As one of 35 global biodiversity hotspots, the Hengduan Mountains have the most intact native ecosystems and complete elevational vegetation spectrum in China. The Ningjing-Yunling Mountain range is at its core, with two large rivers (Lancang and Jinsha) running in parallel on its west and east side. The Ningjing-Yunling Mountain Range has a typical elevational vegetation spectrum, characterized by shrubs in arid and warm low elevation valleys, mixed broad-leaved and coniferous forests in the mid-elevation, and evergreen shrubs and meadow above the alpine treeline (Yang et al., 2016). Such conspicuous variation in vegetation appearances indicates a strong environmental gradient and divergence in primary plant assembly processes. Our study site, the Hongla Snow Mountain (29°40′ 48′ N, 98°35′ 35′ E), is located in the northern Ningjing-Yunling Mountain range (Figure 1a). The summit of Hongla Snow Mountain is 5,036 m a.s.l., whereas the water surface elevation is at 2,620 m for the Lancang River and 2,495 m for the Jinsha River.

The climate of this region is characterized by marked synchronous seasonality in precipitation and temperature, mainly influenced by the Indian Ocean monsoon coming from the west in summer (Li et al., 2011), and a cold high pressure that forms over the Tibetan Plateau in winter. The interaction between the monsoon and the steep mountain range creates vertical bands of vegetation with significant differences. The valleys at the bottom of the mountain range have a dry-hot climate, resulting in a savanna-like vegetation type, dominated by shrubs and grass with scattered trees; the mid-elevations have a cool and humid climate, resulting in forests dominated by evergreen broad-leaved and coniferous trees; and the high elevations have a cold and humid climate, resulting in alpine vegetation dominated by dwarf evergreen scrubs and meadow.

In July–August 2020, we established 25 transects along the elevation gradient on the west and east slopes of Hongla Snow Mountain, including 12 on the west slope and 13 on the east slope (Figures 1b,c). Each transect area was 0.1 ha (10 m × 100 m), with the long side perpendicular to the slope. The elevations of the transects ranged from 2,700 to 4,900 m on the west slope, and from 2,500 m to 4,700 m on the east slope. The elevational difference between adjacent transects was 200 m. Each transect was divided into ten plots of 100 m² (10 m × 10 m) (Figure 1d). We investigated the plants in the community at the plot scale. We recorded the species name, diameter at breast height (DBH), stem height, population abundance, and coverage for woody plants. The species name, clump stem, height, population abundance, and coverage for herbaceous plants. After the survey, 74 woody species were recorded, with 48 occurring on the western slope and 56 on the eastern slope, and 348 herbaceous species were recorded, with 197 occurring on the western slope and 261 on the eastern slope. In each transect, we collected four or more samples of leaves, bark, and wood for each woody species. All samples were collected from at least four healthy individuals with average DBH, growing under a normal light condition. Fully developed mature leaves were collected on the sunny surface of the outer canopy of each plant. The wood and bark of trees were cut together in a 10 cm × 5 cm area from the main trunk. A 10-cm section of the trunk was cut from each plant for shrubs. All samples were packed in self-sealing bags and taken to the laboratory for the following processing.

A total of 45 species of woody plants were collected on the west and east slopes, with 353 individual plants of 34 species on the west slope and 424 individuals of 38 species on the east. The total coverage of woody species in the plots on the east and west slopes were 95.28 and 92.93% on average (Supplementary Table 1), respectively. The fresh mass of leaf samples was measured using 1/1,000 analytical balance, and the leaf area was measured using the CanoScan LiDE 300 portable scanner. Three points were uniformly selected to measure the thickness along the direction of the central leaf vein using vernier calipers (0.01 mm resolution), avoiding the veins. The thickness of a leaf is the mean of three thickness measurements, and the mean leaf thickness of three measured individuals represents that of a species. The fresh weight of wood samples was weighed with bark stripped, and the volume was measured using the drainage method. In addition, bark thickness was measured with the method of leaf thickness measurement to avoid bark bumps. The leaf samples were oven-dried at 105°C for 20 min, then at 65°C for 72 h to a constant mass, and weighed for dry mass. The wood samples were first exposed to air for about 1 month, then dried at 65°C for 72 h to reach a constant mass and weighed for dry weight. The leaf, bark, and wood traits were calculated according to the following (Cornelissen et al., 2003):

Two chloroplast gene regions (rbcL and matK) and the internal transcribed spacer (ITS) of nuclear ribosomal DNA were downloaded from The National Center for Biotechnology Information (NCBI)¹ using accession numbers for phylogenetic reassembly. For species with no sequences in NCBI, the sequences of their close relatives in the same genus were used as a proxy (Supplementary Table 2). The sequences were aligned in MAFFT version 7 and then concatenated in PhyloSuite v1.2.2. The best substitution model for each gene region was chosen by using jModelTest 2.1.7. Bayesian reconstructions were performed to estimate phylogenetic relationships and divergence times using BEAST v2.6.3. We performed the phylogeny reconstructions with an uncorrelated lognormal relaxed molecular clock model and a Yule speciation tree prior. Six-time calibration priors were set according to Zanne et al. (2014; Supplementary Table 3). A Markov chain Monte Carlo simulation for 100 million generations was set for the reconstructions. Tracer v1.7.1 was used to examine the log file and ensure that all effective sample size values were above 200. The output tree was annotated using TreeAnnotator v2.6.3, with 25% pre-burn-in (see Figure 2).

The elevational vegetation spectrums on the west and east slopes of Hongla Snow Mountain were similarly composed of a wide band of shrubs communities in the low-elevation dry valley, a forest band at mid-elevations, and an alpine band of shrubs and meadow (Figure 3A). On the west slope, shrub communities of variable densities extended from the river bank (2,650 m a.s.l.) upward to 3,300–3,500 m a.s.l., mainly dominated by deciduous species Sophora davidii, Indigofera calcicola, and Ceratostigma minus. The lower part of the forest zone comprised temperate forests ranging from 3,300 to 4,100 m a.s.l., jointly dominated by sclerophyllous evergreen Quercus aquifolioides and coniferous Pinus densata in a patchy pattern. The upper part of the forest zone started from ∼4,100 m, mostly dominated by Picea likiangensis var. balfouriana, and extended to the alpine treeline at ∼4,500 m a.s.l. Alpine dwarf shrubs and cushion vegetation patches extended up to 4,900 m a.s.l., with Sabina pingii var. wilsonii, several species of Rhododendron, and Arenaria densissima as the dominant species. On the east slope, the shrubs extending in the dry-hot valley at 2,500–3,100 m a.s.l. were dominated by Sageretia horrida and Bauhinia brachycarpa; The forest zone at 3,100–3,900 m a.s.l. comprised broad-leaved forest patches dominated by Q. aquifolioides, and the subalpine forests at 3,900–4,300 m a.s.l. were generally dominated by Abies georgei var. smithii. The alpine vegetation at 4,200-4,700 m is composed of patches of shrubs and meadow. The evergreen R. nivale and deciduous Potentilla fruticosa were dominant shrub species, whereas multiple species of Pedicularis, Saussurea, Primula, Poa, and Carex comprised the colorful alpine meadows.

The mean plant species density (per 100 m²) showed two peaks at similar elevational patterns on the east and west slopes of Hongla Mountain (Supplementary Figure 1). The first peak appeared at 3,100–3,300 m a.s.l. and was mainly induced by a peak of SR in shrubs and herbaceous species, whereas a richness peak mainly caused the second peak only in herbaceous species. In contrast, tree SR, which was very low compared with that of shrubs and herbs, showed a standard hump-shaped pattern along the elevational gradient, peaking at 3,900–4,300 m a.s.l. in the mixed forests of deciduous and coniferous species.

For woody plants measured at the transect scale (1,000 m²), the elevational patterns of SR on the west and east slopes of the mountain both revealed a nonstandard hump-shaped pattern with two peaks. A primary peak similarly appeared at the boundary between the valley shrub zone and the forest zone (3,100 and 3,300 m on the east and west slopes, respectively), and a secondary peak just above the alpine treeline (∼4,300 m a.s.l.). SR of the east slope was higher than that of the west slope at the same elevation, except in the alpine vegetation above 4,300 m, where it was lower (Figure 3B). After controlling for the effect of SR, the pattern of Corrected PD showed an increasing trend along with elevation, with two peaks on the western slope and three peaks on the eastern slope. The trend of Corrected PD turned decreasing above 4300 m a.s.l. on both slopes (Figure 3C). The elevational patterns of Corrected FD increased upward from the dry valleys, peaked at around the alpine treeline, and then turned to decrease at higher elevations (Figure 3D).

Measured by the SESMPD, the community structure of phylogeny and function showed comparable elevational patterns on both aspects of the mountain (Figure 4). In general, the phylogeny and function of the shrub communities in the dry-hot valleys tended to be clustered (or converged) rather than random; at mid-elevations, the community structure of phylogeny and function changed gradually from clustered to over-dispersed. At elevations above the alpine treeline, the community phylogeny remained over-dispersed, whereas the function turned to be clustered again. These contrasting patterns between the phylogenetic and functional traits structure occurred on the west and east slopes.

Concerning the six functional traits of woody plants, the CWM values of five traits showed unimodal elevational patterns on both the west and the east slopes; among which MH, LT, and BT had similar patterns, peaking in the subalpine conifer forests, whereas SLA showed an opposite pattern, with its minimum value occurring at the same elevation. In contrast, peak CWM values of LA occurred in the middle of the forest zone, corresponding to the temperate mixed forests of deciduous and coniferous species. The WD value decreased from low to high elevations (Figure 5). When the complete elevational gradient was classified into three sections (i.e., low-elevation shrublands, mid-elevation forests, and alpine shrubs and meadow), the (log-transformed) median value of the six functional traits were all significantly different (P < 0.05) among the sections, and the values of MH, LA, SLA, BT, and WD of non-forest communities were all higher in the dry valleys than in the alpine habitats, whereas LT was higher in the alpine than in the dry valley habitats (Figure 6).

The CVs of six functional traits from large to small were LA, MH, BT, LT, SLA, and WD. The CVs of alltrait values were larger at mid-elevations than in the dry-hot valleys and cold alpine habitats (Supplementary Figure 2). Compared with the mid-elevation forest communities, the CV values of six traits were not significantly different between the non-forest sections at lower and higher elevations (Supplementary Figure 3).

The elevational patterns of SR on both aspects of Hongla Snow Mountain differed from either the declining pattern typical of humid tropical mountains (Homeier et al., 2010; Clark et al., 2015; Worthy et al., 2019), or the hump-shaped unimodal pattern often observed in temperate regions (Rahbek, 1995; Shen et al., 2004). Instead, it showed a bimodal distribution (Figure 3B), which was similar to the results of the surrounding area (Wang et al., 2007; Wu et al., 2014; Yang et al., 2016). The primary peaks in SR on the east and west slopes occurred at 3,100 and 3,300 m, respectively, representing the ecotones between dry-hot valley shrubland and the mid-elevation forest belt. The difference between the two aspects of the mountain was that the lower part of the forest belt comprised mixed deciduous-coniferous forests and the sclerophyllous evergreen broadleaf Q. aquifolioides forest, respectively. The secondary peaks in SR occurred at the alpine treeline ecotone, the upper boundary of the forest belt. It is notable that, from the valley bottom upward, the elevations of vegetation belt boundaries on the west slope were all higher than those on the east slope, which appears to be determined by the elevation of the valley bottom, an indicator of the mass elevation effect on climate (Zhao et al., 2015). The trend of Corrected PD on the east and west slopes is not consistent (Figure 3C), and the absence of a peak in the low-elevation dry-hot valley on the west slope compared to the east slope may be related to the severe drought stress on the west slope. In tropical mountains areas, FD generally decreases with increasing elevation (Ding et al., 2019; Worthy et al., 2019). However, in the current study, the trend in Corrected FD showed non-standard hump-shaped curves with smaller values only in the lower-elevation section (Figure 3D). This may be related to the convergent evolution due to intense drought stress at low altitudes.

The SR values of shrubland in the valley on the east slope were higher than those on the west slope. Meanwhile, plant functional traits, including MH, LA, BT, and WD, also showed larger variability on the east slope than on the west slope. This indicated that environmental stress at low elevations on the west slope is more intensive, consistent with conditions on Baima Snow Mountain in the central section of the Ningjing-Yunling Mountain range (Yang et al., 2016). The SR, Corrected PD, and Corrected FD values of forest communities on the east slope were also higher than those on the west slope, which might be related to the higher precipitation in this section, especially in the upper part of the forest belt (Zhang, 1998), as also found by Yang et al. (2016). However, the SR, Corrected PD, and Corrected FD values of alpine vegetation on the west slope were higher than those on the east slope. In the field, the alpine landscape on the east slope was characterized by shrub patches scattered among homogeneous alpine meadows, whereas, on the west slope, the alpine landscape comprised amore heterogeneous stony scrubland with smaller meadow patches (Figure 3A). This difference in the heterogeneity of micro-topography could have a significant role in assembling different plant communities at different elevations.

Hongla Snow Mountain has an elevational vegetation spectrum typical of the Hengduan Mountains region, which comprises a mid-elevation forest belt with shrub communities at high and low elevations (Liu et al., 1985). In this study, the community structure of phylogeny and function tended to cluster on both the west and east slopes at low elevations (Figure 4), suggesting environmental filtering as a dominant driver of community assembly. This is generally linked to the typical hot and arid climate in the valleys, which forms a unique “valley savanna” type vegetation dominating arid valleys along the main rivers in the Hengduan Mountain region (Jin and Ou, 2000; Yang, 2007). The dry-hot valley environmental context results from interactions among historical, geomorphological, and anthropogenic drivers (Xu et al., 1985; Zhang et al., 1997; Shen et al., 2016), and has significant effects on the flora composition, phylogeography, and plant trait evolution (Liu et al., 2011). The community structure of phylogeny and function at mid-elevations went from random to over-dispersed. This is generally interpreted as enhanced interspecific competition having a dominant role in community assembly under suitable mid-elevational environmental conditions. Plants living above the alpine treeline face extreme freezing stress; thus, environmental filtering has a prominent role in community assembly, as supported by unique evidence of plant trait adaption, such as flower morphology evolution (Peng et al., 2012) and plant camouflage (Niu et al., 2017). However, unlike at low elevations, the over-dispersed phylogenetic structure in the alpine communities implies a distinct assembly process, which could be linked to facilitation among species (Callaway et al., 2002; Schoeb et al., 2012). For example, studies in the surrounding areas have also discovered interspecies facilitation that helps plants to adapt to the extreme alpine climate and maintain SR (Luo et al., 2016). Yang et al. (2010) found that associations between the cushion plant Arenaria polytrichoides (Caryophyllaceae) and other (less related) plant species indicated a positive effect of species coexistence in these communities.

The CWM and CV of the functional traits were higher in the mid-elevation forest zone than in the non-forest vegetation zones, and the community structure of phylogeny and function were random to over-dispersed, indicating that environmental stress was weak and that interspecific competition was dominant in community assembly (Kluge and Kessler, 2011; Montano-Centellas et al., 2020). In contrast, the community structure of function was clustered in the dry valley and alpine habitats, showing similar trends although resulting from different environmental filters. For example, low plant heights occurred in arid environments because evapotranspiration from a high canopy could not be supported by the water supply from the root system (Hartmann, 2011), whereas tree height was limited in the alpine environment by the temperature threshold that blocked the formation of the plant cell (Körner, 1998). Small leaf area was also common but induced by distinct mechanisms. To avoid excessive transpiration and water loss in arid environments, plants tended to form a small leaf area (Shields, 1950), whereas, in a low-temperature environment, such as alpine meadows, the growing season for leaves was short. Plants can maintain a stable photosynthetic carbon acquisition capacity by increasing the nitrogen content of leaves, whereas the cost of respiration and transpiration of leaflets is lower (Reich et al., 1992). The smaller mean BT is mainly related to stem size (Lawes et al., 2011; Rosell et al., 2017). Shrubs dominated in the dry valley and alpine habitats, whereas trees had a much larger DBH; thus, the CWM value of BT was generally much higher in forests. In contrast, woody density showed a clear decreasing trend with increasing elevation (Chave et al., 2006). The wood density of drought-tolerant species was higher (Niinemets, 2001; Poorter and Markesteijn, 2008), whereas drought-intolerant species tended to have the lower woody density to store water in the stem (Markesteijn and Poorter, 2009), which was related to increasing precipitation with elevation.

Ecosystems in dry-hot valleys and alpine regions are subjected to strong environmental stresses and, thus, are very sensitive to climate change and natural disasters (Yand et al., 2007; Kim and Donohue, 2013). Species living in such environments have evolved over millions of years to adapt to such unique and extreme conditions (Engler et al., 2011). Therefore, anthropogenic global warming at an unprecedented rate, accompanied by extreme climatic events, poses particularly significant threats to these ecosystems and the species living in them. Thus, the precise identification of such threats, and protection of these ecosystems using suitable and sustainable approaches should take priority in biodiversity conservation policies and action plans.