Abstract
Several studies have shown that plant height changes along environmental gradients. However, altitudinal patterns of plant height across species are still unclear, especially in regions sensitive to climate change. As canopy height decreases dramatically near the tree line in alpine areas, we hypothesize that plant height across all species also decreases with increasing altitude, and distinct thresholds exist along this gradient.
Using a large dataset of maximum plant height and elevation range (400 to 6000 m a.s.l.) of 4295 angiosperms from the regional flora of the Tibetan Plateau, we regressed plant height for every 100 m belt against elevation to explore the relationships. To identify the approximate boundaries where dramatic changes in plant height occurs for herbaceous plants, shrubs, trees, woody plants and all angiosperms, we used piecewise linear regression. Phylogenetically independent contrast was used to test the potential evolutionary influences on altitudinal patterns at the family level.
Results showed that for herbaceous plants, shrubs, trees, woody plants and all angiosperms, plant height decreases significantly as altitude increases. In addition, we found that altitude, a proxy for many environmental factors, had obvious thresholds (breakpoints) dictating patterns of plant height. The results of phylogenetically independent contrast also emphasized the importance of evolutionary history in determining the altitudinal patterns of plant height for some growth forms. Our results highlight the relative intense filtering effect of environmental factors in shaping patterns of functional traits and how this could vary for different ranges of environmental variables.
INTRODUCTION
Understanding patterns and underlying mechanisms of plant functional traits along environmental gradients (Hardy and Senterre 2007) is key to both community ecology (McGill et al. 2006) and macroecology (Swenson and Enquist 2007). Functional traits have frequently been used to explain how species survive in an environment or how they are filtered by the environment (Pavoine et al. 2011; Stahl et al. 2014; Stubbs and Wilson 2004; Violle et al. 2007). Disentangling the relationship between traits and environmental factors is a powerful approach to understanding species distribution and the underlying ecological processes (Baraloto et al. 2010; Kraft et al. 2008; Lamarque et al. 2014; Reich et al. 2014; Wright et al. 2004).
Maximum plant height, often a proxy for plant body size, is a key functional trait indicating variability in plant form, function and diversity (Swenson and Enquist 2007; West et al. 1999). Plant height is a crucial strategy for plants to compete for light while adapting the vertical structure (Falster and Westoby 2003), and also an important trait associated with other functional traits, such as flowering phenology, seed reproduction and seed dispersal (Cornelissen et al. 2003; Moles and Leishman 2008; Sun and Frelich 2011; Thomson et al. 2011). The phenomenon of changing plant height with elevation has been studied extensively (e.g., Smith 1980; Macek and Lepš 2008). The altitude where vegetation canopy height changes dramatically is often defined as ‘tree lines’ or ‘grass lines’ in alpine mountain ecosystems, and these lines have been found to be sensitive to climate change (Kullman 2007; Payette 2007). The sensitivity of these regions to climate change make them a key and interesting ecotone for ecological studies (Kullman and Oberg 2009; Peringer and Rosenthal 2011; Trant et al. 2011). It is widely recognized that canopy height declines significantly as elevation increases, especially at tree lines or grass lines. However, most plant species observed in the communities of alpine forest ecosystem around ‘tree lines’ are gymnosperms and are limited to few genera, such as Picea, Pinus or Abies (Kullman 2005; Kullman 2007; Liang et al. 2011). Currently, there is little knowledge on altitudinal patterns of plant height across species in angiosperms, the most diverse and dominant plant group in global terrestrial ecosystems.
Using a comprehensively compiled dataset of plant heights and their elevation distribution records ranging from 400 to 6000 m a.s.l. on the Tibetan Plateau (Zheng and Yao 2006), we aimed to answer two fundamental questions: (i) How plant height changes with altitude between different growth forms? and (ii) Are there evident breakpoints for plant height along the altitudinal gradient?
METHODS
Data source
The study site is the Tibetan Plateau located in southwest China with a latitude ranging from 26°44′ to 36°32′N and a longitude ranging from 78°25′ to 99°06′E and (Fig. 1). We gathered the height and altitudinal range for all flowering plant species recorded on the Tibetan Plateau. Plant maximum heights of angiosperms were compiled from Flora of Xizangica (FOX; Wu 1983–1987), and the elevational distribution records for each species were compiled from FOX and The Vascular Plants and Their Eco-geographical Distribution of the Qinghai-Tibetan Plateau (Wu 2008). Cultivated and alien species were excluded from the analysis. A total of 4295 species (varieties and sub-species were treated as independent species) plant height and elevational distribution were used in our analyses, including 3127 herbaceous plants, 813 shrubs and 355 trees, respectively. The number of species analyzed in this study accounted for more than 70% of all angiosperms on the Tibetan Plateau. The species were matched to the Angiosperm Phylogeny Group III (APG III) classification (Bremer et al. 2009).
The location of the Tibetan Plateau.
Statistical analyses
We divided species elevation range into 100 m elevation belts (from 400 to 6000 m a.s.l.), and statistical analyses were conducted for each belt. Plant height and elevation were regressed using linear models for five growth forms: (i) herbaceous plants, (ii) shrubs, (iii) trees, (iv) all woody plants, and (v) all angiosperms. To test the existence of breakpoints along the elevational gradient, we used piecewise linear regression (Toms and Lesperance 2003) to identify the approximate boundaries where dramatic changes of height were found. The definition of these breakpoints along the elevational gradient was performed using the function ‘piecewise.linear’ in the R package ‘SiZer’ (Sonderegger 2012). The Pearson’s correlation coefficients between elevation and all records (r) as well as the two different subsets (as identified by the piecewise regression) separated by breakpoints (r1 and r2) were calculated (Fig. 2).
Boxplot of plant heights for herbaceous plants, shrubs, trees and all angiosperms. The sample sizes are 3127, 813, 355 and 4295, respectively.
Phylogenetic relatedness between species could also influence the results (Felsenstein 1985). The trait of one species might not be regarded as independent in statistical analyses so phylogeny must be taken into account (Swenson and Enquist 2007; Zhang et al. 2011). Since there is still no completed phylogenetic tree for all the species, we tested the potential influence of phylogenetic independence at family level. The mean height for all the species in a family were correlated with the mean elevation of those species. Pearson’s correlation coefficients between these two variables were computed among different growth forms for all families. The herbaceous plants, shrubs, trees, woody plants and all angiosperms investigated in this study belongs to 75, 64, 55, 89 and 129 families, respectively. We constructed a phylogenetic trees using ‘Phylomatic’ based on ‘phylomatic’ tree (R20100701), and used the ‘bladj’ in Phylocom (http://www.phylodiversity.net/ phylocom; Webb et al. 2008) to assign branch lengths for each family using the age provided by Wikstrom et al. (2001). The remaining unresolved clades were treated as polytomies in our supertree, and these polytomies were resolved randomly using the function ‘multi2di’ in the R package ‘ape’ (Paradis et al. 2004). Phylogenetically independent contrasts (PIC r; Felsenstein 1985) were performed using the function ‘pic3’ in the ‘picante’ package (Kembel et al. 2010) of R program (R Core Team, 2014).
RESULTS
On the Tibetan Plateau, maximum plant heights were different between growth forms (Fig. 2) and the mean maximum plant heights were 44.5, 247.6, 1433.7 and 197.7 cm for herbaceous plants, shrubs, trees and all angiosperms, respectively. The results showed that the proportion of growth forms at species level changed significantly with elevation (Fig. 3). The proportions of woody plants (shrubs and trees) were higher than herbs at low elevations and generally decreased as elevation increased (Fig. 3).
Growth form spectra along altitudinal gradients. The elevational ranges are from 700 to 5600 m. The proportions of herbaceous and woody plants are approximately equal at 2100 m.
Generally, maximum plant height of the four growth forms and all angiosperms decreased with increasing elevation according to the results of linear regressions (Fig. 4). The two partitions of piecewise linear regression for herbaceous plants were relatively consistent, with a correlation coefficient, r = −0.386 (P < 0.001). A breakpoint was observed at 3400 m and the slope of the linear regression for data below this altitude was much gentler than above it. The Pearson’s correlation coefficients for the high- and low-altitude partitions were r1 = −0.160 (P < 0.001) and r2 = −0.318 (P < 0.001), respectively. For shrubs, in the lower altitudinal partition (<3377 m), plant height changed minimally with elevation and the absolute values of correlation were also very weak (r1= −0.038, P < 0.05). In contrast, in the higher altitudinal partition, shrubs height decreased sharply as elevation increased and the correlation became stronger (r2 = −0.333, P < 0.001). The breakpoint for trees was lower (2153 m) than for shrubs and herbaceous species. Tree height decreased significantly as elevation increased (r = −0.139, P < 0.001). Trees height did not change significantly with elevation (r1 = −0.001, P > 0.05) at lower altitudes but did significantly change at higher altitudes (r2 = −0.151, P < 0.001). For all woody plants, plant heights also decreased with increasing elevation in the lower partition. The breakpoint was 1891 m and the two correlation coefficients were r1 = −0.023 (P > 0.05) and r2 = −0.291 (P < 0.001) for higher and lower altitude partitions, respectively. For all angiosperms, correlation coefficients for all elevations and the two partitions were all significant (r = −0.331, P < 0.001; r1 = −0.218, P < 0.001; r2 = −0.216, P < 0.001), and the breakpoint was around 3331 m.
Plant height patterns along with altitudinal gradients for herbaceous plants (a), shrubs (b), trees (c), woody plants (d) and all angiosperms (e). Blue lines are linear regression lines for elevation and the plant heights of species in every100 m belt. Red lines are piecewise linear regression lines. The dot dashed red lines indicated the breakpoint elevations. Sampling sizes of statistical analyses are 22764, 5726, 1967, 7693 and 30457 for herbaceous plants, shrubs, trees, woody plants and angiosperms, respectively. *P < 0.05; ***P < 0.001; ns, not significant.
The analyses at family level showed that altitudinal patterns of plant height were not always consistent across taxonomical ranks, but the trends of decreasing plant height with increasing altitude was consistent though not always significant (Figs. 4 and 5). Heights of shrubs and trees showed a non-significant negative relationship with elevation at the family level (P > 0.05). However, after taking in to account phylogenetic non-independence using PIC, the correlation coefficient between the plant heights of trees and elevation became significant (PIC r = −0.324, P < 0.05), but did not change for shrubs (PIC r = −0.159, P > 0.05).The relationship between mean plant height of families and mean distribution elevations for herbaceous plants was also significant (P < 0.01), but the PIC result was not significant (P > 0.05). For woody plants and all angiosperms, these relationships were all significant either using original data or PIC (r = −0.451, PIC r = −0.258 for woody plants and r = −0.502, PIC r = −0.432 for angiosperms).
Correlations between plant heights and elevation across growth forms and all angiosperms (a, c, e, g and i are Pearson’s correlations; b, d, f, h and j are phylogenetically independent contrast correlations at family level). Original data were transformed as log (y+1), including values of plant height and elevation. *P < 0.05; **P < 0.01; ***P < 0.001; ns, not significant, P > 0.05. All data were log-transformed (the unit for plant height is centimeter, and for the altitude is meter). Abbreviations: HP = herbaceous plants; WP = woody plants.
DISCUSSION
A goal of the following study was to test whether plant height changes with altitude and whether this varies by growth form. Previous studies of plant height along an altitudinal gradient have focused on only one or a limited number of species (Fernandez-Calvo and Obeso 2004; Macek and Lepš 2008; Moles et al. 2009). Moles et al. (2009) showed that latitude is a better predictor for plant height patterns at global scales than altitude. Previous studies also show that globally at higher latitudes there are a greater number of small plant species than in lower latitudes (Aarssen et al. 2006; Niklas et al. 2003). Using a comprehensive dataset of plant heights and their altitudinal distribution from a geographically limited plateau, our results showed that for angiosperms altitude was a good explanatory variable for the spatial distribution of plant height. However, the patterns were variable between different growth forms (Fig. 4).
Several hypotheses have been proposed to explain spatial patterns of plant height at fine scales including the hydraulic limitation hypothesis (Ryan et al. 2006), competition for light in vertical structure (Falster and Westoby 2003), and ‘physical space niche’ limitation (Aarssen et al. 2006). Elevation is often a complex indirect gradient with many environmental variables (Fierer et al. 2011), and along altitudinal gradients, many climatic variables change predictably, such as temperature, radiation and other variables important to plant growth and survival (Körner 2007). For example, temperature has shown very similar trends with changing latitude and altitude (Halbritter et al. 2013). In these hypotheses, tall plants might be filtered by limited water and energy, and space becomes a limiting factor at high altitudes. In other words, altitude can be used as a proxy for climate, soil, radiation and other environmental variables. The results demonstrate that altitudinal gradients and associated variables have a significant impact on plant height. Changes in growth form along environmental gradients are considered a key survival strategy for plants to shift into freezing-prone environments (Zanne et al. 2014). Since the ratio of plant growth form changes with altitude (Fig. 3) and plant height differs significantly by growth form (Fig. 2), plant height along altitudinal gradients might be also controlled by this key survival strategy.
Another goal of this study was to test whether the existence of thresholds as environmental filters or breakpoints for plant height along altitudinal gradients at the species level as well as ‘tree lines’ and ‘grass lines’. Based on the results of piecewise linear regressions, for shrubs, trees, woody plants and all angiosperms, we found visible thresholds shaping the altitudinal patterns of plant height. Generally, plant height decreased with increasing elevation significantly. However, these relationships were non-existent or non-significant in low-altitude regions for trees and woody plants (Fig. 4). This suggests that in lower regions, the variations of environmental variables associated with elevation had limited effects on plant maximum height. In contrast, at higher altitudes, the effect of environmental filtering was much stronger, especially for shrubs and woody plants (Fig. 4b and d). Our results show that on the Tibetan Plateau, two elevation belts exist and should be taken into account in ecological research of the area. One is the belt ranging from 1800 to 2200 m, where the heights of trees and woody plants changed significantly. The other belt is from 3300 to 3400 m, which is also a transition zone for plant height of herbaceous plants and shrubs. Many plant functional traits are highly correlated with environmental variables, such as leaf economics spectrum, seed mass, leaf nitrogen and woody density (Chave et al. 2009; Shipley et al. 2006; Swenson and Weiser 2010; Wright et al. 2004). However, previous studies showed that altitude is not a good predictor for some key plant traits, including plant height (Moles et al. 2009), seed mass (Moles et al. 2007), wood density (Swenson and Enquist 2007) at global scales. It has been reported that woody density and modulus of rupture were not related to altitude <1000 m, but decreased at altitudes >1000 m (Zhang et al. 2011). The underlying mechanisms could be similar for plant height. The reason that altitude is not considered a good predictor for some key plant functional traits at the global scale is because the latitudinal range of datasets is often too broad and most elevation records don’t exceed thresholds that we identified for plant height. Tree lines at high elevations are sensitive to climate change (Grace et al. 2002; Harsch et al. 2009), thus above thresholds proposed by our results some ecological processes influencing plant height may change dramatically under rapid climate change. More detailed examination is needed around these two potential ecological thresholds, rather than only focusing on tree lines.
The results from PIC analysis show that at the family level, phylogeny of angiosperms influences the patterns of plant heights along altitudinal gradients for some growth forms. The statistical results show that plant heights tend to decrease with increasing elevation at the family level but not for shrubs and trees. At high elevations, environmental filtering are stronger and species phylogenetic relationships and functional traits are more clustered (Pavoine et al. 2011). Additionally, most angiosperms at high elevation are herbaceous. That could explain why the heights of herbaceous plants have weak latitudinal gradients in alpine areas when phylogenetic non-independence is taken into account.
Our results showed that plant height at the species level decreased significantly with increasing altitude at the regional scale. The research area in our study was limited to the Tibetan Plateau where latitude spans ~10°. Climate conditions also change along latitudinal gradients (Halbritter et al. 2013; Stephenson and Das 2011), which could influence the robustness of our results. However, most species observed in our study were distributed in the southern Tibetan Plateau, especially in the Himalaya region (~3° of latitude). Therefore, we believe that the impact from changing of environmental conditions associated with latitude on our results is limited. However, two other aspects need to be improved in future work. Firstly, to test our hypotheses the results could be more robust if the variations of plant height at the intra-species level are considered. Secondly, there are a limited number of phylogenetic trees which cover many plant species. This could lead to some gaps in the results based on different taxonomic ranks when considering the influence of phylogenetic non-independence. Our results indicate that plant height patterns might be influenced by evolutionary history at family level but it is still unclear how it performed at the species level. Therefore, more detailed analyses based on phylogenetic trees with higher resolution at the species level are needed in future studies.
ACKNOWLEDGEMENTS
We are grateful to the constructive comments and suggestions from the editor and two anonymous reviewers. We also thank the financial support from The Fundamental Research Funds for Central Public Welfare Research Institutes (2016) for Shengbin Chen. We would also like to thank Alison Beamish at the University of British Columbia for her assistance with English language and grammatical editing of the manuscript.
