Phylogenetic clustering and overdispersion for alpine plants along elevational gradient in the Hengduan Mountains Region, southwest China


  • Xin-Hui LI,

    1. Key Laboratory of Biodiversity and Biogeography, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming, China
    2. University of Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Xin-Xin ZHU,

    1. Key Laboratory of Biodiversity and Biogeography, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming, China
    2. University of Chinese Academy of Sciences, Beijing, China
    Search for more papers by this author
  • Yang NIU,

    1. Key Laboratory of Biodiversity and Biogeography, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming, China
    Search for more papers by this author
  • Hang SUN

    Corresponding author
    1. Key Laboratory of Biodiversity and Biogeography, Kunming Institute of Botany, Chinese Academy of Sciences, Kunming, China
    Search for more papers by this author


To better understand the elevational pattern of phylogenetic structure shown by alpine taxa and the underlying causes, we analyzed the phylogenetic structure of each elevational belt of alpine plants in the Hengduan Mountains Region, measured by net related index (NRI) and net nearest taxon index (NTI). We found both the indices of phylogenetic diversity indicated that alpine plants tended to show phylogenetic overdispersion at low elevational belts, implying that the distribution of alpine plants in these belts was mainly determined by interspecific competition. Alpine plants at higher elevational belts tended to phylogenetic clustering indicated by NRI, and NTI revealed phylogenetic clustering at the belts between 4300 m and 5500 m, which presumably suggested environment filtering and rapid speciation. Above 5500 m, NTI indicated that the phylogenetic structure became random again, perhaps due to the low intensity of filtering and the large distances between plants at the top of the scree slopes. We concluded that phylogenetic structure was, indeed, influenced by the environmental filter, interspecies interaction, rapid speciation during the uplift of the Qinghai–Tibet Plateau, and distance between plants.

Since the concept of phylogenetic diversity was first presented, phylogenetic frameworks have increasingly been applied to study biodiversity patterns in plants (Webb, 2000; Webb et al., 2006; Bryant et al., 2008; Kluge & Kessler, 2011) and animals (Graham et al., 2009; Auguet et al., 2010; Jones & Hallin, 2010; Machac et al., 2011; Wang et al., 2012). For example, the phylogenetic study of ferns along the elevational gradient indicated that phylogenetic overdispersion (high differentiation) under mild environmental conditions was presumably due to interspecific competition or another reason, whereas phylogenetic clustering under extreme environmental conditions was due to environmental filtering (Kluge & Kessler, 2011). The same phylogenetic structure was found in microbe communities along the elevational gradient (Wang et al., 2012). Despite phylogenetic frameworks having been developed for some plants, until recently, few studies have focused on alpine plants, which grow in one of the harshest environments on Earth.

To explain the phylogenetic structures along an elevation, the niche conservatism assumption is very important. For example, many species of Saussurea are confined to alpine regions, some only to scree slopes, such as Saussurea medusa Maxim., and many plant clades are found only under forest. Niche conservatism describes the tendency for closely related species to retain their niches and related ecological traits over time (Wiens & Graham, 2005; Wiens et al., 2010). Where niche conservatism and environment filtering exist, they will tend to create phylogenetic clustering (Webb et al., 2002). A case in ant communities indicated that those communities at high elevation tended to phylogenetic clustering, due to environmental filtering (Machac et al., 2011). We assume that when species migrate from lower elevation to higher elevation, if niche conservatism exists, habitat filtering will only allow some lineages to survive the harsh environment and lead to phylogenetic clustering when the plants pass through “filters” which, in alpine areas, can be low temperatures, limited precipitation, scarce pollinators, high solar radiation, or strong winds, all of which change greatly along elevation.

The Hengduan Mountains Region is the core of the south-central China biodiversity hotspot (Myers et al., 2000) and is located on the eastern edge of the Qinghai–Tibet Plateau, the youngest plateau in the world (Wu, 1988). Due to the complexities of geographic isolation, tectonic uplift, climatic changes, strong microhabitat differentiation, and a varied history of migration and/or evolution in the Hengduan Mountains Region, the region supports many alpine plant taxa (Wu & Wang, 1983; Li & Li, 1993). We collected 3411 alpine seed plant species (including subspecies and varieties) distributed throughout the Hengduan Mountains Region. Excluding subspecies and varieties, 3030 alpine seed plants are recorded, of which 12 are gymnosperms. The worldwide alpine flora has been estimated to be composed of 8000–10 000 species, approximately 4% of all known vascular plant species (Körner, 2003). According to this data, the number of alpine plants in the Hengduan Mountains Region accounts for approximately 30% of all alpine plant species globally, probably the largest number of alpine plant species in any alpine region globally, much greater than that in the European Alps, where approximately 650 alpine taxa were documented (Agakhanjanz & Breckle, 1995). In addition, the Hengduan Mountains Region is the important differentiation center of many temperate plant species and is a center of distribution and speciation for many alpine taxa (Sun, 2002; Sun & Li, 2003), such as Rhododendron, Pedicularis, Saxifraga, Gentiana, Corydalis, Saussurea, Primula, Aconitum, Delphinium, Cremanthodium, and Meconopsis. Thus, this area has captured the interest of many researchers worldwide.

In this study, we examined phylogenetic structure along elevation. We tested: (i) whether alpine plants at lower elevation with relatively favorable conditions showed phylogenetic overdispersion; (ii) whether alpine plants at high elevation with relatively harsh climate showed phylogenetic clustering; and (iii) which pattern the phylogenetic structure would exhibit, unimodal, or increase along the elevational gradient. We also tried to interpret the underlying reasons for the pattern of phylogenetic structure.

1 Material and methods

1.1 Study area

The region of the Hengduan Mountains extends from 24°40′N to 34°00′N and 96°20′E to 104°30′E, covering ca. 364 000 km2, and includes western and northwestern Yunnan, western Sichuan, southeastern Tibet, southeastern Qinghai, and southwestern Gansu (Li, 1987) (Fig. 1). Seven mountain chains and six rivers make up the main geographic features across the region from north to south. The average elevation in this region decreases from northwest to southeast. Gongga Mountain, in western Sichuan, at 7556 m, is the highest mountain in this area (Li, 1989); Yulong Snow Mountain, in northwestern Yunnan, at 5596 m, is the closest glaciated area to the equator in Eurasia (He et al., 2002). The Hengduan Mountains Region is greatly affected by the Indian and Pacific Ocean monsoon circumfluence, resulting in a distinct wet season from mid-May to October and a dry season from October to mid-May the next year (Zhang, 1989).

Figure 1.

Location and terrain of the Hengduan Mountains Region (available from Gongga Mountain, in western Sichuan, the highest mountain in the region, and Yulong Snow Mountain, the southernmost glaciated area in Eurasia, are shown. The six big rivers are also marked, Min River, Dadu River, Yalong River, Jinsha River, Mekong River and Salween River, the first four of which are in the Yangtze River basin.

The alpine plant region is the zone above the treeline and below the nival belt (Grabherr et al., 2003). The typical alpine region is composed of alpine shrub at low elevation, with alpine meadows above and then alpine screes at high elevation in the Hengduan Mountains Region. The alpine area is one of the harshest zones on the planet as a result of low temperature, limited precipitation, lack of pollinators, fluctuating weather, strong winds, and short growing periods (Körner, 2003; Hodkinson, 2005). The average altitude of the treeline is usually around 4000–4200 m in most of northwest Yunnan and southeast Xizang, and it is 3600–4000 m in northwest Sichuan (Wang et al., 2004); the scree slopes above the treeline always start at approximately 4400 m in this region.

1.2 Data sources

Our data were based on substantial surveys carried out since the 1950s and many former materials (particularly specimens collected by G. Forrest, H. Smith, and so on), including comprehensive scientific expeditions to the Qinghai–Xizang Plateau by the Chinese Academy of Sciences and the expeditions related to the biodiversity of the Hengduan Mountains Region carried out between 1984 and 2010, mainly supported by China and the USA, published monographs including Vascular plants of the Hengduan Mountains (Wang et al., 1993, 1994), Flora of Gaoligong Mountains (Li et al., 2000), The vascular plants and their eco-graphical distribution of the Qinghai–Tibetan Plateau (Wu, 2008), and websites built for introducing the collections, including the Biodiversity of the Hengduan Mountains and adjacent areas of south-central China database (, and the Chinese Virtual Herbarium ( All of these materials were used to generate the database in this study. The information in the database includes species identity, genus, family, their altitudinal range, and the fruit type for each entry. This database contains 3411 taxa (3030 species), which belong to 371 genera and 66 families (according to the Angiosperm Phylogeny Group (APG) III, 2009). Of these 3030 species, 3018 are angiosperms.

1.3 Constructing phylogenies

A phylogenetic tree was generated using the informatics tool Phylomatic (Webb & Donoghue, 2005) which followed the APG III system (Angiosperm Phylogeny Group, 2009) at the level of family. The phylogenetic tree used in this study included the 3018 angiosperm species in the alpine region (Fig. S1). Estimated node ages (Table S1) were used to assign branch length in the phylogenetic tree using the “bladj” algorithm in the software Phylocom (

1.4 Assessing the phylogenetic structure of each belt

We divided the study region into 27 elevational belts at 100-m vertical intervals. Each species was treated as spread at every 100-m belt between its top and bottom elevation. For instance, Primula minor Balf. f. & Kingdon-Ward, distributed between 4300 and 5000 m, was treated to be present in each belt of 4300, 4400, 4500, and so on up to 5000 m. This method has been commonly applied to research species richness patterns along elevational gradients (Rahbek, 1997; Grytnes & Vetaas, 2002; Wang et al., 2007; Zhang et al., 2009).

The fruit type and dry or not when ripe, is thought to be a consequence of the plants adapting to the living environment (Bolmgren & Eriksson, 2005; Peng et al., 2012). Due to numerous small seeds easily spread by wind, the proportion of capsules, a type of dry fruit, increases along the elevational gradient (Pellissier et al., 2010). So the fruit type was chosen to test niche conservatism, which was divided into dry fruit or not, and 11 categories: achene; capsule; caryopsis; pome; cremocarp; follicle; legume; nut; silique; berry; and drupe. We calculated Blomberg et al.'s (2003) K statistic of each belt using the fruit type to test niche conservatism (Zanne et al., 2005; Swenson & Enquist, 2007), using the package “picante” implemented in the R software environment (Kembel et al., 2010). K < 1 would imply that closely related taxa are more different from that expected by chance, whereas K > 1 would imply that closely related taxa are more similar than expected, which is a strong phylogenetic signal indicating niche conservatism (Blomberg et al., 2003). Significance was determined using randomization tests with 999 permutations (Kembel et al., 2010).

After testing for niche conservatism, we estimated the phylogenetic structure of each belt using two indices: NRI based on mean phylogenetic distance (MPD); and NTI based on mean nearest taxon distance (MNTD). MPD is an estimate of the average phylogenetic relatedness between all possible pairs of taxa in one belt, and MNTD is an estimate of the mean phylogenetic relatedness between each taxon in one belt and its nearest relative in another belt (Webb et al., 2002). Based on MPD and MNTD, we calculated NRI and NTI indices, which were defined as

display math
display math

We calculated and compared the NRI and NTI (based on presence/absence) of each belt, to 999 randomly calculated NRI and NTI for the same numbers of species as in the observed belts, using the phylogeny representing the entire species pool of alpine angiosperm plants in the Hengduan Mountains Region. To assess whether the observed NRI/NTI values differed significantly from zero, a two-tailed test was used to evaluate significance at P = 0.05, such that an observed number of less than 25 or more than 975 bigger than zero was assumed to indicate significant overdispersion or clustering, respectively. These analyses were carried out in Phylocom 4.2 (Webb et al., 2008).

1.5 Environmental variables

To examine the relationship between phylogenetic indices of each 100-m belt and climate, represented by mean annual precipitation and mean annual temperature, we extracted mean annual precipitation and mean annual temperature for each elevational belt from the WorldClim version 1.4 database (; Hijmans et al., 2005) using ArcGIS 9.3 (ESRI, 2008).

The digital elevation model of the Hengduan range with resolution of approximately 90 m was rasterized into 0.1 × 0.1° grid cells. Combined with the WorldClim version 1.4 database with resolution of approximately 1 km2, the climate data of each belt (represented by mean annual precipitation and mean annual temperature) was calculated with the center points represented in each quadrant.

1.6 Analyses

We examined elevation from 3000 to 5700 m, divided this range into 27 belts at 100-m vertical intervals, then calculated the area of each belt as an explanatory factor. There were three reasons why we chose 3000 m as the lowest elevation. First, hundreds of alpine plants were found below the treeline, so only using data from above the treeline may lead to biased results. Second, in the north of the Hengduan Mountains Region, the treeline is much lower than the central area, for example, it is at 3200 m at Erlang Mountain, northern Sichuan (Jiang et al., 2004). Finally, all vegetation was quite similar above 3000 m, which represented the limit of the dark coniferous Pinus, Picea, and Abies forest (Liu et al., 1984).

We related the area of each belt and the climate variables (mean annual temperature and precipitation) to the phylogenetic structure (NRI, NTI) of each belt using generalized linear models. In the models, area, mean annual temperature, and annual precipitation were taken as explanatory variables. In all cases, a polynomial regression was carried out, comparing models with the first-order term (i.e., elevation), the second-order term (i.e., elevation, elevation2), and the third-order polynomial (i.e., elevation, elevation2, elevation3). We used the maximum likelihood procedure to fit each model, and compared the models using the Bayesian information criterion and R2. These analyses were carried out in R 2.14.2 (R Core Team, 2012).

2 Results

The results showed that K-values of fruit type were all bigger than 1 except the highest belt from 5600 to 5700 m, and the K-values of dry or not mature fruit were more than 1 under 4300 m. These indicated strong evidence for niche conservatism of the traits we examined (Table S2).

Mean annual temperature declined linearly with elevation (F = 1086.922, P < 0.0001); mean annual rainfall also declined linearly with elevation with a little fluctuation (F = 166.907, P < 0.0001; Fig. 2). The two phylogenetic indices of the belts showed different patterns along the altitudinal gradient (Fig. 3): NRI values increased along the altitudinal gradient, described by a significant regression model with the third-order term (Table 1). NRI was significantly less than zero (phylogenetic overdispersion) under 3700 m a.s.l., then the value increased but was not significantly different from zero as elevation increased (phylogenetically random), and no belts had values significantly greater than zero (there was no phylogenetic clustering). NTI values also showed a third-order relationship along the altitudinal gradient, with a unimodal pattern with elevation. NTI increased at lower elevational belts and, after peaking between 4500 and 5000 m, where there were screes, the value decreased at higher elevational belts. The phylogenetic structure at low elevation tended towards significant overdispersion (six belts, Fig. 3: B), then it became random with increasing elevation (seven belts, Fig. 3: B). At higher elevation, it tended to exhibit clustering (12 belts, Fig. 3: B), although when the elevation exceeded 5500 m, the NTI tended to indicate random phylogeny again (two belts, Fig. 3: B).

Figure 2.

Relationships between elevation and mean annual temperature (A) and mean annual rainfall (B).

Figure 3.

Patterns of phylogenetic structure along the altitudinal gradient (A, B) and climate gradients of mean annual temperature (C, D), and mean annual precipitation (E, F). The fitted relationship is indicated by a solid line. Significantly overdispersed belts are indicated by white points, and clustered belts by black points, belts approaching random phylogenetic structure are indicated by gray points.

Table 1. Summary of regression statistics for tests of associations between net related index (NRI) and net nearest taxon index (NTI) and area, temperature, and precipitation
CategoryResponse variabled.f.Polynomial orderR2BIClogLikP-value of model
  1. Polynomial first-, second-, and third-order models were tested against one another; the most parsimonious model was identified using the Bayesian information criterion (BIC). The fitted model is depicted in bold. d.f., degrees of freedom; logLik, log-likelihood.

The environmental variables had a high explanatory power (all R2 > 0.88) for phylogenetic structure (Table 1, bold text). High correlations between phylogenetic structure and temperature and rainfall were detected in the Hengduan Mountains Region (Fig. 3; Table 1). For NRI, precipitation was the most important factor in the Hengduan Mountains Region, whereas for NTI, temperature was the most important.

3 Discussion

Branch length estimated for the phylogeny was quite crude, but was a marked improvement over assigning one to all branch lengths (Webb, 2000). To some extent, the phylogeny, generated using APG III which mainly classifies at the familial level at present, was weakly resolved at the species level, and had a large number of terminal nodes left unresolved, but it is helpful for understanding the phylogenetic structure in the Hengduan Mountains Region. Although WorldClim data presented some problems, the data were used widely in other studies of elevational diversity gradients (McCain, 2009; Machac et al., 2011). The average of the mean annual precipitation and mean annual temperature data were calculated for the each belt, so the climate data were representative.

The value of NRI revealed two kinds of phylogenetic structure (overdispersion and random) along the elevational gradient. NTI, more sensitive than NRI, identified all three possible types of phylogenetic structure (overdispersion, random, and clustering), and more importantly, it reflected changes in the alpine habitat (Fig. 3). The elevation between 3000 and 3600 m was under the forest, the NTI showed phylogenetic overdispersion. At the alpine treeline, in the transition zone at around 3600–4200 m, the NTI showed a random phylogenetic structure. The alpine meadow and alpine screes are mainly distributed over 4200 m, and the NTI indicated phylogenetic clustering. However, the phylogenetic structure was random at altitudes over 5500 m, where the environment is extremely harsh, for example, with the lowest annual temperatures. In the young mountain area, the NTI index may be useful because of a relatively short evolutionary history.

Phylogenetic overdispersion was detected at lower elevations, between 3000 and 3700 m for NRI or between 3000 and 3600 m for NTI, which has been observed in many other studies (Ackerly et al., 2006; Cavender-Bares et al., 2006; Silvertown et al., 2006) and could indicate that negative interactions (e.g., competition) are important in community assembly (Graves & Gotelli, 1993; Webb et al., 2002). This result was consistent with the concept that interspecific competition shapes the composition of a community when niche conservatism exists (Webb et al., 2002). So it is suggested that competition may structure the alpine plants' arrangement in low sites with relatively mild climates (Choler et al., 2001). This is not to say that facilitation or other factors, like habitat heterogeneity, which can lead to less related species in a belt, are unimportant in mild conditions, only that they are not the most important forces to form the phylogenetic structures in low sites with relatively favorable climates.

When the elevation exceeded 3700 m for NRI or 3600–4300 m for NTI, the phylogenetic structure became random, compared to overdispersion at the lower elevation. The habitats between 3800 and 4200 m generally represented the treeline and alpine thickets, which is the transition zone in the Hengduan Mountains Region. Under the treeline, there are many forest gaps for the settle of alpine herbs from high elevation regions; over the treeline in alpine thickets, there are shade places for some plants which mainly live under forests, which means that these belts are buffer zones between areas of lower elevation, where the phylogenetic structure showed overdispersion, and areas of higher elevation where the phylogenetic structure showed clustering.

The phylogenetic structure showed clustering at elevations over 4300 m, above the treeline in the Hengduan Mountains Region. Our results suggested that habitat filtering may play a vital role in the assembly of alpine plants. The existence of niche conservatism was checked in this region, and phylogenetic clustering should be the consequence of habitat filtering (Webb et al., 2002). The alpine flora of the Hengduan Mountains Region, at the southeast of the Qinghai–Tibet Plateau, developed in a totally new region that was progressively uplifted during the Paleocene (Shi et al., 1998). So most of the alpine plant species should be, relatively, much younger than those of other mountains, and are likely to be immigrants from lower regions. Such plants would have been exposed to filtering, if niche conservatism existed, when they moved from low to high elevation as a result of the great change in climate (including temperature and rainfall). After the filtering of environment, the alpine plants that survived were always relatives because they have similar traits to survive in the harsh climate, resulting in phylogenetic clustering. Many close relatives, many of which speciated due to strong microhabitat differentiation or other factors during the uplift of the Qinghai–Tibet Plateau (Zhang et al., 2009), are more closely related taxa than expected by chance (phylogenetic clustering). This may be another cause behind phylogenetic clustering in these belts.

Interestingly, the NTI index showed a random phylogenetic structure above 5500 m, at the top of alpine screes near the permanent snow cover. This pattern might be explained by two reasons. First, because the plants were sparsely distributed and the distance between them became progressively greater with increasing elevation at the top of the screes, and large distances between plants indicates weak interspecific relationships (Vogt et al., 2010; Yuan et al., 2011), the plants became solitary and independent. Second, although the environment changed at elevations over 5500 m, the change was much smaller than that found around the treeline, therefore, the filter effect was not as strong as around the treeline, where there was a transition zone. The alpine plants adapted to the harsh environment gradually with increasing elevation, thus the filtering effect was too weak for the phylogenetic structure to cluster further, and there was a return to a random phylogenetic structure.

4 Conclusion

In summary, our results showed a pattern of phylogenetic overdispersion at low elevation while clustering at high elevation, which suggested interspecific competition was the main force in phylogenetic structures at low elevation belts, whereas environmental filtering (especially mean annual temperatures) and rapid speciation were the main factors to affect lineages present at high elevation belts. In our study, the pattern of NTI showed a tendency towards a second zone of phylogenetic randomness at the belts above 5500 m, presumably due to the low filtering intensity and the large distances among plants at these belts. Our results implied that interspecific interactions, niche conservatism, distance between plants, rapid speciation, and temperature shaped the distribution of species in the alpine region of the Hengduan Mountains Region.


We sincerely thank Wen SHA for her assistance with ArcGIS manipulation. We are also particularly grateful to Zhi-Qiang ZHANG, Jian-Guo CHEN, Guo-Dong LI, and the anonymous reviewers for their helpful suggestions and comments on the manuscript. This work was supported by the Strategic Priority Research Program (B) of the Chinese Academy of Sciences (Grant No. XDB03030112), the National Natural Science Foundation of China (NSFC) (Grant Nos. 40930209, 31061160184), Yunnan Natural Science Foundation and NSFC joint Project (Grant No. U1136601 to Prof. Hang SUN), and the Hundred Talents Program of the Chinese Academy of Science (Grant No. 2011312D11022 to Prof. Hang SUN).