Rachael V. Gallagher, Department of Biological Sciences, Macquarie University, North Ryde, NSW, 2109, Australia. E-mail: email@example.com
Aim Understanding large scale patterns in trait variation in climbing plants (lianas, vines, scramblers, twiners) is important for the development of a stronger theoretical understanding of climbing plant ecology and for more applied issues such as prediction of community assembly under changing climatic conditions. We compared values of five key functional traits for 388 species of climbing plant from tropical and temperate regions of Australia to quantify variation between these two biogeographic regions.
Methods Data on dispersal mode, growth habit, leaf form, leaf size and seed mass were compiled from field measurements and published sources. Comparative analyses were performed in three ways: (1) across species where each species was treated as an independent data point, (2) using evolutionary divergence analyses for each trait, and (3) in multidimensional space using a matrix of similarities between species.
Results Tropical climbing plants had 22-fold greater seed mass and four times greater leaf size than did temperate species. Tropical climbers were more likely to be woody (63%) than were temperate species (40%). Surprisingly we found a similar proportion of animal-dispersed seeds in the two regions, although we expected animal-dispersed seeds to be more prevalent in the tropics. We also found similar proportions of simple- and compound-leaved species between the two regions. All of our findings were consistent between cross-species and phylogenetic analyses indicating that patterns in present-day species are reflected in the evolutionary history of Australian climbers. Multivariate analyses suggested that there is a spectrum of variation among climbing plants, with tropical species having greater seed mass, leaf size and woody growth compared with temperate climbing plant species.
Main conclusions Tropical and temperate climbers of Australia exhibit a mixture of similar and contrasting traits and ecological strategies. Understanding strategy variation along latitudinal gradients will be particularly informative for predicting ecosystem and community structure with climate change.
Functional traits provide insight into how organisms gain, deploy and capture resources within the constraints of their physical environment. In trait-based comparative studies emphasis is placed on how the physical characteristics of species, rather than their identity, vary between sites or assemblages. As a result, important ecological variation that may otherwise have been hidden in comparisons based on taxonomic diversity, can be captured and related to the emergent properties of ecosystems, such as carbon cycling (Cornwell et al., 2008) or above-ground net primary productivity (Garnier et al., 2007). This approach has offered insight into a wide range of questions in plant ecology in recent decades, including predicting the responses of communities to disturbance (Díaz et al., 2001; Keith et al., 2007), evaluating differences in native, naturalized and invasive species (van Kleunen et al., 2010; Leishman et al., 2010) and defining trade-offs that shape major strategy dimensions of ecological variation in leaves (Ackerly & Donoghue, 1998; Wright et al., 2004), seeds (Leishman, 2001; Moles et al., 2005) and wood (Chave et al., 2009).
Extending this trait-based approach to the investigation of tropical and temperate climbing plant species is crucial for our understanding of large-scale patterns of variation in this ecologically important group. Furthermore, the practical importance of identifying gradients in climbing plant traits will become increasingly clear as climate change progresses. For example, temperature increases resulting from anthropogenic warming have been associated with the movement of tropical species towards the poles (Walther et al., 2002; Parmesan, 2006) and may promote the establishment of tropical climbers in temperate forests in the coming decades. The ability to predict the impact of these potential changes in species composition using functional trait data may aid forest management in the future. Adopting trait-based approaches to understanding the ecology of climbers in temperate compared with tropical regions will complement the extensive literature about variation at the species level (reviewed in Schnitzer & Bongers, 2002; Isnard & Silk, 2009), and facilitate the development of trait-based models that predict vegetation dynamics for this functional group (Lavorel & Garnier, 2002).
The climbing habit has evolved multiple times throughout plant evolution and has been described as a ‘key innovation’ due to its link to radiations in species richness across the angiosperm phylogeny (Gianoli, 2004). On average, climbers contribute around a quarter of the woody stem biomass and species diversity in tropical lowland systems of the world (Schnitzer & Bongers, 2002). In sites outside the tropics the diversity of climbers drops to around 10% of all woody species and abundance differs by an order of magnitude between northern temperate forests and most tropical sites (Gentry, 1991). Despite these differences in abundance between tropical and temperate regions there have been no systematic investigations of how the functional traits and ecological strategies of climbing plants differ between these regions.
In this study we used a comparative trait framework to examine differences in the ecological strategies of tropical and temperate climbing plants in Australia. Australia is divided roughly in two by the Tropic of Capricorn, with a monsoonal climate in the north and temperate seasonal climate in the south. Average temperatures range between 21 and 27 °C north of the tropic and between 2 and 21 °C south of the tropic (for mean annual temperature maps see Australian Bureau of Meteorology website: http://www.bom.gov.au/jsp/ncc/climate_averages/temperature/index.jsp?maptype=6&period=an#maps). The tropics are characterized by open savannas, monsoon forest, wetlands and tropical rain forests, whereas temperate areas are dominated by eucalypt woodlands and hummock grasslands and there is a high diversity of climbing plants in both regions (George et al., 1982). Five functional traits that influence how plants establish, gain carbon and convey fitness to the next generation were used to quantify strategy differences between climbers of tropical and temperate regions. We chose a mix of regenerative traits (seed mass and dispersal mode) and traits relating to the deployment and support of photosynthetic surfaces for carbon capture (leaf form, leaf size and growth habit). Below we outline briefly the reasoning for the selection of these five traits (seed mass, dispersal mode, leaf form, leaf size and growth habit) and the specific hypotheses tested.
Seed mass and dispersal mode
Seed mass and dispersal mode are central to our understanding of plant fecundity and seedling establishment (Howe & Smallwood, 1982; Leishman, 2001; Moles et al., 2004). In plants, maternal investment in reproduction can be spent on few, well provisioned seeds with a higher chance of surviving environmental hazards during establishment, or on many small seeds whose fate rests with their quantity rather than quality (Smith & Fretwell, 1974). Coexistence of these differing seed mass strategies within communities is achieved through the integration of reproductive investment over the lifetime of a species, such that the apparent advantage of increased fecundity in small-seeded species is balanced by the longer life spans and larger canopies associated with large-seeded species (Moles et al., 2004). Hence seed size provides a vital link between investment in reproduction and vegetative growth in plants.
Seed mass is also associated with dispersal mode (Leishman et al., 1995), with seeds over 100 mg often being moved around the landscape by vertebrates, and smaller seeds more likely to be wind dispersed or experience unassisted dispersal (Hughes et al., 1994). Dispersal mode is an informative surrogate for estimating the distance a propagule can travel, and its colonization potential in the landscape (Petit et al., 2003). In addition, the provision of nutritious rewards, such as fleshy fruits or elaiosomes, in return for dispersal mediates mutualistic interactions within communities (Burns, 2004).
We offer two hypotheses about the regeneration strategies of climbing plants. Firstly, we expect that tropical climbers will possess larger seeds than do temperate species. Secondly, we propose that a higher proportion of species will be dispersed by animals (biotic dispersal) in tropical regions than in temperate regions. Larger seeds may facilitate the transition from heterotrophic to autotrophic growth in light-limited environments such as closed canopy forests, which are more prevalent in the tropics. Across a range of plant growth forms, vertebrate-dispersed species have been found to be more common in the tropics, and temperate species are more likely to be wind dispersed (Lord et al., 1997; Moles et al., 2007). Equally, seed mass and dispersal mode may simply be linked along an ecological strategy dimension, with larger seeds more likely to be dispersed by animals than by wind.
Leaf form and leaf size
Leaves can be either simple, with a single lamina arising from a node, or compound, consisting of repeated leaf units supported on a rachis. Leaf size ranges over six orders of magnitude globally (Wright et al., 2007). Selection may favour leaf forms and sizes that maximize water use efficiency (Parkhurst & Loucks, 1972), optimize leaf temperature through convective heat loss (Gates, 1980; Niinemets, 1998; Vogel, 2009), increase light-foraging efficiency (Givnish, 1978; Falster & Westoby, 2003) or minimize herbivore damage (Moles & Westoby, 2000) in any given environment. Alternatively, leaf size may simply be traded-off against leafing intensity (the numbers of leaves produced per unit shoot volume) (Kleiman & Aarssen, 2007). In addition, leaf size is correlated with twig size, plant height and the size of seeds and infructescences, making it a central trait in the allometry of plants (Ackerly & Donoghue, 1998). Compound leaves have been shown to be selectively favourable in water-limited environments where the dissection of simple leaves into individual leaflets reduces effective leaf size, leading to increased convective cooling (Niinemets, 1998). Leaf size also varies across environmental gradients, being smaller in sites characterized by extreme conditions, such as high radiation or low water availability (Webb, 1968; Givnish, 1987; Pickup et al., 2005). In environments characterized by high resource availability, such as the tropics, selective pressure to reduce leaf size to conserve limiting resources such as water may be eased. Therefore we expected that tropical climbing plant species will, on average, possess larger leaves than do temperate species, and that simple leaf forms will be favoured in tropical environments.
Herbaceous species tend to have higher relative growth rate, specific leaf area, and net assimilation rate and lower basal areas than do woody species (Hunt & Cornelissen, 1997). Herbaceous and woody climbers also respond differently to abiotic factors. For instance, the abundance of woody climbers in tropical regions is negatively associated with mean annual rainfall and positively with seasonality of precipitation (Schnitzer, 2005), yet the abundance and richness of tropical herbaceous climbers shows no significant relationship to climatic factors (Bhattarai & Vetaas, 2003). There is also evidence that the abundance of woody climbers is negatively correlated with extremes of minimum temperature, which render lianas more susceptible to freezing-induced embolisms in the xylem (Jiménez-Castillo et al., 2007). Although sub-zero temperatures are not a common feature of the Australian climate, largely due to the very low relief of the continent, minimum temperatures do decline markedly between tropical regions in the north and temperate regions in the south (for minimum temperature maps see the Australian Bureau of Meteorology website:
We hypothesized that differences in the seasonality of rainfall and minimum temperatures between tropical and temperate regions of Australia would lead to a higher proportion of woody climbers in the tropical zone.
In summary, we use data for these five traits for 388 species to address the following questions.
1 How do the values for each individual trait differ in cross-species analyses of tropical and temperate climbers? On average we expect tropical climbers to have a higher proportion of species with larger leaves and seeds, woody growth habits, simple leaf form and biotic dispersal.
2 How does shared ancestry affect these univariate patterns? This allows us to assess the role that phylogenetic history plays in driving present-day patterns in cross-species relationships.
3 What is the pattern of multivariate variation when all five traits are considered in concert? This allows us to determine which combinations of traits co-vary and whether the resulting axes of variation differentiate tropical and temperate climbing plant species.
Materials and methods
We compiled a list of 388 Australian vascular plants described as possessing a climbing habit from herbarium collections, published floras and plant databases. Climbers were assigned to two groups – temperate or tropical – on the basis of distributional information from Australia’s Virtual Herbarium (AVH; http://avh.rbg.vic.gov.au/avh/public_query.jsp). We defined ‘tropical’ species as those whose geographic range was entirely contained north of the Tropic of Capricorn (23°30′ S) and ‘temperate’ species as those whose entire geographic range was south of the Tropic. We deliberately avoided including any species that occurred in both the tropical and temperate regions of Australia to ensure we were contrasting species adapted to abiotic conditions in each region. In total 220 species were classified as tropical and 168 as temperate for this analysis (see Appendix S1 in the Supporting Information). Nomenclature followed the accepted conventions of the Australian National Herbarium in Canberra.
Data on mean dry seed mass were primarily sourced from the Royal Botanic Gardens Kew Seed Information Database (SID; http://data.kew.org/sid/) and Millennium Seed Bank projects in New South Wales and Western Australia, and were supplemented with datasets collected in the field by the authors and via personal communications with J. Lord (University of Otago, New Zealand) and R. Kooyman (Macquarie University, Australia). This yielded seed mass data for 90 species. Due to the scarcity of species-level seed mass data for climbing plants, measurements for a further 115 species were estimated from the average seed mass of their genus using the equation: log (seed mass) = 0.864 × log (average seed mass) + 0.280 (r2 = 0.89, P <0.001, n =43). This relationship was developed using the average seed mass of the genus as the predictor variable and known seed mass as the response in a linear regression model using data taken from SID. Only genera for which seed mass data for 10% or more of the constituent species were available were used in the calculation (see Appendix S2 for further explanation of the technique used to estimate seed mass). To confirm that the patterns revealed using the genus level data were consistent with the 90 species for which seed mass was known we analysed both data sets and report the results for each.
Field-based measurements of leaf size were used wherever available and were taken from datasets assembled by the authors and colleagues (n =20). For a further 270 species, leaf size was calculated using length and width measurements compiled from floras and published literature, using the equation area = length × width × 0.7 (r2 = 0.98, P <0.001) reported in Kraft et al. (2008). This equation has been shown to estimate leaf size reliably for 742 Neotropical species and has subsequently been used to estimate the leaf size of 231 tree and shrub species from Australian subtropical rain forests (see Kraft et al., 2008; Kooyman et al., 2010). The length and width of leaves of the 20 species collected in the field were consistent with those reported in published floras. We excluded species with linear or deeply lobed leaves from estimations of leaf size using this technique (n =14). For compound leaves the size of an individual leaflet was calculated, as this is defined as the minimum photosynthetic unit and is comparable to simple leaves (Kraft et al., 2008).
Dispersal mode was coded as either abiotic (anemochory or hydrochory) or biotic (endo/exo-zoochory) and the dataset contained no species with ballistic dispersal. Species with seeds with unassisted dispersal were excluded from the dataset (n =42). Leaf form was coded as simple or compound, and species with one-foliate leaves were excluded from the dataset (n =4). We classified growth habit as herbaceous or woody according to the state of the stem at maturity. Species listed as semi-woody in data sources were included as woody in this analysis.
For all analyses seed mass and leaf size were analysed as continuous variables and were log10-transformed before analysis. The remaining three traits were analysed as binary variables.
We compared tropical and temperate climbing plants by considering each of the five life history traits in isolation. Differences in seed mass and leaf size between regions were assessed using one way analysis of variance (ANOVA). For categorical traits (dispersal mode, growth habit, leaf form) we used the chi-square goodness-of-fit test to compare the proportion of species in each region possessing a given trait state. All statistical analyses were considered significant at the 0.05 α level and were performed in spss 16.0.1 (SPSS Inc., Chicago, IL, USA).
We used evolutionary divergence analyses, also known as phylogenetically independent contrasts, to further investigate patterns identified in cross-species analyses. Primarily, we were interested in assessing whether relationships detected in cross-species analyses are conserved across the phylogenetic history of our sample of Australian climbing plants. This technique also allows us to assess if present-day patterns in trait associations result from a small number of large divergences across trait values in the evolutionary history of the species examined. For instance, a cross-species analysis that identified significantly larger seeds in the tropics may result from the inclusion of a few highly diverse tropical families or genera of climbing plants that possess large seeds.
The tree building application Phylomatic was used to array our species on a phylogenetic tree (http://www.phylodiversity.net/phylomatic/phylomatic.html). This program matches taxa on the basis of their genus or family to an angiosperm megatree compiled from published phylogenies. The phylogeny R20091120 was used as the basis for tree construction. This phylogeny is based on the Angiosperm Phylogeny Group’s most recent published phylogeny (APG3 megatree).
The values of trait divergences across nodes of the tree were calculated using the ‘Analysis of Traits’ module (AOT) in Phylocom (v. 4.1) (Webb et al., 2008) following the methods outlined in Felsenstein (1985). Phylocom AOT calculates divergences between two continuous traits or a continuous and a binary trait. As yet this program does not accommodate analyses with both binary dependent and independent variables. Therefore, for each species we replaced the binary predictor region (tropical or temperate) with a continuous measurement by substituting the midpoint of its latitudinal range. The midpoint was determined by averaging the latitude of all collection records for the species in the AVH, which is a comprehensive database of Australian plant distributions (over 6 million specimens have been digitized) that have been submitted and validated by qualified taxonomists and ecologists. It represents the collective effort of thousands of botanists over time and provides the best estimate of the known distribution of each species on the Australian continent. Divergences for traits coded as binary (dispersal mode, growth habit, leaf form) were then calculated in Phylocom using latitude as a continuous variable.
The relationships between divergence values calculated from two continuous traits (leaf size and latitude, seed mass and latitude) were analysed using a linear model in spss (v. 16.0.1). The regression line was forced through the origin to account for ambiguity in the direction of the subtraction giving rise to the divergence value (see Garland et al., 1992). For divergences based on one binary and one continuous trait (dispersal mode, growth habit, leaf form) significance was assessed using a one-sample t-test against the null hypothesis of no change in latitude across the divergence.
We used permutational multivariate analysis of variance (PERMANOVA; Anderson, 2001) to assess whether climbing plants from tropical and temperate regions occupy different positions in multidimensional space. Data were analysed using a one-factor (region) design with two levels (tropical or temperate). A similarity matrix was assembled using the 90 species for which values for all five traits were known, with traits as columns and species as rows. A multivariate similarity matrix was computed using Pearson’s correlation on normalized data. Significance values were calculated by permutation through the random reassignment of the factor levels using 9999 permutations of the raw data and differences were considered significant at the 0.05 α level.
PERMANOVA, like other multivariate methods for testing for group differences via permutation, is sensitive to the dispersion, or spread around the centroid, of the multivariate data under investigation. That is, significant differences between tropical and temperate species may be not be the result of differences in the location of the data clouds in multivariate space, but by how spread out each group is around its centroid. Therefore, we employed a test for homogeneity of multivariate dispersion (PERMDISP; Anderson et al., 2006) to identify whether differences in tropical and extratropical climbing plants are the result of difference in the dispersion of points between groups rather than their location. This test is analogous to Levene’s test (Schultz, 1985), commonly used to explore equality of variance assumptions in univariate ANOVA.
Principal components analysis (PCA) was used as an ordination method to visualize the multivariate data in a two-dimensional space. Eigenvalues associated with each axis were ranked by their absolute value and the three largest were used for interpretation of which traits were driving clustering patterns on principal component axes.
Multivariate procedures were performed in primer (v. 6) using the PERMANOVA+ add-on (Clarke & Gorley, 2006).
Differences in individual traits between regions
There were highly significant differences between tropical and temperate climbers in three of the five traits examined (Table 1). These differences supported our hypotheses about seed mass, leaf size and the proportion of woody species all being larger in the tropics. On average, seed mass was 22 times larger in tropical species than in the temperate species and was significantly different between the regions (tropical mean = 425.9 mg; temperate mean = 18.9 mg; F =68.5, d.f. = 1, 203, P <0.001, n =205 species; Fig. 1). This pattern of larger seeds in the tropics was consistent when the restricted data set of 90 species for which seed mass was known at the species level was analysed (F =19.9, d.f. = 1, 88, P <0.001). Leaf size was also significantly larger in the tropics (F =115.8, d.f. = 1, 288, P <0.001, n =290 species; Fig. 1), with leaves being four times larger on average in tropical species than in temperate species (tropical mean = 56.9 cm2; temperate mean = 12.7 cm2).
Table 1. Results of analyses across species and across phylogenetically independent contrasts (PICs) for five traits of tropical and temperate climbing plant species from Australia. In cross-species analyses continuous traits were assessed using one-way analysis of variance, and binary traits were tested using chi-square goodness-of-fit tests. In phylogenetic analyses, divergences between two continuous traits (e.g. latitude and seed mass) were assessed using linear regression with no intercept value, and divergences between one binary and one continuous trait were assessed using a one-sample t-test based on the null hypothesis of no change. Region (tropical or temperate) was the predictor variable in all analyses.
Significance values: *P <0.05, **P <0.01.
Tropical climbing plant species were significantly more likely to exhibit a woody growth habit than were temperate species (χ2 = 17.9, d.f. = 1, P <0.001, n =350; Fig. 2b). In the tropics 63% of species were woody, compared to 40% of species in the temperate region. Finally, neither the proportion of species with abiotic or biotic dispersal modes (χ2 = 1.3, d.f. = 1, P =0.25, n =263; Fig. 2a) nor the proportion of species with simple or compound leaves (χ2 = 1.2, d.f. = 1, P =0.27, n =375; Fig. 2c) differed significantly between regions, in contrast to our predictions.
Evolutionary divergence analyses
Phylogenetic analyses corroborated the findings of our cross-species analyses of differences in the traits of tropical and temperate climbing species (Table 1). For example, shifts to a tropical region across nodes in the phylogeny were consistently associated with shifts to larger leaves (106 of 145 divergences, r2 = 0.41, F =99.7, P <0.001, Fig. 3a), consistent with the findings from our cross-species analyses. Tests across divergences for dispersal mode, growth habit, leaf form and seed mass also showed similar patterns of association observed in cross-species analyses (Table 1). Seed mass and growth habit remained significantly different between regions in phylogenetic analyses (seed mass: 63 of 104 divergences, r2 = 0.06, F =6.5, P <0.01, Fig. 3b; growth habit: 25 of 36 divergences, t =2.5, P =0.02, Fig. 4) and dispersal mode and leaf form showed no difference between regions (dispersal mode: 21 of 46 divergences, t = −0.5, P =0.65; leaf form: 10 of 16 divergences, t =2.0, P =0.06, Fig. 4). These results suggest that the cross-species patterns we observe arise as a result of consistent associations between the traits and region throughout the evolutionary history of climbing plants.
Multivariate patterns in the traits of tropical and temperate climbers
We detected highly significant differences between the ecological strategies of climbers in tropical and temperate regions when all five traits were combined in a multivariate analysis of variance (PERMANOVA: pseudo F =5.6, d.f. = 1, 88, P <0.01). Scatter plots of the first two PCA axis scores show clustering of tropical and temperate species in two-dimensional space (Fig. 5). These axes combined explained 62% of the total variation in trait values across the 90 species for which data were available. The similarity in the dispersion of data points from temperate (empty circles; Fig. 5) and tropical (solid triangles; Fig. 5) regions implies that the significant differences between the two regions identified in the PERMANOVA resulted from differences in location of the data clouds in multivariate space. This was confirmed by the non-significant result of a multivariate test of dispersion (PERMDISP: F =1.1, d.f. = 1, 88, P =0.26).
PC axis 1 differentiated tropical and temperate species on the basis of seed mass, growth habit and leaf size, and to a marginally smaller extent dispersal mode. This is indicated by the sign and strength of the association of eigenvalues for these traits with this axis (Table 2). The higher end of PC axis 1 was associated with greater seed mass, woody growth habit and leaf size, and was dominated by tropical species. This interpretation supports the results presented from univariate and phylogenetic analyses that tropical climbers allocate more resources to seed mass, leaf size and structural support in the form of woody tissues than temperate climbers. The two reproductive traits – dispersal mode and seed mass – both had positive eigenvalues along PC axis 1 (0.45 and 0.53, respectively). This indicates that these two traits are coordinated along this axis so that increases in seed mass are accompanied by shifts to biotic dispersal in the tropical climbers at the higher end of the axis.
Table 2. Eigenvalues associated with the first three principal components axes (PCA) derived from the 5 trait × 90 species matrix of Australian climbing plants. Numbers in parentheses represent the total amount of variation explained by each axis. Values are ranked by their absolute value along PC axis 1. Numbers in bold represent the three highest eigenvector scores along each axis.
Axis 1 (40.9%)
Axis 2 (21.1%)
Axis 3 (14.6%)
Tropical and temperate species were not distinctly clustered along PC axis 2 (Fig. 5). This axis was strongly associated with changes in leaf form (Table 2). Leaf form exhibited the highest weighting on the second axis (0.86); however, species from both regions were similarly distributed along this axis, indicating that leaf form failed to distinguish between species from each region. This result reflects the findings of the above analyses of no difference in the proportion of species with simple and compound leaves from each region (Fig. 2c). Dispersal mode and seed mass also ranked as important variables on this axis; however, their eigenvalues were much smaller than those for leaf form.
Australian tropical and temperate climbing plants differ in some key life history traits and ecological strategies. On average, tropical climbers direct reproductive allocation towards large, well-provisioned seeds and gain structural support in the form of woody biomass. They also invest in larger leaves than do temperate species. Tropical and temperate climbers exhibit a similar proportion of simple and compound leaf forms, and abiotic and biotic dispersal mode; however, dispersal mode, along with seed mass, leaf size and growth habit, form an axis of specialization that distinguishes between climbers from each region in multivariate space.
Although this study was limited to a single continent and growth form, the clear differences in three of the traits examined reflect the results of global studies examining latitudinal patterns in fundamental life history traits. For example, Moles et al. (2007) identified declines in seed mass of a similar magnitude to those presented here in a global analysis across a range of growth forms. Vegetation type had considerable explanatory power (R2 = 0.31) in driving seed mass declines over the latitudinal gradient in the global analysis, and there is some evidence that climbers with large seeds are more likely to colonize low light microsites in forests (Gerwing, 2004). These studies support the idea that the presence of a high proportion of closed canopy forests in tropical locations contributes to seed mass differences in present-day Australian climbing plant assemblages.
The relationship between seed mass and latitude found in present-day climbing plant species is reflected in the wider evolutionary history of seed mass. Moles et al. (2005) found significant associations between latitude, and the climatic parameters that co-vary with it, such as temperature and precipitation, and the evolution of seed mass across growth forms. However, correlated divergences with other plant traits, in particular growth form, had more explanatory power in describing changes in seed mass across evolutionary divergences in their study. Divergences in growth form explained 10% of the variation in seed mass contrasts and fossil evidence points to simultaneous radiations in these two traits in the period between 85 and 65 Ma. This implies that the large differences in seed mass between temperate and tropical climbing plants identified in this study may be associated with the shift from woody to herbaceous growth form between the two regions. The coordination between growth habit and seed mass identified in our multivariate analysis lends support to this view.
We found no evidence to show that tropical climbing plants have a higher proportion of seeds dispersed by animals than by abiotic mechanisms such as wind or water. This is a striking departure from previous studies identifying strong latitudinal clines in the proportion of species exhibiting different dispersal modes (Lord et al., 1997; Moles et al., 2007) and is particularly surprising given the paucity of large, wind-dispersed climbing plant families in tropical Australia, such as the Bignoniaceae. In addition, neither wind nor animal dispersal has consistently been associated with shifts between tropical and temperate regions throughout the evolutionary history of Australian climbing plants.
One possibility that may explain the prevalence of wind dispersal in tropical climbers relates to abiotic conditions in the dry season. Dry seasons are characterized by an absence of heavy rain and the presence of strong winds (Strahler & Strahler, 1989). These conditions may favour the deployment of wind-dispersed seeds in species that can maintain growth as conditions become drier. Schnitzer (2005) proposed that woody lianas from tropical regions are able to extend their growth into the dry season by accessing deep repositories of soil moisture. Other studies have linked the high incidence of wind dispersal in tropical climbers to the dominance of deciduous species that shed their leaves and open up the canopy in the dry season (Morellato & Leitao-Filho, 1996). However, tropical vegetation in Australia is predominantly evergreen, reducing the likelihood that wind dispersal in climbers is an adaptation to leafing phenology in our study region (Bowman & Prior, 2005; Heise-Pavlov et al., 2008).
The presence of open woodland, savanna and wetland habitats in tropical Australia is unlikely to explain the persistence of wind-dispersed seeds in tropical climbers. Although wind dispersal is more common in these types of open habitats (Lorts et al., 2008), climbing plants are likely to be rarer in these sites than in closed canopy forests due to the absence of strong competition for light in open habitats. Only a small number of tropical species in our dataset are found exclusively outside of rain forests. In addition, previous studies have identified disproportionately large numbers of climbing plants with wind-dispersed seeds in closed forest habitats (Ibarra-Manríquez et al., 1991; Griz & Machado, 2001), suggesting that the tendency for climbers to have wind-dispersed seeds is independent of vegetation type.
This study joins a number of others that enumerate cross-species latitudinal gradients in leaf size (Webb, 1968; Ohsawa, 1995; Halloy & Mark, 1996); however, it is the first to quantify this pattern in climbing plants. Like these previous studies we detected a strong trend towards larger leaves with decreasing latitude. This is despite the fact that leaf size varied by between two and four orders of magnitude within each region. Unfortunately, the various different ways in which leaf size was measured across these different studies (Raunkiær–Webb size classes, ratio of leaf length to width) makes it difficult to determine whether the magnitude of increase in leaf size with latitude found for climbing species is consistent with that of other growth forms.
The relative stability of environmental conditions in tropical regions may underpin the increase in leaf size. In environments where resource availability is high, such as the tropics, selective pressure to reduce leaf size to preserve limiting resources, such as water, may be diminished. Although the relationship between leaf size and the physical environment is complex, leaves generally decrease in size when resources become limiting. For example, leaf size has been shown to decrease as temperatures become cooler along elevational (Halloy & Mark, 1996) and latitudinal gradients (Webb, 1968), and in drier, nutrient-limited sites (Fonseca et al., 2000). Mechanistic explanations for why leaves become smaller in these environments vary, although in arid sites with strong radiation and low water availability, smaller leaf sizes may lead to more efficient convective heat loss from the leaf surface. Alternatively, in colder sites small leaves may be able to heat up to optimal temperatures for photosynthesis faster due to smaller boundary layers of air. Smaller leaves may therefore represent a better strategy for maintaining optimal leaf temperatures for photosynthesis and preserving water balance in more extreme environments (Parkhurst & Loucks, 1972; Givnish, 1987). However, in tropical sites the relative abundance of rainfall, warmer temperatures and shadier conditions in dense vegetation throughout the year may ease the selective pressure towards reductions in leaf size. If climbers are able to gain access to deep repositories of soil moisture in dry times, as proposed by Schnitzer (2005), this may also be advantageous in sustaining larger-leaved lianas in the tropics (Fetcher, 1981).
An alternative explanation for larger leaves in tropical vines considers light as a limiting factor. Competition for light drives interactions between species in tropical settings (DeWalt et al., 2000) and larger-leaved species may be more capable of monopolizing light resources in the canopy through overtopping and out-shading of competitors. This may be a particularly effective strategy in climbers whose leaf size may not be constrained by branching architecture in a similar way to co-occurring trees. Climbers explore for light gaps and for hosts with leafless leading shoots and they can deploy leaves rapidly into these microsites upon detection (Putz & Holbrook, 1991). These leaves are typically arrayed in a linear fashion along a vertically oriented shoot, which may contribute to reductions in self shading (Hikosaka et al., 1994). The ability to rapidly colonize light gaps with large leaves and maintain these into the dry season may make increased leaf size a superior competitive strategy for climbers in tropical environments.
There is some evidence that species with larger leaves may be able to maintain a higher leaf mass fraction (leaf dry mass per shoot invested) at a given leaf mass per unit area (Pickup et al., 2005). Given their reduced allocation of resources to structural support, climbers may be well placed to capitalize on this strategy and investigating leaf to shoot ratios may represent an important future direction for research.
Finally, woody climbers, like all woody species, are more diverse in humid tropical regions (Schnitzer & Bongers, 2002). This may be the result of warmer minimum temperatures in the tropical regions of Australia reducing the risk of xylem cavitation in wide-vesselled, but poorly insulated stems of lianas (Jiménez-Castillo et al., 2007). The idea that lianas may be able to access deep soil moisture, thereby allowing them to extend their growth into the dry season in tropical seasonal climates, may also explain their competitive advantage over other growth forms, although evidence remains equivocal (see Andrade et al., 2005; Schnitzer, 2005).
We did not detect differences in the proportion of species exhibiting different leaf forms between tropical and temperate regions as hypothesized. Compound leaves reduce drag in high winds and this may prevent damage to the leaf surface in windy environments (Vogel, 2009). They also reduce effective leaf size, facilitating convective heat loss in moisture-limited environments (Niinemets, 1998). However, there is evidence that compound leaves are an adaptation to light competition, as they increase plant height by acting as cheap, throwaway branches (Givnish, 1978). Therefore, different selective pressures may result in the co-existence of different leaf forms in both regions.
Understanding trait variation across climatic regions is important both for our theoretical understanding of climbing plant abundance and diversity and for more pragmatic issues such as predicting community and ecosystem dynamics under changing climatic conditions. The results of this study show that while there are some similarities in the traits of tropical and temperate climbing plants there are also important differences in ecological strategy. For instance, strong decreases in leaf size and seed mass with latitude exist despite substantial variation in these two traits within each region. Seed mass and growth form, along with leaf size and to a lesser extent dispersal mode, form an ecological strategy axis that separates tropical and temperate climbers. Given these strategy differences it is imperative to understand how the extension of tropical conditions towards the poles may affect climbing plants and their hosts as climate change progresses. Other research directions that may improve our understanding of forest dynamics include trait comparisons between trees and climbers within sites and along environmental gradients.
An explosion in the availability of data on traits and species distributions in the last two decades has allowed large-scale comparative studies that address ecological, rather than taxonomic, diversity to be undertaken. Given these advances, a global analysis of climbing plant traits cannot be far off.
We are grateful to the researchers and organizations who provided access to their data, including Rob Kooyman, Janice Lord, Ian Wright, the Threatened Flora Seed Centre at the Western Australian Herbarium, and the NSW Seed Bank of the Botanic Gardens Trust in Sydney. Thanks to Marti Anderson and Peter Wilson for providing germane advice on multivariate statistics and to Cam Web and colleagues for their ongoing work on Phylomatic. Comments from Robyn Burnham and an anonymous reviewer substantially improved the presentation of our findings.
Rachael V. Gallagher is a PhD candidate at Macquarie University in Sydney, where she investigates the functional traits of climbing plants and climate change.
Michelle R. Leishman is an ecologist at Macquarie University, Sydney, where she works on the comparative ecology of traits and ecological strategies of plants.
Angela Moles is an associate professor at the University of New South Wales in Sydney. Her research focuses on global patterns in plant form and function across many species.
Author contributions: R.V.G. and M.R.L. conceived the ideas; A.T.M. and R.V.G. analysed the data; and R.V.G. collected the data and led the writing.