Contrasting patterns of intraspecific trait variability in native and non-native plant species along an elevational gradient on Tenerife, Canary Islands

Abstract

Background and Aims

Non-native plant species are not restricted to lowlands, but increasingly are invading high elevations. While for both native and non-native species we expected variability of plant functional traits due to the changing environmental conditions along elevational gradients, we additionally assumed that non-native species are characterized by a more acquisitive growth strategy, as traits reflecting such a strategy have been found to correlate with invasion success. Furthermore, the typical lowland introduction of non-native species coming from multiple origins should lead to higher trait variability within populations of non-native species specifically at low elevations, and they might therefore occupy a larger total trait space.

Methods

Along an elevational gradient ranging from 55 to 1925 m a.s.l. on Tenerife, we collected leaves from eight replicate individuals in eight evenly distributed populations of five native and six non-native forb species. In each population, we measured ten eco-morphological and leaf biochemical traits and calculated trait variability within each population and the total trait space occupied by native and non-native species.

Key Results

We found both positive (e.g. leaf dry matter content) and negative (e.g. leaf N) correlations with elevation for native species, but only few responses for non-native species. For non-native species, within-population variability of leaf dry matter content and specific leaf area decreased with elevation, but increased for native species. The total trait space occupied by all non-native species was smaller than and a subset of that of native species.

Conclusions

We found little evidence that intraspecific trait variability is associated with the success of non-native species to spread towards higher elevations. Instead, for non-native species, our results indicate that intermediate trait values that meet the requirements of various conditions are favourable across the changing environmental conditions along elevational gradients. As a consequence, this might prevent non-native species from overcoming abruptly changing environmental conditions, such as when crossing the treeline.

Intraspecific trait variation, phenotypic plasticity, plant functional traits, functional diversity, elevation gradient, altitudinal gradient, alien, exotic, invasion success, forbs, near-infrared spectroscopy

INTRODUCTION

Biological invasions are a focal topic in nature conservation, because the ‘invasive’ species among the non-native ones cause ecological damages and threaten native biodiversity (Vilà et al., 2011; Simberloff et al., 2013). The transportation of non-native species into new regions and their ability to successfully establish themselves and spread under a range of novel environmental conditions might be enabled by a high variability in functional traits of such species (Davidson et al., 2011; Matesanz et al., 2012; Colautti and Barrett, 2013). However, as yet, only a few studies have explicitly compared the intraspecific trait variability of native and non-native species along environmental gradients (e.g. Alexander et al., 2009; Canessa et al., 2018).

While most terrestrial ecosystems are affected by biological invasions, mountains represent a rare exception in that only recently non-native plant species have been documented to expand there, and only a small proportion of these become dominant (Becker et al., 2005; Pauchard et al., 2009; Alexander et al., 2016). Two main mechanisms are likely to prevent plant invasions in harsh environments such as high-elevation sites: propagule limitation and biotic resistance from resident species (Zefferman et al., 2015). However, increasing human influence in mountain areas enhances propagule transportation and disturbance frequency and intensity, all supporting the invasion of non-native species at high elevations. Regarding the invasion capacity of non-native species, invasion success along elevational gradients has been linked to high phenotypic plasticity (e.g. Ansari and Daehler, 2010), and genetic adaption to local environmental conditions (e.g. Monty and Mahy, 2009; Haider et al., 2012). The recent increase of mountain invasions is cause for particular concern due to the ecosystem services these regions provide and the role they play in preserving biodiversity (Nagy and Grabherr, 2009; Pauchard et al., 2009). Concomitantly, these incipient invasions can provide a basis for understanding how non-native plant species expand into these high-elevation regions (Alexander et al., 2011).

Elevational gradients represent a combination of various changing environmental factors, including colder climate, decreasing soil depth and less fertile soil, but also reduced human disturbance due to difficult terrain (Körner, 2003). Non-native species are in most cases introduced in and adapted to human-disturbed habitats at low elevations (Pauchard et al., 2009; Lembrechts et al., 2016) and are likely to have higher intraspecific trait variability in these areas due to receiving germinules from a wide variety of population and geographical backgrounds (Alexander et al., 2011). Related to this, non-native species often exhibit acquisitive growth strategies involving rapid accumulation of biomass (Dawson et al., 2012), larger leaf-area allocation (e.g. higher specific leaf area (SLA); van Kleunen et al., 2010), higher leaf nitrogen and lower leaf carbon content (Henn et al., 2019), and increased nitrogen and phosphorus use efficiency (Drenovsky et al., 2008; Funk, 2008). In contrast, the long-term community assembly of native species might have led to an elevational distribution of species according to their trait suitability for the respective habitat. Thus, native plant species at high elevations typically possess functional traits suited to persist under lower temperatures, a shorter growing season and other elevation-related factors (Körner, 1989, 2003; Read et al., 2014; Rosbakh et al., 2015; Bucher et al., 2016, 2018, 2020). Typical traits for plant species exposed to these conditions include an increased leaf dry matter content (LDMC), which is concurrent with a decrease in SLA, and increased leaf nitrogen and phosphorus contents, reflecting an overall more conservative growth strategy (Körner, 1989, 2003; Dubuis et al., 2013). Due to the long-term evolution of native species populations along elevational gradients, we might expect that the basis of intraspecific trait variability are genetic adaptations to the specific local environmental conditions. Additionally, long-term evolution under less favourable environmental conditions is expected to result in high similarity of plant individuals and thus reduced trait variability within populations (‘environmental filtering’; Ordoñez et al., 2009).

In mountain regions, non-native species predominantly spread along and establish in roadside habitats (Seipel et al., 2012; Haider et al., 2018). Therefore, traits of non-native species in mountain areas are not only a response to the environmental gradients described above, but also a result of the specific conditions of ruderal roadside habitats. In particular, such specific suites of traits are favoured that relate to short generation times or high dispersal rates; while clonality, a long lifespan and shade tolerance are important for colonization of adjacent more natural, non-roadside areas (McDougall et al., 2018).

Intraspecific trait variability, irrespective of its source being identified as phenotypic plasticity, genetic differentiation or both, has been found to support the invasion processes of non-native plants in a variety of ways. For instance, Asteraceae forb species non-native to Europe or North America showed similar elevational clines of seed mass and plant height along elevational gradients in their native and introduced ranges, possibly caused by local adaptation (Alexander et al., 2009). In a glasshouse study with the herbaceous species Polygonum cespitosum, which is invasive in North America, seeds from populations distributed across a wide geographical range were planted in a variety of light and moisture treatments (Matesanz et al., 2012). The authors described that individuals from all populations varied in growth, fitness and leaf traits across the different experimental conditions (i.e. displaying high phenotypic plasticity in all populations). Furthermore, intraspecific variability of SLA, leaf chlorophyll concentration and the root-to-shoot biomass ratio of non-native species in a tropical forest had a significant positive correlation with the range that species were able to occupy along a light gradient (Canessa et al., 2018). Based on such findings, it can be assumed that high intraspecific trait variability resulting from both phenotypic plasticity and genetic differentiation might also support the upward spread of non-native species along elevational gradients.

To compare intraspecific trait variability between native and non-native plant species, we conducted a field study along a gradient spanning almost 2000 m in elevation on Tenerife, Canary Islands. Because mountain roads are the primary pathway for non-native species to spread towards higher elevations (Seipel et al., 2012; Haider et al., 2018), we focused on roadside habitats. In our study area on Tenerife, roadside habitats were characterized by more open dry conditions compared to adjacent more natural areas. By contrast, roadside and non-roadside habitats did not differ in terms of soil temperature and showed the same magnitude of temperature decrease with increasing elevation (own data; unpublished), the latter forming the main environmental gradient in our study. For 11 native and non-native species, eco-morphological and leaf biochemical traits characterizing the position of a plant in the leaf economics spectrum from acquisitive to conservative growth strategies (Wright et al., 2004; Reich, 2014) were measured in 83 populations along the elevational gradient, and intraspecific trait variability was calculated and compared within as well as between populations of native and non-native species. Our aim was to investigate the magnitude of trait variability, irrespective of it arising from plastic responses to the environment or from genetic differentiation, which from a functional point of view is not of primary interest. Rather, we intend to give a general idea of non-native species’ capacity to cope with the steep environmental gradients in mountain areas (compare Milla et al., 2009; Albert et al., 2010).

We tested the following three hypotheses as visualized in Fig. 1:

Fig. 1.

Expectations for changes in growth strategy and within-population trait variability with elevation, and total trait space occupied by (A) native and (B) non-native species. For both native and non-native species we expected acquisitive growth strategies to be replaced by conservative growth strategies with increasing elevation (bold red line; Hypothesis 1). Trait variability within populations (vertical blue arrows) was expected to be constant along the elevational gradient for native species, but to decrease with elevation for non-native species (Hypothesis 2). As a consequence, the total trait space occupied (bold vertical black arrows) was assumed to be larger for non-native species (Hypothesis 3).

• H1: With increasing elevation, acquisitive growth strategies (e.g. high SLA, high leaf nitrogen content) are increasingly replaced by conservative growth strategies (e.g. high LDMC, high leaf carbon to nitrogen ratio). The change of traits with elevation is equally strong within non-native species and native species (identical slope of the red line in Fig. 1). However, on average, non-native species follow a more acquisitive growth strategy compared to native species (higher intercept for native species in Fig. 1).

• H2: Across the elevational gradient, within-population trait variability is lower for native compared to non-native species. For native species, we expect the amount of within-population trait variability to remain constant across elevations because these populations are the results of long-lasting environmental filtering processes that in their intensity should not vary with elevation (vertical blue arrows with constant length in Fig. 1A). By contrast, non-native species are expected to display high within-population trait variability in the more favourable parts of the elevational gradient and comparatively lower within-population trait variability in adverse parts (vertical blue arrows with decreasing length in Fig. 1B).

• H3: Overall, the group of non-native species covers a larger trait space compared to the group of native species (longer vertical black arrow for non-native species in Fig. 1).

MATERIAL AND METHODS

Study area

The field study was carried out on Tenerife, Spain. With c. 2000 km² Tenerife is the largest island in the Canary archipelago and is located off the African coast at 28.28°N, 16.15°W. The centre of the island is dominated by the volcanic cone of Pico del Teide, which rises to 3718 m a.s.l. North-east trade winds divide the island into two climatically different parts: the temperate and moist northern half and the more arid south (Fernández-Palacios, 1992) where our studied elevational gradient was located. The natural vegetation along the southern slope begins with coastal and thermophilous scrub in the areas up to 1000 m a.s.l., transitions to forests of Canary pine (Pinus canariensis) up to 2000 m a.s.l. which are followed by high mountain scrub on the central plateau of Las Cañadas from 2000 to 2500 m a.s.l. The alpine regions close to the summit are only inhabited by a few specialized plant species. While the coastal regions of the island are densely settled, the extent of human influence and the resulting disturbance decreases with increasing elevation and is lowest above 1000 m a.s.l. (Otto et al., 2014).

Sampling design and study species

Sampling was conducted during the growing season from April to May 2018 along three roads which ran from the coast to the central plateau (Fig. 2). The three roads are paved over the entire length of the gradient, are open to traffic throughout the year, and receive similar amounts of traffic (Arévalo et al., 2010).

Fig. 2.

Map of the island of Tenerife, with the three roads used for sampling. The roads start in the vegetation zone of coastal and thermophilous scrub (A), follow a continuous transition (B) to forests of Canary pine (C), and end on the central high-elevation plateau with high mountain scrub. Red parts indicate the actual extent of each road where samples were taken. In each plot, we sampled and measured leaves from eight individuals per species.

As this project was conducted within the framework of the Mountain Invasion Research Network (MIREN; Kueffer et al., 2014), we used a subset of the permanent plots established in 2008 according to the standardized protocol of the global MIREN survey (Arévalo et al., 2010; Seipel et al., 2012; Haider et al., 2018). The plots have a size of 50 × 2 m, with the long side parallel and directly adjacent to the road. To avoid a bias in sampling date, we began in the lowest plots of all roads and moved up the elevational gradient over the course of the fieldwork, following the phenology of the different vegetation belts described above (Fig. 2). The majority of sampled individuals of all species were either flowering or fruiting, supporting our intention to measure leaves only of fully developed plants (Supplementary Data, Table S1).

Eleven forb species that exhibited sufficient abundance and large elevational ranges were sampled (Table 1). All species were present in the majority of the plots in 2008 when the first vegetation survey of our permanent plots was made (Arévalo et al., 2010). With the exception of Calendula arvensis and Volutaria canariensis, which belong to the family Asteraceae, all sampled species belong to different plant families (Table 1). Forbs were selected because they are the only growth form of which native as well as non-native species occur along large parts of the elevational gradient. However, an intrinsic limitation of our study system is that native forb species generally occupy narrower elevational bands, reaching lower maximal elevations compared to non-native species. This restriction needs to be considered when interpreting differences between the groups of native and non-native species.

Table 1.

Overview of the study species, their taxonomic family, floristic status on Tenerife (based on Arévalo et al., 2005; Acebes Ginovés et al., 2010; Haider et al., 2010), invasiveness and the elevational range in which they were sampled. For non-native species, ‘+’ means that the species is considered invasive in the sense of rapidly spreading and causing ecological harm, and ‘–’ means that the species is introduced, but not considered harmful (yet) (based on Acebes Ginovés et al., 2010). For native species, we applied the approach of Dawson et al. (2012) and counted the number of global regions mentioned in the ‘Global compendium of weeds’ (Randall, 2017) which are not part of the species’ natural range

Species	Family	Floristic status	Invasiveness	Elevational range (m)
Bituminaria bituminosa	Fabaceae	Native	1	340–1140
Fagonia cretica	Zygophyllaceae	Native	0	55–505
Forsskaolea angustifolia	Urticaceae	Native	0	55–1005
Rumex vesicarius	Polygonaceae	Native	1	130–405
Volutaria canariensis	Asteraceae	Native	0	55–440
Calendula arvensis	Asteraceae	Non-native	–	405–1020
Erodium cicutarium	Geraniaceae	Non-native	–	340–1925
Eschscholzia californica	Papaveraceae	Non-native	+	875–1635
Hirschfeldia incana	Brassicaceae	Non-native	–	340–1425
Malva parviflora	Malvaceae	Non-native	–	130–1140
Silene vulgaris	Caryophyllaceae	Non-native	–	755–1900

Species were assigned a floristic status (native or non-native) based on the ‘Lista de especies silvestres de Canarias’ (Acebes Ginovés et al., 2010), and where the information from that source was ambiguous, other published studies from Tenerife were consulted in addition (Arévalo et al., 2005; Haider et al., 2010). Within the group of non-native species, only Eschscholzia californica is classified as an invasive species on Tenerife (Acebes Ginovés et al., 2010; Table 1). Among the native species, Bituminaria bituminosa and Rumex vesicarius also occur as non-native species outside their native range (Dawson et al., 2012; Table 1). A coarse analysis (not shown) revealed that the studied non-native species have larger climate niches compared to the studied native species, supporting our assumption of larger trait variability for non-native species (Hypothesis 3).

While C. arvensis, Erodium cicutarium, Es. californica, R. vesicarius and V. canariensis are annual or biennial plants, the other species are perennial (Muer et al., 2016). We aimed to sample each of the 11 species in eight plots covering the whole species’ elevational range. This could be realized for most species and in the case of Forsskaolea angustifolia even be overachieved with ten plots sampled, while the sampling goal could not be met for Malva parviflora or R. vesicarius (five plots), or V. canariense (seven plots). In each plot and for each species, eight individuals (representing one population) were measured and 10–50 leaves per individual (depending on leaf size) were collected for further analyses. Overall, 83 populations were sampled in 29 different plots, resulting in 664 individual leaf samples.

Trait measurements

Leaves collected in the field were placed in zip-lock bags with a moist tissue inside, and stored in a cooling bag. On the same day, the saturated fresh leaves were weighed (Sartorius MC1 AC210, Sartorius AG, Göttingen, Germany) and scanned with a flatbed scanner at a resolution of 300 dpi. Leaf area was calculated using the software WinFOLIA (Regent Instruments, Quebec, Canada). The samples were subsequently dried for 72 h at 80 °C in a drying oven, and weighed again to calculate LDMC (leaf dry mass divided by leaf fresh mass; Table 2) and SLA (fresh leaf area divided by leaf dry mass; Table 2) in accordance with the methods described by Kleyer et al. (2008) and Pérez-Harguindeguy et al. (2013).

Table 2.

Traits measured in this study and their position within the leaf economics spectrum. The first group of traits is associated with a conservative growth strategy, typically found at high elevations, while the second group of traits is associated with an acquisitive growth strategy, which is typical for lowlands. The traits are classified as eco-morphological (morph) and leaf biochemical (chem) traits. R² and root-mean-square error (RMSE) show the quality of the trait predictions via near-infrared spectroscopy. Trait means for the eco-morphological and leaf biochemical traits for each species are given in Supplementary Data Table S3.

Growth strategy	Trait	Abbreviation	Unit	Trait type	Formula	r  2	RMSE
	Leaf dry matter content	LDMC	mg g−1	morph	Leaf dry mass/leaf fresh mass	90.23	0.02
Conservative	Leaf carbon content	Leaf C	%	chem	Percentage of total dry mass	80.50	2.91
	Carbon:nitrogen ratio	Leaf C:N ratio	Non-dimensional	chem	Leaf C/leaf N	60.40	5.44
	Plant height	Height	cm	morph	–	–	–
	Specific leaf area	SLA	cm−2 g−1	morph	Fresh leaf area/leaf dry mass	84.85	25.1
	Leaf nitrogen content	Leaf N	%z	chem	Percentage of total dry mass	88.10	0.37
Acquisitive	Leaf phosphorus content	Leaf P	µmol g−1	chem	P/leaf dry mass	49.02	7.22
	Leaf calcium content	Leaf Ca	µmol g−1	chem	Ca/leaf dry mass	82.36	54.3
	Leaf potassium content	Leaf K	µmol g−1	chem	K/leaf dry mass	62.01	146
	Leaf magnesium content	Leaf Mg	µmol g−1	chem	Mg/leaf dry mass	81.22	61

To determine leaf carbon, nitrogen, phosphorus, calcium, potassium and magnesium contents (leaf C, N, P, Ca, K, Mg; Table 2), the samples were ground in an oscillating mill (MM 400, Retsch, Haan, Germany) until they became homogeneous powder.

A nitric acid digestion was carried out using 200 mg of leaf powder per sample. The liquefied sample was then used to measure leaf P with a photometric assay using ammonium heptamolybdate ((NH₄)₆Mo₇O₂₄) and ascorbic acid (C₆H₈O₆) (Pérez-Harguindeguy et al., 2013), and to determine leaf Ca, K and Mg via atomic absorption spectrometry (ContrAA 300 AAS, Analytik Jena, Jena, Germany). Five milligrams of the leaf powder was used to measure leaf C and leaf N gas-chromatographically with the Dumas method (Vario EL Cube, Elementar Analysensysteme, Langenselbold, Germany), from which we further calculated the carbon to nitrogen ratio (leaf C:N ratio).

Due to the high number of leaf samples, we selected only a subset for the laboratory analyses listed above (‘calibration samples’), and predicted the trait values for the remaining samples via near-infrared reflectance spectroscopy (NIRS) as described by Foley et al. (1998). For the calibration samples, one individual from each plot and species was randomly selected, giving a total of 83 samples. First, leaf powder of all samples was scanned with a stationary NIR spectrometer (MPA, Bruker Optik, Ettlingen, Germany). Each sample was scanned three times and the average spectrum over the three measurements was calculated. Second, separate prediction models were created for each trait (LDMC, SLA and leaf biochemical traits; Table 2) based on the analytical results and reflectance data of the calibration samples (software OPUS version 7.1, Bruker Optik). Finally, using these models and the spectroscopy data, trait values for all samples could be predicted. The quality of the prediction models ranged from r² = 0.49 (leaf P) to r² = 0.90 (LDMC; Table 2). Although using predictions with only moderate coefficient of determination runs the risk of increased β-errors, namely failing to detect an existing relationship, we included them in the further analyses because this conservative approach does not increase the risk of false positive results.

In total, ten eco-morphological and leaf biochemical traits were analysed (Table 2).

Statistical analysis

All statistical analyses were carried out in R version 3.5.0 (R Core Team, 2018). To test our hypothesis on intraspecific trait changes along the elevational gradient (Hypothesis 1), we first conducted a redundancy analysis (rda function in the vegan package; Oksanen et al., 2018) to investigate whether elevation, floristic status (native or non-native) and their interaction significantly affected the trait combinations (only including plant individuals with values for all traits; n = 420). Second, to explore the individual responses of the different traits, we fitted linear mixed-effects models using trait values from all measured individuals as separate data points (n = 664). These models were fitted with each of the ten eco-morphological and leaf biochemical traits as the response variable (Table 2), and elevation, floristic status (native or non-native) and their interaction as fixed effects. Species identity and plot were included in the models as crossed random factors.

To verify whether there was an important influence of phylogeny in the ten eco-morphological and leaf biochemical traits (Table 2) and floristic status (native or non-native), we tested for a phylogenetic signal in these. We constructed a phylogenetic tree for the 11 studied species with the function phylo.maker in the R package V.PhyloMaker (Jin and Qian, 2019), and calculated Blomberg’s K and Pagel’s λ for each trait and floristic status with the function phylosig (package phytools; Revell, 2012) (cf. Supplementary Data, Table S2 for more details on the methods and results). We did not find a phylogenetic signal either for the traits or for floristic status. This means that there is no relationship between the trait values and the phylogeny of our studied species, and that the non-native species were phylogenetically not more similar or distant to each other compared to the sampled native species (cf. Münkemüller et al., 2012; Table S2). Therefore, we did not include a phylogenetic correction in the models described above and in all further analyses.

To test whether within-population trait variability changes with elevation (Hypothesis 2), we calculated Rao’s quadratic entropy (Rao’s Q; Rao, 1982; Botta-Dukát, 2005) for each population (n = 83), using the equation in the FD package (Laliberté and Legendre, 2010):

$${\displaystyle <tex-math>}$$

In contrast to the usual method for calculating Rao’s Q of plant communities, where the species’ trait distances are weighted by its relative abundance in the community, the within-population trait variability corresponds to the sum of all pairwise functional distances between individuals weighted by their relative abundances in the population. Therefore, N here equals the number of individuals in a population (with few exceptions, N = 8), p_i and p_j are the relative abundances of individuals i and j (the abundance of each individual is equal to one), and d_ij is the trait distance between individuals i and j in a population. Thus, eqn (1) measures the mean functional distance between randomly chosen individuals in a population. The calculation of within-population trait variability was done for each trait separately and for all traits combined, using the FD package (R package FD; Laliberté et al., 2014) and scaling trait values to unit variance. We then fitted linear mixed-effects models with the within-population trait variability as the response variable, and elevation, floristic status (native or non-native) and their interaction as fixed effects. As for the previous models, species identity was included as a random factor.

All mixed-effects models were fitted with the function lmer in the R package lmerTest (Kuznetsova et al., 2017), and P-values were calculated from F-statistics of type III sum of squares with the Satterthwaite approximation to estimate the denominator degrees of freedom.

Where interactions in any of the models yielded no significant results, the models were simplified by removing the interaction term and refitted.

To test our third hypothesis on how native and non-native species differ in their total trait space, we first performed a principal component analysis (PCA) including the ten eco-morphological and leaf biochemical traits scaled prior to the analysis. Second, we calculated functional richness (FRic) across all native and across all non-native species individuals with both single- and multitrait approaches (R package FD; Laliberté et al., 2014). For the case of single traits, the functional richness of each trait corresponds to the trait range, calculated as Euclidean distance. For the case of multitrait analysis, functional richness is the minimum convex hull volume of the observations of the ten traits included for each of the two groups of species (native and non-native, respectively) distributed in a ten-dimensional trait space (Villéger et al., 2008). The multitrait space (convex hull volume) is analogous to Hutchinson’s multidimensional niche concept, with each functional trait corresponding to a different dimension occupied by species or individuals according to their trait values (Rosenfeld, 2002).

Because the PCA displayed the gradient from conservative to acquisitive traits, we also used it to test if the location of the centroid of native and non-native species differed along the PCA axes (Hypothesis 1), using the envfit function in the vegan package (Oksanen et al., 2018).

RESULTS

Elevational trait patterns

The redundancy analysis (RDA; total inertia: 8.69; proportion explained by constrained eigenvalues: 0.455; proportion explained by unconstrained eigenvalues: 0.545) showed that elevation had a significant effect on the trait combinations (F = 20.24, P < 0.001). Separate mixed-effects models revealed that across all sampled individuals, leaf N and the leaf cation contents decreased significantly with increasing elevation (Fig. 3), while leaf C and leaf C:N ratio increased significantly with elevation (Table 3; Fig. 3B, C). However, except for leaf C (~3 % overall increase) and leaf Ca (~40 % overall decrease), these responses were mainly driven by trait changes in native species, for which leaf N, leaf K and leaf Mg decreased by about 40 % and leaf C:N ratio increased by about 70 % with increasing elevation. In contrast, for non-native species, these traits remained rather constant along the elevational gradient. For LDMC and plant height we did not find a significant main effect of elevation across native and non-native species. While both traits increased with elevation for native species (~12 % for LDMC and >100 % for plant height; Fig. 3A, D), they (slightly) decreased for non-native species (~3 % and 30 %, respectively). There were no significant changes in SLA and leaf P along the elevational gradient. We did not detect any differences between the species’ mean trait values for native compared to non-native species (Table 3; Supplementary Data Table S3).

Table 3.

Results from the linear mixed-effects models for traits of all individuals sampled as a response to elevation and floristic status (native or non-native). Traits are ordered by their association with either a conservative or acquisitive growth strategy (Table 2). F-values and P-values, taken from type III sum of squares with the Satterthwaite approximation to estimate the denominator degrees of freedom (d.f.), are indicated in bold text when significant (P < 0.05). If the interaction was not significant, the model was refitted without the interaction

	Fixed effects
	Elevation			Floristic status			Elevation × floristic status
Response traits	d.f.	F	P	d.f.	F	P	d.f.	F	P
LDMC	49.61	1.22	0.275	9.79	0.42	0.53	490.78	4.65	0.031
Leaf C	41.67	7.27	0.010	9.28	0.15	0.71
Leaf C:N ratio	51.50	18.18	<0.001	29.57	1.04	0.32	331.38	20.73	<0.001
Height	56.46	3.86	0.054	12.26	1.89	0.19	384.97	16.73	< 0.001
SLA	36.95	2.23	0.144	9.11	0.00	0.96
Leaf N	59.45	9.06	0.004	19.36	0.45	0.51	369.90	19.06	<0.001
Leaf P	52.53	2.59	0.114	9.35	0.00	1.00
Leaf Ca	42.53	4.77	0.035	9.59	0.88	0.37
Leaf K	45.26	10.22	0.003	16.25	3.72	0.07	388.33	11.16	<0.001
Leaf Mg	85.48	13.33	<0.001	10.19	0.89	0.37	212.44	5.91	0.016

Fig. 3.

Significant trait changes along the elevational gradient for 11 native and non-native forb species (n = 664). Traits are sorted by their association with either conservative (A–C) or acquisitive growth strategy (D–H). Regression lines are based on model predictions (Table 3). Green lines and points represent native species, while orange lines and triangles represent non-native species.

The differing responses to elevation of native and non-native species were also reflected in the RDA, revealing a significant interacting effect of elevation and floristic status (F = 12.89, P < 0.001). Removal of the interaction resulted in a higher Akaike’s information criterion value (953 vs. 942 including the interaction), i.e. in a lower explanatory power of the model.

Within-population trait variability

Along the elevational gradient, the change of within-population trait variability (calculated as Rao’s Q) of LDMC, leaf C:N ratio and SLA differed significantly between native and non-native species (Table 4). The within-population trait variability increased for native species by 300 % (LDMC; Fig. 4A), 800 % (leaf C:N ratio; Fig. 4B) and 80 % (SLA; Fig. 4C). For non-native species it decreased (by 95 % for LDMC and by 70 % for SLA; Fig. 4A, C) or remained constant (leaf C:N ratio; Fig. 4B). All other traits measured did not display any changes of within-population variability along the elevational gradient (Table 4). Within-population variability of LDMC was about 70 % higher for native than for non-native species, but we did not find any other significant differences in within-population trait variability between the groups of native and non-native species.

Table 4.

Results from the linear mixed-effects models for within-population trait variability (measured as Rao’s Q) of each trait and all traits combined (convex hull volume), using elevation, floristic status (native or non-native) and their interaction as fixed effects. Traits are ordered by their association with either acquisitive or conservative growth strategies. F-values and P-values, taken from type III sum of squares with the Satterthwaite approximation to estimate the denominator degrees of freedom (d.f.), are indicated in bold text when significant (P < 0.05). When the interaction was not significant, the model was refitted without the interaction

	Fixed effects
	Elevation			Floristic status			Elevation × floristic status
Response traits	d.f.	F	P	d.f.	F	P	d.f.	F	P
LDMC	63.41	3.64	0.061	34.40	4.75	0.036	63.41	23.23	<0.001
Leaf C	78.06	0.00	0.966	11.74	3.87	0.073
Leaf C:N ratio	44.15	14.65	<0.001	23.74	0.41	0.528	44.15	14.05	0.001
Height	74.99	0.04	0.835	11.82	0.13	0.723
SLA	70.83	0.47	0.494	30.99	0.12	0.729	70.83	5.04	0.028
Leaf N	79.84	1.97	0.164	10.02	0.11	0.744
Leaf P	73.76	0.14	0.711	9.33	1.48	0.254
Leaf Ca	78.00	0.01	0.923	13.17	2.64	0.128
Leaf K	63.15	0.02	0.899	15.28	0.15	0.702
Leaf Mg	74.71	0.56	0.456	9.47	0.76	0.404
Convex hull volume	72.00	1.73	0.192	72.00	0.01	0.934

Fig. 4.

Significant elevational change of within-population trait variability (n = 83) as measured by Rao’s Q for LDMC (A), leaf C:N ratio (B) and SLA (C). Regression lines are based on model predictions (Table 4). Green lines and points represent native species, while orange lines and triangles represent non-native species.

Total trait space

In the PCA including the ten eco-morphological and leaf biochemical traits (Table 2), the first axis was mainly associated with leaf C:N ratio (negatively), and leaf K and leaf P (positively), and explained 31.8 % of the observed variance (Fig. 5). The second axis of the PCA explained 22.7 % of the trait variability and was mainly associated with leaf Ca (positively) and plant height (negatively). The leaf economics spectrum (Wright et al., 2004; Reich, 2014) was linked to the first PCA axis, with traits reflecting a conservative growth strategy (high LDMC and leaf C:N ratio) and traits reflecting an acquisitive growth strategy (high SLA, leaf N and leaf P) pointing in opposite directions. Fitting floristic status (native or non-native) to the PCA showed that the position of the centroids of both species groups along the first axis did not differ (P = 0.784, r² = 0.0002), and thus that non-native species were not characterized by a more acquisitive growth strategy compared to native species. Also the linear mixed-effects models for separate traits did not show any differences between the species groups (Table 3).

Fig. 5.

PCA indicating the total trait space of the two groups of native and non-native species (Table 1), based on ten eco-morphological and leaf biochemical traits (Table 2). Per cent values show the variance explained by PC1 and PC2. The ellipses are 95 % confidence ellipses. Larger symbols correspond to the ellipse centroids.

The total trait space (calculated as functional richness) of native species was between ~20 % (height, leaf Ca) and 200 % (leaf P) larger than that of non-native species for seven out of ten traits, and 300 % larger in the multitrait analysis (Supplementary Data, Table S4). In contrast, the total trait space of LDMC, leaf N and leaf K was up to 12 % larger for non-native compared to native species. In line with the smaller convex hull volume (multitrait functional richness) of non-native species, the trait space visualized in the PCA was smaller for non-native species and a subset of the native species’ trait space. Specifically along the second axis (associated with leaf Ca), native species were more dispersed.

While the trait space of most species was strongly overlapping, R. vesicarius had the largest expansion along the first PCA axis and F. angustifolia the largest along the second axis (Supplementary Data, Fig. S1).

DISCUSSION

To understand the fairly recent phenomenon of plant invasions in mountain areas (Pauchard et al., 2009; Alexander et al., 2016) and how these introduced species manage to reach high elevations, we compared intraspecific trait variability of 11 native and non-native plant species by measuring functional traits from several populations per species along an elevational gradient on Tenerife. We found that non-native and native species differed in their trait response to elevation. Native species responded to elevation in the majority (8 out of 10) of the traits measured. In contrast, non-native species only barely showed trait changes in response to the varying local conditions along the elevational gradient. Surprisingly, for native species, we did not find any indications for environmental filtering with increasing elevation. Only for non-native species did within-population trait variability of LDMC and SLA decrease along the elevational gradient. For our studied species, non-native species occupied a subset of the total trait space occupied by native species.

Elevational trait patterns

We expected acquisitive growth strategies to be increasingly replaced by conservative growth strategies with increasing elevation (Hypothesis 1). Accordingly, leaf C and leaf Ca showed positive and negative correlations, respectively, with elevation for both native and non-native species. In addition, this hypothesis was confirmed for native plant species, which exhibited further trait shifts with elevation, such as increasing LDMC and leaf C:N ratio, but decreasing leaf mineral nutrients (N, K, Mg), as also described by several studies in different regions and for different life-forms (Cordell et al., 1998; Wright et al., 2004; Dubuis et al., 2013; Liu et al., 2016; Pfennigwerth et al., 2017). The results thus support the general idea that species occurring along environmental gradients exhibit at the less benign end of the gradient shifts in traits towards the ‘slow-return’ of the leaf economics spectrum, with low carbon fixation rates and nutrient contents, associated with longer leaf lifespans and thicker, denser leaves (Reich et al., 2003). These ‘slow’ or conservative traits are particularly advantageous under low-resource conditions, such as high-elevation areas, because resource conservation confers these plants with the ability to better cope with stress, which can increase their survival chances. Interestingly, although the native species studied here did not reach such high elevations as the non-native species, their trait shift towards a more conservative growth strategy was more pronounced compared to the non-native species, which did not show a clear response to elevation. Therefore, our results only partly support our hypothesis, namely that both native and non-native species display trait changes along the elevational gradient, but not that the magnitude of trait changes is similar. While non-native species have been observed to respond to changes in environmental conditions in a meta-analysis (Davidson et al., 2011), several other studies show the opposite. Individual studies (e.g. Brock et al., 2005; Murphy et al., 2016) and a meta-analysis by Palacio-López and Gianoli (2011) failed to find non-native species to be more plastic than native species. Brock et al. (2005) and Murphy et al. (2016) both came to the conclusion that trait values related to fitness, such as seed mass, or resource capture, such as SLA, might be more important for invasion success than overall trait plasticity.

Traffic along the road can result in constant seed dispersal and thus mixture between different plant populations along the gradient, which further inhibits the expression of trends in traits at the local level by a homogenization effect (Haider et al., 2012). These factors could further explain the weak to absent trait response to the gradient for non-native species. This is in contrast to the native species, which are assumed to disperse from locally adapted populations inside the natural habitats along the elevational gradient and thus do exhibit shifts in traits as described above. Roads acting as a corridor for non-native species’ dispersal can aid their movement to upper elevations by bringing in individuals from different backgrounds (and thus contributing to within-population variability), but at the same time preventing an expected shift in trait values with elevation.

For the studied eco-morphological and leaf biochemical traits, we did not observe non-native species to follow on average a more acquisitive growth strategy compared to native species, which contrasts with several studies that found them to exhibit higher SLA (van Kleunen et al., 2010), leaf N (Drenovsky et al., 2008) and P use efficiency (Funk, 2008). In contrast, Funk and Vitousek (2007) found invasive species to out-perform native ones under low nutrient conditions by resource conservation traits. In accordance with our results, when comparing non-native and native forb species in different fertilization treatments, Scharfy et al. (2011) observed almost no differences between the two groups. Scharfy et al. (2011) concluded that the difference between the fertilization treatments they used might not have been significant enough to give non-native species a clear advantage. Similarly, the roadside habitats in our study might not have been sufficiently nutrient-rich to induce non-native species trait values that are clearly differentiated from those of the native species.

Within-population trait variability

We hypothesized that native species display overall lower, but constant, within-population trait variability across the elevational gradient when compared to non-native species, while non-native species should display high within-population trait variability in suitable parts and comparatively lower variability in adverse parts of the elevational gradient.

While in our study within-population trait variability of native and non-native species did not differ per se, we found significant differences in the direction of change of within-population trait variability of LDMC, leaf C:N ratio and SLA along the elevational gradient between the two species groups. The decrease of within-population trait variability for LDMC and SLA for non-native species concurs with our expectation that these species display decreasing within-population trait variability as they expand into higher elevations where environmental conditions become less suitable. Similarly, Lang et al. (2019) also found a significant response of within-population trait variability for SLA along a gradient from dry to moister conditions in Mongolian rangelands, but with trait variability peaking at intermediate moisture conditions.

In our study, higher levels of within-population trait variability of non-native species at lower elevations might be explained by repeated introductions from multiple source populations to the coastal regions (Arteaga et al., 2009; Haider et al., 2010; Alexander et al., 2011). The changing environmental conditions with increasing elevation (lower temperatures and lower nutrient availability) gradually restrict the range of suitable trait characteristics, resulting in lower functional diversity in non-native populations at higher elevations. This trait filter could be another factor in the comparatively slow expansion of non-native species into mountain ecosystems, next to lower propagule pressure, lower disturbance and less benign environmental conditions when compared to the lowlands (Pauchard et al., 2009). However, with our study design we cannot conclude whether filtering of locally adapted genotypes is the actual mechanism underlying the within-population trait variability observed, which could only be inferred from experimental or genetic analyses.

For native species, we found for most traits no change of within-population trait variability along the elevational gradient. Similarly, Lang et al. (2019) detected changes of within-population trait variability only for a subset of their traits measured along a precipitation gradient, indicating that not all traits respond to the same environmental gradient, which in our case was considered to be a temperature gradient. In contrast to our expectation of climate filtering towards higher elevations, within-population trait variability of LDMC, leaf C:N ratio and SLA for native species increased with elevation. A possible explanation might be that biotic interactions in the form of hierarchical fitness differences cause this surprising result: Although soil moisture is low across the whole elevational gradient (<10 vol.%), it might be a more limiting resource at lower elevations where the evaporational demand is higher and soil nutrient supply is greater. It has been shown that competition for a single resource might not lead to trait overdispersion as a result of limiting similarity, but to trait clustering and thus lower trait variability (Mayfield and Levine, 2010; Gallien, 2017). Additionally, as our sampled native species populations did not reach higher than 1140 m a.s.l., while non-native species were found as high as 1925 m a.s.l., it is likely that the effect of the climate filter towards higher elevations could only be captured in our study for non-native species.

Total trait space

Our third hypothesis was that the group of non-native species is more variable than the group of native species. However, multidimensional analysis including ten eco-morphological and leaf biochemical traits across all individuals revealed that non-native species occupied a subset of the total trait space occupied by native species. Mainly two species were responsible for the larger native species’ trait space: R. vesicarius along the first axis and F. angustifolia along the second axis of a PCA (Supplementary Data, Fig. S1). In the case of R. vesicarius, an agricultural analysis found high concentrations of numerous mineral nutrients including K, Ca and Mg for this species (Kambhar, 2014), in addition to large intraspecific trait variability being reported for several congeneric species (e.g. Rumex crispus; Hume and Cavers, 1982) which might indicate the high trait variability potential inherent to the genus.

Furthermore, the trait range across all non-native species was smaller than for native species for most of the traits considered. This was surprising, because the non-native flora specifically at low elevations results from multiple introductions from different source regions for each of the species (Alexander et al., 2011), and other studies found higher intraspecific trait variability also within single non-native species (e.g. Alexander et al., 2009; Davidson et al., 2011; Canessa et al., 2018). However, our results are similar to those of Scharfy et al. (2011), who also found no differences in intraspecific trait variability between non-native and native forb species. Also, Murphy et al. (2016) found no significant correlation between intraspecific trait variability of leaf and growth traits and the global invasion success of different Rosa species in a glasshouse experiment. The lack of consensus indicates that intraspecific trait variability might be strongly dependent on the studied species, measured traits and other local factors (e.g. Kichenin et al., 2013; Bucher et al., 2016, 2019). For our case, some of the studied species were highly variable regarding specific traits (R. vesicarius and F. angustifolia extended the native species’ trait space especially regarding leaf biochemical traits, such as leaf Ca and leaf K), while SLA surprisingly did not show any response to elevation.

The observed large total trait space of native species recorded along the elevational gradient might be interpreted as the ability of native species to exhibit high fitness homeostasis under different resource availabilities, as observed in a meta-analysis (Davidson et al., 2011). The combination of strong responses of native species to elevation and their large trait space indicates that these species, being present there for longer time, are locally adapted to the specific conditions of each site along the elevational gradient, while still maintaining high trait variability across the whole gradient. The fact that the sampled native species responded strongly to the elevational gradient despite being restricted to narrow elevational bands reinforces this conclusion.

The results presented here give new insights into the role traits and intraspecific trait variability play in allowing range expansions of non-native species towards high elevations. In contrast to other studies (Alexander et al., 2009; Davidson et al., 2011), we found only few differences in trait variability between native and non-native species. Thus, there is little evidence that intraspecific trait variability is associated with the success of non-native species to spread towards higher elevations. Rather, it seems that being able to express a certain set of traits is more useful for the successful upwards spread of non-native species than having larger intraspecific trait variability than native species, similar to the conclusions Murphy et al. (2016) drew in their study. This also concurs with a meta-analysis carried out by Dawson et al. (2012), who found that a high plasticity of resource-capture traits was not correlated to the global distribution range of species, and that this success was rather correlated with the ability to rapidly increase biomass in suitable conditions. Similarly, Helm et al. (2019) found no evidence that intraspecific trait variability supported the recolonization success of typical species of Mediterranean steppe communities.

Along elevational gradients, environmental filters gradually restrict the functional suitability of non-native species. This might make it increasingly difficult for non-native species to propagate through the different vegetation types along the gradient and bridge the gap between abruptly changing habitat types, for example crossing the treeline or percolating away from the road into natural plant communities.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Fig. S1: PCA showing the total trait space occupied by each of the 11 native and non-native species studied. Table S1: Relationship between the phenological status of sampled individuals and sampling elevation. Table S2: Testing for phylogenetic signal in eco-morphological and leaf biochemical traits and in floristic status. Table S3: Mean and standard deviation for each trait and species. Table S4: Functional richness of eco-morphological and leaf biochemical traits for the two groups of native and non-native species.

ACKNOWLEDGEMENTS

We are grateful to José Ramón Arévalo and Miguel Mederos (Universidad de La Laguna) for their support during fieldwork on Tenerife. We also thank Gunnar Seidler and Erik Welk for their help with GIS, specifically for the calculation of climate niches. We thank two anonymous reviewers for their helpful comments.

FUNDING

The project was conducted in the framework of the Flexible Pool project (W47014118) of S.H. funded by the German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig.

Author notes

P.K. and A.R.B. contributed equally to this study and should be considered joint first authors.