑨, exhibited intercept‐driven uniqueness and acted as a paleo‐endemism center, harboring evolutionarily distinct relict taxa. Type IV encompassed regions characterized by significant endemism and certain threatened species, including mixed‐endemism centers such as the Taiwan Mountains ⑦ and the Laojun Mountain–Dawei Mountain ④, as well as paleo‐endemism centers in the Tianmu Mountain–Yandang Mountain in Zhejiang ⑥, and the Qian Mountain–Nuluerhu Mountain in Liaoning ⑧. Furthermore, adjacent areas of the Daba Mountain–Qinling Mountain ②, which contain rare threatened species such as R. roxieoides (Endangered) and R. purdomii (Vulnerable), warrant special conservation attention and further investigation.

Figure 6.

Priority conservation areas, climate change vulnerability, and current protected areas (PAs) coverage for Rhododendron in China

(A) The top 10% of priority areas, identified as evolutionary and functional hotspots (i.e., ‘islands’ and ‘hills’), and the minimum area required to protect threatened species, including: ① Eastern Himalayas and Hengduan Mountains, ② Qinling Mountain–Daba Mountain, ③ Daliang Mountain–Wumeng Mountain, ④ Laojun Mountain–Dawei Mountain, ⑤ Nanling, Miaoling, and Luoxiao Mountains, ⑥ Tianmu Mountain–Yandang Mountain, ⑦ Taiwan Mountains, ⑧ Qian Mountain–Nuluerhu Mountain, and ⑨ Changbai Mountains. (B) Climate change levels across the distribution of Rhododendron. (C) Proportion of each grid cell covered by PAs. (D, E) Comparison of climate change levels (D) and PAs coverage (E) between priority and non‐priority areas using the Wilcoxon test. In (D, E), violin plots with embedded boxplots depict the distribution of the data. In the boxplots, the center line indicates the median, the box shows the interquartile range (IQR), and the whiskers extend to 1.5 × IQR.

Notably, these priority areas are predominantly concentrated in mountainous regions, which are particularly vulnerable to the impacts of future extreme climate change (2081–2100; SSP585 scenario) (Figure 6B, D). Encouragingly, the proportion of formally protected land within priority areas is significantly higher than that in non‐priority areas (Figure 6C, E). However, 14.2% (31 grid cells) remain uncovered by protected areas, primarily distributed in regions ①, ③, ⑤, and ⑧ (Figure S7A). When the conservation target was raised to 17%, 25%, and 30%, the identified priority areas expanded outward from the original 10% priority conservation areas (Figure S8), with the conservation gap growing to 18.9%, 22.6%, and 22.5% (Figure S7B–D).

DISCUSSION

Patterns of Rhododendron diversity in China and their key drivers

Rhododendron species are widely distributed in the mountainous regions of southern China. The spatial distribution patterns of Rhododendron in China exhibited significant consistency across diversity dimensions, with the Three‐Parallel‐Rivers Region in the southern Hengduan Mountains emerging as the core area (Figure 1). This pattern aligns with findings reported for Primulaceae (Bai et al., 2020) and for flowering plants more broadly (Lu et al., 2023). Our results showed that temperature seasonality and elevation range served as representative indicators for climatic seasonality and habitat heterogeneity, which were crucial factors influencing the multidimensional diversity of Rhododendron in China (Figure 4). At the taxonomic level, temperature seasonality and elevation range emerged as the primary drivers of diversity (Figure 4A, B), aligning with the findings of Shrestha et al. (2018a) on Rhododendron species richness in China. On a global scale, studies on Rhododendron richness patterns similarly supported the importance of elevation range (Xia et al., 2022; Guan et al., 2024;). These studies primarily focus on the influence of contemporary climate and habitat heterogeneity on species diversity patterns, while the global analysis by Guan et al. (2024), based on island biogeography theory, further identified mountain area as the strongest predictor of richness. Notably, our study found that, although energy factors exerted a relatively weak effect at the taxonomic level, they were dominant factors in influencing phylogenetic and functional structures (Figure 4C, D).

Previous studies have demonstrated that the difference in species richness between northern and southern China is closely related to temperature seasonality (Zhao et al., 2023b; Zhu et al., 2025). In alignment with the climate stability hypothesis, the relatively low temperature seasonality and minimal fluctuations in southern China create stable conditions that promote the accumulation of species diversity (Klopfer, 1959; Klopfer and Macarthur, 1961). These processes result in increased species richness and greater phylogenetic and functional diversity.

Within the broadly defined southern China (south of the Qinling‐Huaihe line), where temperature seasonality is relatively weak (Shrestha et al., 2018a), southwestern China exhibited higher species richness compared to the narrowly defined southern China (Figure S3B), particularly for threatened species (Figures 1A, S1A). This pattern is closely associated with macrotopographical features of China. The primary distribution centers of Rhododendron are concentrated in the eastern Himalayas and Hengduan Mountains, where steep elevation gradients and the interdigitation of mountain ranges and valleys create a highly heterogeneous habitat mosaic (Rahbek et al., 2019). Such heterogeneity promotes species coexistence and accelerates population differentiation through geographic isolation and dispersal barriers (Liu et al., 2024b; Yuan et al., 2025). Consequently, this often results in narrow species ranges (Liu et al., 2022), which are more vulnerable to anthropogenic pressures and climate change.

The high mountain barriers and complex terrain of the Hengduan Mountains further create relatively stable and diverse microclimatic environments that buffer against large‐scale climatic fluctuations (Scherrer and Körner, 2011; Chang et al., 2023). These refugial environments facilitate the persistence of plant populations and maintain genetic diversity. Indeed, the Hengduan Mountains have been identified as major glacial refugia for alpine plants (Ding et al., 2020). Consequently, this region not only preserves intermediate‐aged Rhododendron lineages but also promotes rapid radiation and adaptive diversification, giving rise to exceptionally high phylogenetic diversity and centers of mixed endemism (Figure 2A).

Additionally, temperature‐related energy factors, MTWQ and MAT, significantly shaped phylogenetic and functional structures, respectively (Figure 4C, D). The Himalaya–Hengduan Mountains serve as a center for the rapid species diversification of Rhododendron, where low temperatures in alpine habitats act as an important environmental filter (Zhang et al., 2023a). This has led to phylogenetic conservatism in cold‐tolerance functional traits, such as leaf texture (leathery) and life form (evergreen) (Table S3), resulting in both phylogenetic and functional clustering. In contrast, in the warm lowlands of southern China, abundant energy alleviates abiotic environmental stress. In this context, biotic interactions, particularly interspecific competition, play a prominent role in community assembly, promoting niche differentiation and trait divergence among species. Although human activities exert certain influences on diversity patterns, their impact remains relatively limited (Figure S9).

Regional differences in uniqueness patterns

In the context of the phylogenetic uniqueness‐distance increase model, the mechanisms underlying phylogenetic uniqueness exhibited significant regional variability, reflecting the deep biogeographical history of Rhododendron, specifically, its early Paleocene origin at high northern latitudes, subsequent southward migration during the Neogene, and extensive evolutionary radiation of species in the Himalaya‐Hengduan Mountains and Southeast Asia during the late Miocene and thereafter (Xia et al., 2022). The Changbai Mountains, in particular, demonstrated low species richness but relatively high phylogenetic uniqueness (Figures 1H, 3A, C). The high intercepts in this region indicate the preservation of ancient lineages such as R. redowskianum and R. tomentosum. The existence of these ancient lineages in northeastern China supports the hypothesis of a high‐latitude origin for Rhododendron. A comparable evolutionary trajectory is also observed in Saxifraga, which originated in North America and subsequently underwent rapid diversification in the Hengduan Mountains, accompanied by ongoing orogeny (Ebersbach et al., 2017).

In contrast, the Hengduan Mountains showed the highest level of richness and phylogenetic uniqueness, characterized by both high intercepts and steep slopes (Figures 1H, 3A, C). This pattern reflects the accumulation of intermediate‐aged and numerous younger lineages (high intercepts) (Ding et al., 2020; Cao et al., 2025), along with rapid community differentiation driven by complex topography and pronounced environmental heterogeneity (steep slopes) (Qian et al., 2020). These findings are consistent with the historical process of extensive radiation in Rhododendron after migrating southward to the Himalaya‐Hengduan Mountains. In Southern China, an intermediate level of uniqueness was observed, characterized by low intercepts but steep slopes (Figures 1H, 3A, C). This suggests that local floristic communities are not markedly distinct from adjacent floras at short distances, but differentiation increases rapidly with distance. This pattern likely reflects the region's relatively stable geological history compared to the Hengduan Mountains, with the predominantly hilly landscape imposing fewer dispersal constraints (Ye et al., 2017), particularly for the small, lightweight Rhododendron seeds primarily dispersed by wind (Stephenson et al., 2007; Wang et al., 2014).

At the functional level, the uniqueness‐increase model exhibited similar patterns to those observed at the phylogenetic level. However, regions with high intercepts and steep slopes were more widespread, covering the entire Hengduan Mountains (Figure 3A, B, D). This broader distribution implies that the radiation of Rhododendron in the region has contributed to both high phylogenetic diversity and the profound divergence of functional traits. The marked spatial environmental heterogeneity likely drives adaptive differentiation in key functional traits (Stark et al., 2017), and the resulting functional differences enable species to effectively partition ecological niches (Kraft et al., 2015). This niche differentiation, together with the ensuing complementarity, reduces competition among species with different trait combinations, thereby supporting species coexistence, boosting community biodiversity, and driving the evolution and maintenance of diversity in the region (Cadotte, 2017; Liu et al., 2024a). This interpretation aligns with findings from Xia et al. (2022), which highlighted the role of leaf functional trait adaptation in reinforcing the evolutionary radiation of Rhododendron.

Priority conservation areas identification and conservation implications

While biodiversity hotspots based on the 30%‐criterion showed high spatial concordance among metrics, this concordance decreased under more restrictive definitions (Figure S10). A pronounced divergence was observed in the Changbai Mountains, the Taiwan Mountains, and southeastern Yunnan (Figure S10D–I).

Although these diversity indices capture distinct dimensions of biodiversity, their intrinsic correlations complicate conservation prioritization and provide limited complementary information. Therefore, we selected a subset of key indices to identify priority conservation areas from both evolutionary and functional perspectives, which were emphasized in the Convention on Biological Diversity, while also incorporating the minimum distribution ranges of threatened species. Ultimately, four categories of priority conservation areas were identified (Figure 6A). Consistent with previous studies, our results reaffirmed the conservation significance of the eastern Himalayas and the Hengduan Mountains, while also recognizing secondary hotspots, including southeastern Yunnan, the Daba Mountain–Qinling Mountain, the Nanling Mountains, and the Taiwan Mountains (Yu et al., 2017; Shrestha and Wang, 2018; Zhao et al., 2023b). Our multi dimensional framework further revealed critical paleo‐endemism centers, such as the Changbai Mountains and the Qian Mountain–Nuluerhu Mountain in northeastern China, and the Tianmu–Yandang Mountains in southeastern China, which have largely been overlooked by taxonomy‐only approaches.

Given the spatiotemporal heterogeneity of biodiversity hotspots, conservation strategies should be developed with adaptability and specificity. Hotspots, including the Hengduan Mountains ①, Daliang–Wumeng Mountains ③, Laojun Mountain–Dawei Mountain ④ in southwestern China, and the Taiwan Mountains ⑦ in southern China, exhibited high vulnerability to climate change (Figure 6A, B). Notably, conservation gaps persisted at the peripheries of regions ① and ③. In alignment with Global Biodiversity Framework Target 12, a key conservation priority is to fill protection gaps in these critical areas and enhance ecological connectivity among reserves, thereby strengthening the climate resilience and integrity of the overall conservation network (Zhao et al., 2023a; Zhu et al., 2025).

The hilly and mountainous regions of southern China constitute another diversity hotspot for Rhododendron, including several species listed as National Key Protected Wild Plants, such as R. huadingense, R. jingangshanicum, and R. kiangsiense. Our study identified the Miaoling, Nanling, and Luoxiao mountains as priority areas in this region ⑤. However, the current network of protected areas in this region remains fragmented and patchy. The establishment of a “Southern Hilly and Mountainous Belt” conservation network in the future would significantly enhance the long‐term persistence and systematic conservation of flora and fauna in this region. In densely populated eastern China, Rhododendron hotspots ⑥ are more effectively protected through the implementation of plant micro‐reserves (Ma et al., 2013; Zhou et al., 2023). This fine‐scale protection advances Global Biodiversity Framework Target 4 by preventing the loss of threatened populations due to habitat fragmentation in anthropogenic landscapes.

Compared to other regions, northeastern China, particularly the Changbai Mountains and the Qian Mountain–Nuluerhu Mountain, serves as a center of paleo‐endemism for Rhododendron, harboring the oldest lineages such as R. redowskianum, and threatened species such as R. dauricum. These regions are crucial for alleviating the irreversible loss of relict Rhododendron lineages and conserving rare species. However, under the 30% conservation target scenario, this region exhibits extensive protection gaps, necessitating comprehensive conservation efforts (Zhao et al., 2024).

Collectively, our study uses Rhododendron, a widely distributed montane genus encompassing both ancient relict and rapidly radiated lineages, as an ideal model for evaluating the effectiveness of conservation efforts aimed at protecting mountain plant diversity and key habitats in China. By integrating multidimensional diversity data, our study delineated four types of priority areas, comprising nine montane conservation units. However, it is important to note that Rhododendron does not fully represent the conservation needs of other ecosystems, which are shaped by different ecological processes, evolutionary histories, and anthropogenic pressures, requiring tailored conservation strategies. Integrating more comprehensive data in the future will be essential for a thorough assessment of conservation priorities.

MATERIALS AND METHODS

List of Rhododendron species in China

A checklist of Rhododendron species in China was compiled from multiple sources: Flora of China (FOC: 571 species, http://www.iplant.cn/foc), Plants of the World Online (which lists five additional species recorded in China beyond those listed in FOC, https://powo.science.kew.org), protologues for 30 newly described species not included in FOC after 2004 (Table S4), one rediscovered species (R. wuyishanicum), and a species of Ledum merged into Rhododendron (e.g., R. tomentosum). Three cultivated species (R. indicum, R. obtusum, and R. pulchrum) and two putative hybrids (R. × inopinum and R. × paradoxum) were excluded. Our study focused solely on species‐level diversity, with a total of 603 Rhododendron species recorded in China (as of October 5, 2023) (Table S5).

Predicting species' distribution ranges

The distribution data for the Rhododendron species comprised over 70,000 records, sourced from online databases, field surveys, and protologues of newly described species (see Zhu et al. (2025) for further details). We implemented a rigorous filtering process: excluding incomplete, duplicate, and cultivated records; re‐identifying anomalous specimens; and georeferencing records lacking coordinates using Google Earth, based on locality descriptions, including administrative and habitat information from specimen labels. Finally, we spatially rarefied the dataset to retain only one record per species within each 5 × 5 km grid cell. The final dataset contained 11,198 records for 603 Rhododendron species in China (Figure S3B).

To estimate distribution ranges, we employed a dual approach. We constructed species distribution models for species occurring in seven or more grid cells (304 species) and generated a 50 km radius buffer around each point for species with fewer than seven grid cells (299 species) to delineate their distribution range (Wang et al., 2024).

We selected 28 environmental variables influencing plant growth and distribution, which were categorized into five groups: 19 climate variables, one elevation variable, three land cover variables, four soil variables, and one population density variable (for further details, see Supplementary Methods in the supporting information) (Xie et al., 2022; Yu et al., 2017). All variables were resampled to a resolution of 5 km and projected to the WGS 1984 UTM Zone 48 N coordinate system (EPSG: 32648). We assessed collinearity among all variables using the variance inflation factor (VIF) and Spearman correlation (ρ) with the usdm package in R 4.1.3 (Naimi and Araújo, 2016), excluding variables with a VIF > 10 or |ρ| > 0.75. Ultimately, 13 variables remained for model construction (Table S6).

Species distribution models were developed using the maximum entropy algorithm (MaxEnt) (Phillips et al., 2006), a widely recognized method for its efficacy and robustness in predicting potential species distributions from presence‐only data (Elith et al., 2006; Kaky et al., 2020). For each species, the model calibration area was defined by constructing a minimum convex polygon encompassing all distribution points, with an additional 200 km buffer (Mota et al., 2022). To enhance model performance, we systematically tuned MaxEnt parameters by combining 10 regularization multipliers (ranging from 0.5 to 5.0 in 0.5 increments) with three feature classes (linear, quadratic, and product), yielding 70 candidate models for each species. Model cross‐validation was performed using block partitioning for species with more than 15 occurrence points and the jackknife method for those with fewer occurrence points (da Costa‐Pinto et al., 2024). From these candidate models, the top 10% of performing models, determined by the lowest omission rate and highest area under the curve (AUC), were averaged to create a final consensus model. Continuous probability outputs were converted to binary (presence/absence) predictions using the “ten percentile training presence” threshold, which reduced sensitivity to potential errors in occurrence records (Raes et al., 2009). For the 12 species with an AUC below 0.7, we excluded model predictions and relied solely on distribution points (da Costa‐Pinto et al., 2024).

A 50 × 50 km grid was overlaid on the map of China, and grid cells with less than 50% of their area within the national boundary were excluded. Potential habitats for species were defined as grid cells where the centers were covered by either the model predictions or the buffers. For four species, where no grid cell center was covered, we considered cells with spatial overlap as potential areas for species occurrence.

Construction of plastome phylogenetic tree for Rhododendron in China

A phylogeny comprising 442 Rhododendron species and three subspecies native to China, with the subspecies included as taxonomically similar to data‐deficient species, was derived from the genus‐level plastome phylogenetic tree of Mo et al. (unpubl. data), which was constructed using plastome sequences for 588 species (including 21 subspecies). To incorporate species lacking plastome sequences, we used a taxonomic similarity substitution approach, following the species description and classification system employed in FOC and/or the genus Rhododendron of China (Geng, 2014). Specifically, 137 species were inserted as polytomies with their morphologically closest relatives within the same subsection or section. Additionally, the remaining 24 species, which lacked clear morphological affinities or molecular data, were randomly inserted within their respective subsection or section in polytomies.

Construction of a functional dendrogram

To quantify the functional diversity of Rhododendron species in China, we collected seven key traits for each species: growth form (shrubs or trees), plant height, leaf length, leaf width, leaf texture (leathery, papery, or membranous), number of flowers per inflorescence, and corolla length. These functional traits were compiled from FOC, The genus Rhododendron of China, and other relevant literature (see Table S4). Collectively, these traits captured major axes of ecological strategy, spanning resource acquisition and stress tolerance (growth form, plant height, leaf size, and texture) as well as reproductive investment and pollination strategy (floral traits). Specifically, growth form reflects variation in plant architecture and light‐acquisition strategies; plant height captures the balance between light competition and increasing hydraulic constraints with plant stature (Wang et al., 2019; Luo et al., 2024); leaf size and texture indicate trade‐offs among light capture, thermal and water regulation (Zhang et al., 2020; Zou et al., 2021); and floral traits describe variation in pollinator attraction and the degree of morphological matching with pollinators. For continuous traits, we utilized the maximum recorded value for each species. Based on these traits, we computed Gower's distance matrix to quantify pairwise functional dissimilarity among the species (Pavoine et al., 2009). Subsequently, we constructed a functional dendrogram using hierarchical clustering with the unweighted pair group method with arithmetic mean (Wang et al., 2023).

Diversity assessment

To characterize biodiversity patterns, we calculated taxonomic, phylogenetic, and functional diversity metrics for each grid cell. Taxonomic diversity was quantified as SR and WE. PD was calculated as the total branch length of all species present in a grid cell, while PE was calculated by weighting each branch according to the inverse of the geographic range of all its descendant species (Rosauer et al., 2009). PU for each grid cell was quantified as the average Ruggiero's phylogenetic dissimilarity between its community and all others, defined as the proportion of the total branch length that is unique to the focal cell in each pairwise comparison (Baselga et al., 2025). FD, FE, and FU were derived in a similar manner from the functional dendrogram. EDGE for each cell was computed as the sum of $\sum \left[\ln \right(1+\mathit{ED})+\mathit{GE}\times \ln 2]$ across all resident species (Isaac et al., 2007), where ED represents the evolutionary distinctiveness of each species and GE is the global endangerment score based on the China Biodiversity Red List—Higher Plants (2020) (Ministry of Ecology and Environment, Chinese Academy of Sciences, 2023; Table S5). Species classified as Data Deficient (108 species, 17.9%) were assigned a GE value of 3.5 based on the Kruskal‐Wallis test, as their range sizes (grid cells) fell between those of Endangered and Critically Endangered species (Figure S11). The metrics (SR, WE, PD, PE, FD, FE) were also computed exclusively for threatened species.

To assess whether the observed PD and FD deviate from random expectations, we calculated standardized effect sizes (SES) for PD, FD, mean pairwise phylogenetic distance (MPD), and mean pairwise functional distance (MFD) (i.e. sesPD, sesFD, sesMPD, and sesMFD) using null models that randomized species positions on the phylogenetic tree. Negative or positive SES values indicated phylogenetic/functional clustering or overdispersion, respectively, after accounting for SR. All analyses were conducted in R 4.1.3 using the packages picante (Kembel et al., 2010), phyloregion (Daru et al., 2020), betapart (Baselga and Orme, 2012), vegan (Oksanen et al., 2022), and FD (Laliberte and Legendre, 2010), with PU and FU calculations following Baselga et al. (2025). Spatial patterns of biodiversity metrics were mapped in ArcGIS 10.7 (ESRI, 2019).

Uniqueness analysis based on the distance‐increase model

To determine whether the spatial pattern of PU and FU in Chinese Rhododendron species arises from fine‐scale or broad‐scale differentiation, we applied the distance‐increase model for PU proposed by Baselga et al. (2025). This framework distinguishes two scenarios: “hills” characterized by steep distance‐decay slopes (indicating high regional similarity that decays rapidly with distance), and “islands” characterized by high intercepts (indicating inherent distinctiveness even from nearby communities). Similarly, we applied the model for FU to explore the differences and consistencies in phylogenetic and functional uniqueness patterns. The model was constructed based on the formula $\mathrm{d}=1-\mathrm{a}\times e^{b\times s}$, where d represents PU or FU, a is the intercept, b is the slope, and s is the spatial distance.

Categorical analysis of neo‐ and paleo‐endemism (CANAPE)

We identified centers of neo‐ and paleo‐endemism using the canaper package (Nitta et al., 2023), following the CANAPE framework (Mishler et al., 2014). The approach evaluates the statistical significance of both PE and relative phylogenetic endemism (RPE), where RPE quantifies the ratio of observed PE to the PE calculated from a null model tree with equal branch lengths. Grid cells are then classified into four categories: neo‐endemism centers, paleo‐endemism centers, mixed‐endemism centers, and super‐endemism centers.

To elucidate differences among endemism centers, we applied the diversification rate framework of Mo et al. (unpubl. data). For each species, we extracted the net diversification rate and the most recent divergence time with its sister species. We then calculated the average net diversification rate and divergence time for all species (442 consistent species) within each grid cell and compared these values across the delineated endemism centers.

Identifying priority conservation areas

We evaluated the efficiency of various diversity metrics (SR, WE, PD, PE, PU, EDGE, FD, FE, FU) in optimizing the conservation of phylogenetic and functional diversity. For each metric, we ranked grid cells in descending order and progressively incorporated them into a virtual reserve network. After each addition, we calculated the cumulative proportion of protected PD or FD relative to the total available diversity and plotted them against the fraction of area placed under protection (protected cells /total cells) (Baselga et al., 2025).

Our results indicated that PE facilitated the most rapid accumulation of PD, while WE and FE were the most efficient in enhancing FD. To delineate uniqueness regions, we ranked grid cells by PE and selected the top 10% as candidate cells, which were then partitioned using the slope and intercept parameters of the phylogenetic uniqueness‐distance model (Baselga et al., 2025). Within these candidate cells, those with slopes in the top 10% were designated as “evolutionary hills” (stronger distance‐related differentiation), while those with intercepts in the top 10% were classified as “evolutionary islands” (high distinctiveness at local scales). We applied the same procedure to WE and FE to define corresponding “functional hills” and “functional islands”. To align with the phylogenetic‐level analysis, we selected the FE‐based results for further examination. We further repeated the delineation under 17%, 25%, and 30% thresholds as a sensitivity analysis.

To maximize species protection with minimal area required, we employed a complementary algorithm to identify the smallest areas capable of safeguarding all species, including those under threat (Dobson et al., 1997). The algorithm initiated by selecting the grid cell with the highest species richness, excluding species already present in previously selected grids from subsequent iterations. The next grid was chosen based on the highest number of additional unprotected species. In cases where multiple grids contained the same number of unrepresented species, one was selected randomly. This process terminated once every species was present in at least one selected grid.

The identified “hills” and “islands” along with the minimum area encompassing all threatened species, were designated as priority conservation areas. To assess the magnitude of climate change faced by priority versus non‐priority areas, we performed a principal component analysis (PCA) on 19 bioclimatic variables (Bio1–Bio19) under both current (1970–2000) and future (2081–2100; SSP585 scenario) climate projections (Cui et al., 2024). These projections were derived from three general circulation models: MRI‐ESM2‐0, MIROC6, and IPSL‐CM6A‐LR. For each grid cell, we quantified climate change by computing the distance between its current and future climate positions within the PCA‐defined climate space (PC1 and PC2) (Cui et al., 2024). Differences in climate change distance between priority and non‐priority areas were evaluated using the Wilcoxon test. Furthermore, we overlaid the priority areas onto the existing network of nature reserves (sourced from the China Nature Reserve Specimen Resources Sharing Platform) to evaluate the extent of protection coverage.

Key factors influencing diversity patterns

To investigate the relative contributions of environmental factors to the spatial patterns of Rhododendron diversity in China, we selected 18 environmental variables commonly used in macroecological studies (Wang et al., 2024; Yang et al., 2022). These variables were classified into six categories: energy, water, climatic seasonality, heterogeneity, human influence, and historical climate (Table 1), representing contemporary climate conditions, habitat characteristics, human activities, and climatic stability since the Last Glacial Maximum.

Table 1.

Explanatory variables selected to explain diversity patterns

Variable	Description	Abbreviation	Source
Energy	Mean annual temperature	MAT	https://worldclim.org/
	Mean temperature of the warmest quarter	MTWQ	https://worldclim.org/
	Mean temperature of the coldest quarter	MTCQ	https://worldclim.org/
	Potential evapotranspiration	PET	https://www.global-ai-pet.org/global-aridity-index-pet-database
Water	Annual precipitation	MAP	https://worldclim.org/
	Precipitation of the driest quarter	PDQ	https://worldclim.org/
	Arid index	AI	https://www.global-ai-pet.org/global-aridity-index-pet-database
	Actual evapo‐transpiration	AET	https://figshare.com/articles/dataset/Global_High-Resolution_Soil-Water_Balance/
	Precipitation of the wettest quarter	PWQ	https://worldclim.org/
Climatic seasonality	Temperature seasonality	TSN	https://worldclim.org/
	Precipitation seasonality	PSN	https://worldclim.org/
	Mean diurnal air temperature range	DATR	https://worldclim.org/
	Annual air temperature range	ART	https://worldclim.org/
Heterogeneity	Elevation range	ELER	https://www.earthdata.nasa.gov/data/instruments/srtm
	Shannon diversity of enhanced vegetation index	SDI	http://www.earthenv.org/
Historical climate	Temperature anomaly	TCL	https://worldclim.org/
	Precipitation anomaly	PCL	https://worldclim.org/
Human influence	Human influence	HII	http://sedac.ciesin.columbia.edu/wildareas/

We initially employed a random forest model to preliminarily assess the relative importance of each environmental predictor in explaining diversity patterns (Breiman, 2001), using the rfPermute package (Archer, 2023). We also calculated Spearman's ρ and VIF for all predictors to assess multicollinearity. Based on the importance rankings, we sequentially removed predictors with |ρ| > 0.75 or VIF > 10, and retained the top five predictors to construct generalized additive models (GAMs) and simultaneous autoregressive model with spatial error (SAR_err) (Hastie and Tibshirani, 1990; Dormann et al., 2007). GAMs were used to capture potential nonlinear relationships between responses and predictors, whereas SAR_err models were used to account for spatial autocorrelation and reduce the risk of spurious significance driven by spatial structure. For both GAM and SAR_err, we performed model selection by ranking candidate models using AICc and selecting the model with the lowest AICc as the optimal model. In the GAM, we further evaluated the relative importance of the variables. For the SAR_err model, we parameterized SAR_err models by testing neighbor distances (50–2,000 km, in 50 km increments) and spatial weights (row‐standardized, binary, and variance‐stabilizing coding schemes). The optimal distance and weight were selected by minimizing the sum of absolute values of Moran's I for the top 20 distance classes (minRSA) (Kissling and Carl, 2008), after which the best predictor set was identified based on AIC under this optimal spatial parameter combination.

Given the high correlations among SR, PD, and FD, as well as among WE, PE, and FE, we used sesMPD and sesMFD as response variables to identify environmental factors of phylogenetic and functional structure. These indices quantified deviations from null expectations while controlling for species richness. Prior to SAR_err modeling, SR and WE were Log₁₀‐transformed to reduce skewness and the influence of extreme values, and all environmental variables were standardized to ensure comparability across predictors. Due to the lack of a linear relationship between the WE of threatened species and the candidate variable PCL, we focused solely on the results of the GAM model for threatened species. All statistical analyses were performed using the spdep (Bivand et al., 2013), spatialreg (Bivand and Piras, 2015), mgcv (Wood, 2017), and ncf packages (Bjornstad, 2022).

CONFLICTS OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

L.‐M.G., M.‐S.Z., and D.‐Z.L. conceived the study. M.‐S.Z., Z.‐Q.M., Y.‐H.L., and C.‐N.F. designed the methodology. M.‐S.Z. and Z.‐Q.M. performed the analyses and wrote the original draft. L.‐M.G., T.Z., and J.C. provided resources and curated the data. M.‐S.Z., L.‐M.G., M.M., W.Z., and D.‐Z.L. reviewed and edited the article. L.‐M.G. and D.‐Z.L. acquired funds. All authors read and approved the paper.

Supporting information

Additional Supporting Information may be found online in the supporting information tab for this article: http://onlinelibrary.wiley.com/doi/10.1111/jipb.70253/suppinfo

Figure S1. Spatial patterns of diversity metrics for threatened Rhododendron species in China

Figure S2. Spatial patterns of standardized effect sizes for phylogenetic and functional diversity metrics of Rhododendron in China

Figure S3. Distribution of Rhododendron species in China

Figure S4. Variable selection and importance for threatened Rhododendron species richness and weighted endemism

Figure S5. Evolutionary and functional “hills” and “islands” across thresholds (10%, 17%, 25%, and 30%), defined as the intersections of top‐ranked phylogenetic/functional endemism and the top quantiles of the slope or intercept from the corresponding uniqueness‐distance increase models

Figure S6. Comparison of functional priority areas identified by weighted endemism (WE) and functional endemism (FE)

Figure S7. Grid cells within priority conservation areas not covered by current protected areas (PAs)

Figure S8. Priority areas identified by evolutionary and functional hotspots (i.e., ‘islands’ and ‘hills’) and the minimum area required to protect threatened species

Figure S9. Random forest model for selecting important variables explaining Rhododendron diversity patterns

Figure S10. Spatial distribution of individual metrics for Rhododendron diversity hotspots

Figure S11. Kruskal‐Wallis test comparing the differences in the number of grid cells occupied by species across different threat levels

Table S1. Correlations among Rhododendron diversity metrics

Table S2. Correlations among diversity metrics of threatened Rhododendron species

Table S3. Phylogenetic signal analysis of functional traits in species distributed in the Hengduan Mountains

Table S4. Main reference books and journals for the newly published Chinese Rhododendron species

Table S5. Endangered status and scores for Rhododendron species used in China

Table S6. Predictor variables and their description

Biographies

Zhu, M.‐S. , Mo, Z.‐Q. , Luo, Y.‐H. , Fu, C.‐N. , Zhang, T. , Möller, M. , Cai, J. , Zheng, W. , Li, D.‐Z. , and Gao, L.‐M. (2026). Rhododendron diversity patterns provide new insights for conserving China's montane flora. J. Integr. Plant Biol. 68: 3727–3745.

Data availability statement

The phylogenetic tree and functional dendrogram of Rhododendron, together with the raw data used for their construction and the dataset containing 15 diversity metrics and 18 environmental variables, are available on ScienceDB (https://doi.org/10.57760/sciencedb.32027).