12490
Summary
1. Analysing functional traits along environmental gradients can improve our understanding
of the mechanisms structuring plant communities. Within forests, vertical gradients in light
intensity, temperature and humidity are often pronounced. Vascular epiphytes are particularly
suitable for studying the inuence of these vertical gradients on functional traits because they
lack contact with the soil and thus individual plants are entirely exposed to dierent environ-
mental conditions, from the dark and humid understorey to the sunny and dry outer canopy.
2. In this study, we analysed multiple aspects of the trait-based ecology of vascular epiphytes:
shifts in trait values with height above ground (as a proxy for vertical environmental gradients) at
community and species level, the importance of intra- vs. interspecic trait variability, and trait
dierences among taxonomic groups. We assessed ten leaf traits for 1151 individuals belonging to
83 epiphyte species of all major taxonomic groups co-occurring in a Panamanian lowland forest.
3. Community mean trait values of many leaf traits were strongly correlated with height and
particularly specic leaf area and chlorophyll concentration showed nonlinear, negative trends.
4. Intraspecic trait variability was pronounced and accounted for one-third of total observed
trait variance. Intraspecic trait adjustments along the vertical gradient were common and sev-
enty per cent of all species showed signicant traitheight relationships. In addition, intraspe-
cic trait variability was positively correlated with the vertical range occupied by species.
5. We observed signicant trait dierences between major taxonomic groups (orchids, ferns,
aroids, bromeliads). In ferns, for instance, leaf dry matter content was almost twofold higher than
in the other taxonomic groups. This indicates that some leaf traits are taxonomically conserved.
6. Our study demonstrates that vertical environmental gradients strongly inuence functional
traits of vascular epiphytes. In order to understand community composition along such gradi-
ents, it is central to study several aspects of trait-based ecology, including both community and
intraspecic trends of multiple traits.
Key-words: carbon isotope ratio d13C, LDMC, leaf chlorophyll concentration, leaf thickness,
leaf nitrogen concentration, leaf water content, nitrogen isotope ratio d15N, specic leaf area
(SLA), variance partitioning, vertical zonation
Cornwell & Ackerly 2009). Interestingly, studies based on replacement of species with unsuitable traits. However,
global trait data sets show that large-scale changes in cli- recent studies have also highlighted that intraspecic trait
matic conditions only explained a small proportion of variability can be quite substantial and that individuals
observed variation in leaf traits, while trait variation within species can adjust their traits in response to the
among coexisting species within study plots was relatively environment (Bolnick et al. 2011; De Bello et al. 2011;
high (Wright et al. 2004, 2005). In fact, the environment Kichenin et al. 2013). Additionally, high intraspecic trait
at small scales can be very heterogeneous, promoting the variability might be associated with a large ecological
occurrence of species with dierent traits and ecological breadth or ecological generalism, possibly increasing the
strategies. Moreover, particularly in forests, environmen- potential vertical range of epiphytes within forests (Van
tal factors such as light intensity, temperature and humid- Valen 1965; Sides et al. 2014).
ity normally show marked vertical gradients. Such vertical Vascular epiphytes are a taxonomically diverse group.
gradients, in turn, have the potential to explain a substan- Orchids account for 68% of all epiphyte species, but
tial part of trait variations at plot scale, and it has been ferns and lycophytes, bromeliads and aroids are also
demonstrated that several leaf traits of trees change sig- prominent taxa (Zotz 2013). It is generally assumed that
nicantly along vertical light gradients (Rozendaal, Hur- traits are taxonomically conserved and, consequently,
tado & Poorter 2006; Markesteijn, Poorter & Bongers trait dierences between taxonomic groups can be
2007). expected. Moreover, epiphyte taxa independently evolved
Vascular epiphytes, plants growing non-parasitically on a variety of dierent morphological and physiological
other plants without contact to the soil (Zotz 2013), are characteristics (e.g. velamen radicum, phytotelmata, spe-
particularly suitable for studying the inuence of vertical cialized trichomes) to cope with nutrient and water limi-
environmental gradients on functional traits, because tation (Benzing 1990). Such dierences between-taxa
individuals are entirely exposed to dierent environmen- might aect the response of leaf traits to environmental
tal conditions from the dark and humid understorey to conditions.
the sunny and dry outer canopy. As the leaf weight ratio To analyse the multiple aspects of trait-based ecology
(leaf mass/total plant mass) is generally high in epiphytes along vertical environmental gradients, we studied ten leaf
(Zotz & Assho 2010), leaf traits should be pivotal to traits for 1151 individuals of 83 epiphyte species of all
their performance. The frequently pronounced vertical major taxonomic groups co-occurring in a Panamanian
stratication of epiphyte species has long been recognized lowland forest. We tested the following hypotheses: (H1)
(Schimper 1888; Kr omer, Kessler & Gradstein 2007), but trait means and trait syndromes change with height at the
few studies have attempted to relate their vertical distri- community level; (H2) variations in traitheight relation-
bution to functional leaf traits. Most of these studies ships among species inuence community trait structure;
assessed dierences between sun and shade plants (e.g. (H3) vertical ranges of epiphyte species correlate with their
Mantovani 1999) or used pre-dened zones within forests intraspecic trait variability; and (H4) trait means, trait
or trees (e.g. Johansson zones; Johansson 1974) as surro- syndromes and traitheight relationships dier among tax-
gates for dierent environmental conditions (Andrade & onomic groups.
Nobel 1997; Hietz & Briones 1998; Stuntz & Zotz 2001).
Zotz (2007) pointed out that height above ground might
be more suitable to approximate the environmental gradi- Materials and methods
ents within forests than pre-dened zones. To our knowl-
edge, only a single study related height above ground to STUDY SITE
leaf traits of vascular epiphytes (Cavaleri et al. 2010). This study was conducted at the San Lorenzo Canopy Crane
However, as this study focused on leaf mass per area Site at the Atlantic coast of Panama (9170 N, 79580 W,
(LMA) and did not include epiphytes from important 130 m a.s.l.; Wright et al. 2003). Mean annual precipitation in
this old-growth lowland tropical rain forest is around
taxonomic groups such as orchids or bromeliads, many
3100 mm, with a pronounced dry season from January to
aspects of the vertical leaf trait distribution of epiphytes March. Canopy height is variable and emergent trees reach
are still largely unexplored. maximum heights of c. 45 m. The use of a gondola attached to
Along vertical environmental gradients, shifts in com- a construction crane allowed access to all strata of the forest
munity mean trait values of functionally important leaf within an area of c. 09 ha. A comprehensive census of the vas-
cular epiphyte ora at the study site was conducted in 2010
traits can be expected. An increase in specic leaf area
2012 and yielded >22 000 individuals of >100 species (G. Men-
(SLA = LMA1), for instance, increases the light-capture dieta-Leiva & G. Zotz, unpublished data; see Zotz & Schultz
eciency, which is advantageous under low-light condi- 2008 for methodology).
tions in the understorey (Wright et al. 2004). In contrast, Among vertical environmental gradients, the light gradient is
an increase in leaf thickness can prevent overheating and considered as most inuential on leaf traits (e.g. Poorter 1999;
Markesteijn, Poorter & Bongers 2007). Changes in light intensity
minimize transpiration losses, which is favourable under
with height above ground were measured in situ with light inten-
drier and sunnier conditions in the canopy (Cornelissen sity loggers (HOBO UA-00264; Onset Computer Corporation,
et al. 2003; Rozendaal, Hurtado & Poorter 2006). Such Cape Cod, MA, USA; for more details, see Fig. S1 in Supporting
shifts in community trait means might be caused by Information).
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
Functional traits of vascular epiphytes 3
Analyses were done in R 3.0.1 (R Development Core Team 2013). To assess the general importance of intraspecic trait variability,
Analyses for each hypothesis are described separately in the fol- we rst carried out variance component analyses (R package
lowing. varcomp), which partition observed trait variability into
within-species (intraspecic) and between-species (interspecic)
components (Messier, McGill & Lechowicz 2010). Subsequently,
H1 Trait means and trait syndromes change with height we calculated two measures of trait variability for each species:
the coecient of variation (CV) and the trait range (TR: abso-
at the community level
lute dierence between maximum and minimum trait value
We used linear models (LMs) to analyse the relationship divided by the maximum, given in %). The relationship between
between leaf traits and height. To test for nonlinearity, simple these measures of intraspecic trait variability and species
LMs (trait ~ height) and LMs including a quadratic term (trait vertical ranges was analysed with LMs (vertical range ~ trait
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
4 G. Petter et al.
variability), whereby the vertical range for each species was esti- lated with height (Fig. 1). While SLA, Chlmass and Nmass
mated based on its maximum and minimum height observed in showed decreasing, nonlinear trends with height, leaf
the census.
thickness, LWCarea and d13C showed positive linear
trends with height (Fig. 1). LDMC and d15N showed
H4 Trait means, trait syndromes and traitheight rela- slightly negative trends, but rather weak correlations.
tionships differ among taxonomic groups Observed community trends were largely consistent with
those considering the entire censused community, indicat-
Dierences in trait means among the major taxonomic groups
(aroids, bromeliads, orchids, ferns; see Table 1 for number of spe- ing no substantial sampling bias (compare Figs 1 and S2,
cies and individuals), based on trait means of associated species, as well as Tables S3 and S4).
were compared using max-t tests for multiple comparisons that Many traits covaried signicantly (Table S5; P < 005):
account for unbalanced group sizes, non-normality and heterosce- for instance, leaf thickness and LWCarea (r = 084), Chlmass
dasticity (R packages multcomp and sandwich; see Herberich,
and Nmass (r = 067), as well as SLA and Chlmass
Sikorski & Hothorn 2010).
Dierences in trait syndromes among the taxonomic groups (r = 064). The rst two PCA axes explained 45% and
were tested using a permutational multivariate analysis of variance 25%, respectively, of variation in leaf traits. Height
(PERMANOVA, adonis from vegan R package; Anderson 2001). explained 16% of variation along the rst axis and 7%
Additionally, we used the PCA results to visualize dierences along the second axis (P < 0001).
among taxonomic groups.
Dierences in traitheight relationships among the taxonomic
groups were analysed using generalized linear mixed models (see H2 VARIATIONS IN TRAITHEIGHT RELATIONSHIPS
Fig. S3 and Table S2 for details).
AMONG SPECIES INFLUENCE COMMUNITY TRAIT
STRUCTURE
Results The MAM for SLA, LDMC, leaf thickness and Chlmass
included the interaction between species and height, indi-
H1 TRAIT MEANS AND TRAIT SYNDROMES CHANGE
cating that the community trait structure was best
WITH HEIGHT AT THE COMMUNITY LEVEL
explained when considering that species dier in both
All leaf traits were signicantly correlated with height their trait means and their trait responses to height
(P < 005, Fig. 1, Table S3). The strongest correlations (Table S6). In contrast, for Chlarea and LWCarea only
between community trait means (for 1-m height intervals) between-species dierences in trait means were signicant
and height were observed for SLA (R2 = 089), Chlmass (Table S6).
(R2 = 076), leaf thickness (R2 = 072), d13C (R2 = 066) Seventy per cent of all species had at least one trait that
and LWCarea (R2 = 064). Traitheight correlations were was signicantly correlated with height (P < 005). Signi-
generally much weaker when, instead of community cant intraspecic traitheight relationships were most com-
means, traits of all sampled individuals were used as mon for SLA, for which 45% of all species revealed a
response variable: in this case only d13C (R2 = 035), SLA signicant relationship with height, followed by LDMC
(R2 = 030) and Chlmass (R2 = 016) were moderately corre- with 33% (Table S7; see Figs S4S9 for intraspecic
Table 1. Mean leaf trait values SD of the major taxonomic groups of vascular epiphytes (aroids, bromeliads, orchids, ferns) in a Pana-
manian lowland forest. Species from all other taxa are summarized in Others. CAM species were excluded from d13C analyses. Dier-
ences between taxonomic groups were analysed using max-t tests for multiple comparisons of means, and signicant dierences in trait
means (P < 005) are indicated by dierent letters. Proportions of sampled individuals and species are given in parentheses.
Individuals 149 (129%) 62 (54%) 435 (378%) 379 (329%) 126 (109%)
Species 13 (157%) 5 (6%) 32 (386%) 24 (289%) 9 (108%)
Height (m) 122 75A 147 74AB 210 64B 111 72A 154 30A
SLA (mm2 mg1) 221 109AB 177 75AB 140 57A 242 148B 276 112B
LDMC (g g1) 017 005A 018 004A 020 010A 034 011B 008 005C
Thickness (mm) 038 013A 059 069AB 070 052B 026 012A 096 059B
LWCarea (g H2O m2) 297 102A 378 278ABC 488 344B 168 128C 632 325B
Chlmass (mg g1) 101 29A 54 32AB 63 26B 75 27AB 83 39AB
Chlarea (lg cm2) 553 164A 313 91B 476 153A 414 191AB 323 96B
d13C (&)* 297 27A 299 07A 297 21A 312 16A 304 20A
d15N (&)* 18 19AB 28 10AB 23 10AB 17 11A 38 11B
Nmass (mg g1)* 145 64A 75 21B 116 55AB 120 36A 118 40AB
Narea (g m2)* 107 029A 057 036A 092 033A 096 038A 069 029A
*For the NC traits, not all species were sampled. Species numbers assessed for NC traits were as follows: Aroids: n = 10, Bromeliads:
n = 5, Orchids: n = 19-24, Ferns: n = 17, Others: n = 5, with on average 34 replicates per species (Table S1).
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
Functional traits of vascular epiphytes 5
Thickness (mm)
60 = 030*** = 001*** = 011*** 1500
LDMC (g g1)
20
06
40 15 1000
04
10
20 500
02 05
0 00 00 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30
(e) 30 (f) 200 (g) 15 (h) 6
R M2 = 076*** R M2 = 026** R M2 = 066 *** R M2 = 023**
25 4
Chlarea (g cm2)
150 20 2
20
13C ()
15N ()
0
15 100 25
2
10
50 30 4
5 6
0 0 35 8
0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30 0 5 10 15 20 25 30
(i) 30 (j) 25
Height (m) Height (m)
R M2 = 039*** R M2 = 013*
25 20
R C2 = 009*** R C2 = 004**
Nmass (mg g1)
Narea (g m2)
Fig. 1. Traitheight relationships of vascular epiphytes for ten leaf traits: (a) SLA: specic leaf area, (b) LDMC: leaf dry matter content,
(c) Thickness: leaf thickness, (d) LWCarea: leaf water content per leaf area, (e) Chlmass: mass-based chlorophyll concentration, (f) Chlarea:
area-based chlorophyll concentration, (g) d13C: carbon isotope ratio, (h) d15N: nitrogen isotope ratio, (i) Nmass: mass-based nitrogen con-
centration, (j) Narea: area-based nitrogen concentration. Simple LMs (trait ~ height) and LMs including a quadratic term (trait ~ height
+ height2) were tted and compared by AIC. Nonlinear models were preferred when DAIC10 (see Table S3 for summary statistics).
R2M: amount of variance in community means explained by height. R2C: amount of variance in individuals trait values explained by
height. Asterisks indicate signicance levels of traitheight relationships (***P < 0001, **P < 001, *P < 005). Shaded areas indicate
95% CI.
traitheight relationships of all species). The directions of (R2 = 024, P < 0001) than for mean CV (R2 = 010,
intraspecic traitheight relationships were largely consis- P = 0009).
tent within traits. For instance, for SLA and Chlmass,
slopes were invariably negative (Table S7). However, for
H4 TRAIT MEANS, TRAIT SYNDROMES AND TRAIT
LDMC and thickness, there were a few species showing
HEIGHT RELATIONSHIPS DIFFER AMONG TAXONOMIC
opposing trends (Table S7).
GROUPS
signicant. In contrast, the highest nitrogen and chloro- with height, but strength and direction of correlations
phyll concentrations were consistently found in aroids. varied considerably. The strongest correlations among the
Taxonomic groups also diered signicantly in height extensively sampled traits were found for SLA and
distributions. The mean height of orchid species Chlmass, whose negative trends from the forest oor to the
(210 64 m) was signicantly higher (P < 005, max-t upper canopy are consistent with dierences between sun
test) than that of aroids (122 75 m) and ferns and shade leaves of tropical trees (Rozendaal, Hurtado &
(111 72 m), but did not dier signicantly from that of Poorter 2006; Markesteijn, Poorter & Bongers 2007) and
bromeliads (147 74 m; Table 1). trends along tree height gradients (Rijkers, Pons & Bon-
The PERMANOVA indicated signicant dierences in trait gers 2000). When considering that SLA and Chlmass covar-
syndromes among all taxonomic groups (P < 0001). The ied considerably and that Chlarea did not show a strong
dispersion of species in PCA trait space showed that several vertical trend, it seems likely that changes in Chlmass were
species of dierent taxa shared similar trait syndromes, but mainly driven by changes in SLA (Chlmass =
also that there were unique tendencies within taxonomic Chlarea SLA). In soil-rooted plants, vertical gradients in
groups (compare, e.g. orchids and ferns; Fig. 2). SLA are commonly related to vertical light gradients
The xed-eect structure of the MAMs did not include (Poorter 1999; McMurtrie & Dewar 2011), but hydraulic
the interaction between height and taxonomic group for constraints have also been discussed (Rijkers, Pons &
any leaf trait, indicating that slopes of traitheight relation- Bongers 2000; Koch et al. 2004). A comparative study by
ships did not dier signicantly among taxonomic groups Cavaleri et al. (2010) found that epiphytes were the only
(Table S2). No signicant dierences in slopes or intercepts plant group for which light was most important in
were observed for SLA and all NC traits (Fig. S3). explaining vertical SLA proles, which seems logical as
epiphytes lack a hydraulic connection to the ground.
Because SLA relates the light-capturing leaf area to
Discussion
investment in dry mass, an increase in SLA increases the
potential carbon gain per biomass investment. However,
H1 TRAIT MEANS AND TRAIT SYNDROMES CHANGE
increased light-capture eciency via high SLA tends to
WITH HEIGHT AT THE COMMUNITY LEVEL
be associated with higher respiration rates and shorter
Our results support the hypothesis that community trait leaf life spans. Several such correlations between leaf
means of vascular epiphytes are signicantly correlated traits capturing fundamental aspects of leaf economics
Aroids Bromeliads
06
LDMC Chlarea
04
Orchids Ferns
02
PC2 (25%)
00
LWCarea
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
Functional traits of vascular epiphytes 7
have been observed (worldwide leaf economics spectrum; ance in tissue d13C can be attributed to dierences in
Wright et al. 2004). Theoretical models have demon- water-use eciency. These results agree with observations
strated that, when considering these between-trait correla- for leaves of tropical trees (Medina & Minchin 1980). In
tions, the carbon gain over the leaf life span is maximized contrast, the dierence in d13C signals of epiphytes
when SLA increases nonlinearly with decreasing light between the upper and lower tree zones of a lowland rain
(Sims, Gebauer & Pearcy 1994; McMurtrie & Dewar forest in Costa Rica was smaller (< 2 &; Wania, Hietz &
2011). The nonlinearly decreasing community means of Wanek 2002), possibly due to a less pronounced gradient
SLA with height (Fig. 1a) agree with these expectations of water stress than in our system (precipitation at that site
and corroborate the notion that light is the main driver is >6000 mm year1). Alternatively, the discrepancy may
of vertical SLA proles in epiphytes. indicate that height above ground is a better predictor for
The observed increase in leaf thickness with height is water stress than the pre-dened Johansson zones used by
consistent with within-individual, intra- and interspecic Wania, Hietz & Wanek (2002), which subdivide host trees
vertical trends found in trees (Rozendaal, Hurtado & according to their principal structure without considering
Poorter 2006; Markesteijn, Poorter & Bongers 2007). absolute height.
Apart from maximization of carbon gain, avoidance of In line with Wania, Hietz & Wanek (2002), we observed
damages and water loss minimization are also require- a negative trend in tissue d15N with height (Fig. 1h). The
ments of optimal leaf functioning: an increase in leaf thick- d15N signatures of plants are mostly aected by their
ness is regarded as adjustment to prevent overheating and assimilatory pathway, but also by form (NO
4 , NH3 , N2)
15
to balance carbon gain and transpiration water loss under and d N signature of the nitrogen source (Evans 2001).
drier and sunnier conditions (Cornelissen et al. 2003; Ro- Epiphytes use a blend of dierent autochthonous (e.g. can-
zendaal, Hurtado & Poorter 2006). Without anatomical opy soil, leachates) and allochthonous nitrogen sources
adjustments that change the leaf tissue density, a decrease (e.g. wet and dry deposition), which can vary substantially
in SLA would induce an increase in leaf thickness, which in d15N signatures (Wania, Hietz & Wanek 2002). The
partially explains their covariance (r = 048). Neverthe- observed negative trend with height indicates an increasing
less, the linear increase in leaf thickness (Fig. 1c) in con- contribution of atmospheric N to epiphyte N in the upper
trast to the nonlinear decrease in SLA (Fig. 1a) suggests canopy. However, as we did not measure source d15N sig-
that the trend in leaf thickness is not only related to SLA, natures, caution is needed when interpreting tissue d15N
but also to independent morphological adjustments which trends.
are probably more inuenced by the vertical gradient in In summary, we found only moderate to weak correla-
potential evapotranspiration than by the vertical light gra- tions between leaf traits/leaf trait syndromes and height
dient. when considering all individuals, but often strong correla-
In general, LDMC also tends to scale with SLA and is tions between community means and height. This also
sometimes regarded as an alternative predictor of plant reects that height is a suitable proxy of general vertical
strategies (Wilson, Thompson & Hodgson 1999). Interest- trends in environmental conditions, although it does not
ingly, the observed covariance between LDMC and SLA capture all relevant factors and small-scale environmental
was rather low at community level (r = 016) and the variability (Fig. S1).
LDMC-height correlation was rather weak (Fig. 1b). This
suggests that plant functioning captured by SLA is more
H2 VARIATIONS IN TRAITHEIGHT RELATIONSHIPS
relevant along vertical gradients within forests.
AMONG SPECIES INFLUENCE COMMUNITY TRAIT
It is well-established that the proportion of epiphytes
STRUCTURE
with CAM increases with height (e.g. Zotz 2004). The
positive trend in d13C of C3 plants documented here For four out of six traits, the community trait structure
(Fig. 1g) has arguably the same ecological background: could be best explained when including dierences in intra-
more demanding water relations result in increasing sto- specic trait response to height, which supports our
matal limitations (Farquhar, Ehleringer & Hubick 1989). hypothesis for most traits. Intraspecic trait responses to
Tissue d13C correlates with water-use eciency, and d13C height were particularly important for SLA, which was the
is thus used as indicator of water stress. However, along trait with the highest frequency of signicant traitheight
vertical gradients in forests, interpretation may be con- relationships (45% of all species) and consistently showed
founded as the atmospheric d13C signature also shows a only negative trends. SLA captures essentials of leaf eco-
vertical trend (Quay, King & Wilbur 1989). Nevertheless, nomics (Wright et al. 2004) and is a suitable trait for intra-
the strongest increase in atmospheric d13C signature occurs specic adjustments because it can be relatively easily
within a few metres above the forest oor due to soil respi- adjusted by varying the size, number and cell wall thick-
ration, and above this zone, the gradient in d13C is gener- ness of dierent leaf cell types (Shipley et al. 2006; Kiche-
ally weak (Quay, King & Wilbur 1989). In contrast, our nin et al. 2013). In general, although we cannot rule out
model predicted a linear trend with an average change of genetic variation as source of intraspecic trait variability,
~55 & in d13C from the trunk base to the upper canopy we argue that, considering the spatial scale of our study,
(Fig. 1g), suggesting that a large part of the observed vari- phenotypic trait plasticity in response to the environment
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
8 G. Petter et al.
is probably more important (also see Grassein, Till-Bott- tribution. Sides et al. (2014) conducted a comparable
raud & Lavorel 2010). study of 21 herbaceous perennials along an elevational
Interestingly, the second most frequent signicant intra- gradient of c. 700 m, using CV as measure of trait
specic traitheight relationships were found for LDMC, variability. They observed a stronger correlation between
which, in contrast, was rather weakly correlated with intraspecic trait variability in SLA and elevational range
height at the community level. The high frequency might (R2 = 051). The weaker correlation in our study might
partly be explained by correlations between SLA and partly be explained by the uncertainties associated with
LDMC, which were often stronger at the species level than the height gradient as approximation of environmental
at the community level (compare Figs S4 and S5). How- gradients. Furthermore, Sides et al. (2014) pointed out
ever, species-specic dierences in strategies might also that intraspecic trait plasticity should be essential when
play a role (Wilson, Thompson & Hodgson 1999). For strong trends in community mean trait values exist. Com-
instance, in Elaphoglossum doanense only LDMC was munity mean trends were less pronounced in our study,
strongly correlated with height (R2 = 081), while there indicating that height was a weaker lter than elevation.
was no signicant correlation for any of the other traits. In summary, epiphyte species that can adjust their leaf
Although intraspecic trait response to height was com- traits to the environment can potentially occupy larger
mon in epiphytes, the absence of a signicant intraspecic vertical ranges. However, the substantial amount of unex-
traitheight relationship was not always accompanied by plained variance also emphasizes that unstudied charac-
limited trait variability. Most species that lacked a signi- teristics (e.g. root traits, specic morphological and
cant traitheight correlation had a pronounced intraspe- physiological characteristics) or other processes (e.g. ger-
cic trait variability unrelated to height. Apart from the mination, seedling survival) might be likewise important
uncertainties associated with height as proxy for environ- in explaining why some species are restricted to smaller
mental gradients, plant size and age are additional sources vertical ranges than others.
of intraspecic trait variability (Zotz 2000; Wanek et al.
2002; Hietz & Wanek 2003), which might weaken trait
H4 TRAIT MEANS, TRAIT SYNDROMES AND TRAIT
height relationships. It is therefore striking that height
HEIGHT RELATIONSHIPS DIFFER AMONG TAXONOMIC
emerged as signicant factor for intraspecic changes in
GROUPS
leaf traits.
In summary, our results corroborate the growing evi- For most traits, we found signicant dierences in trait
dence that not only dierences in trait means, but also dif- means between taxonomic groups, which partly conrms
ferences in intraspecic trait response to environmental our hypothesis. Trait dierences were, however, often only
gradients among species are non-negligible aspects of com- signicant between certain pairs of taxonomic groups. The
munity assembly (Bolnick et al. 2011; Kichenin et al. frequent absence of pairwise dierences was mainly due to
2013). the high trait variation between species within taxonomic
groups, and, to a lesser extent, due to similarities in group
trait means. The pronounced within-group trait variation
H3 VERTICAL RANGES OF EPIPHYTE SPECIES
and associated among-group trait overlap become appar-
CORRELATE WITH THEIR INTRASPECIFIC TRAIT
ent when comparing species trait syndromes in the multi-
VARIABILITY
variate trait space (Fig. 2). Nevertheless, the unique
Intraspecic variability explained almost one-third of the tendencies within taxonomic groups indicate that some leaf
observed variance in our trait data, which is in the same traits are taxonomically conserved (Fig. 2).
range as observed for terrestrial plants (Albert et al. 2010; The marked dierences in morphological leaf traits
Hulshof & Swenson 2010). This supports previous nd- between orchids and ferns were consistent with previous
ings underlining the importance of considering trait vari- studies reporting orchids having thicker leaves and lower
ability not only between but also within species (Albert SLA (Stuntz & Zotz 2001; Cardel us & Mack 2010). Com-
et al. 2010). Such intraspecic variability seems to be munity means of leaf thickness and SLA were strongly
important for species spatial distribution, as our results correlated with height, which emphasizes their functional
supported the hypothesis that species occupying larger relevance along the vertical gradient. It is thus unsurpris-
vertical ranges tended towards higher leaf trait variability ing that dierences in these traits were reected in dierent
(Fig. S11). The inherent ability of species to vary their height distributions of these taxa (Table 1; also see Fig. 3).
leaf traits might increase their ability to tolerate a wider This pattern might be partly explained by environmental
range of environmental conditions (Van Valen 1965). In ltering of species with unsuitable traits, but intraspecic
this context, it is not surprising that TR explained a lar- leaf trait adjustments, particularly for SLA, might also be
ger amount of variation in species vertical ranges than important. Interestingly, SLA was the only extensively
CV (TR: R2 = 024, CV: R2 = 010). This is because TR sampled trait without signicant dierences in slopes or
is based on extreme trait values and is thus a better intercepts among the taxonomic groups (Fig. S3a). This
approximation of the theoretical maximal trait range of a suggests an optimal SLA value at a given height indepen-
species, whereas CV is aected by the trait frequency dis- dent of taxonomic group and further indicates that
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
Functional traits of vascular epiphytes 9
Light Species
13C*
intensity Individuals
30 Leaf
Temper- thickness*
Height above ground (m)
ature
LWCarea*
Temp.
fluctuation SLA* Chlarea
20
Wind Chlmass* Narea
speed
Nmass*
Air Evapo-
humidity transp.
10
15N
Substrate 13C
diameter (air) LDMC
0
Aroids Bromeliads Orchids Ferns
Fig. 3. Schematic diagram illustrating main ndings. Arrows on left side: environmental factors commonly changing with height above
ground within forests. In this study, only the vertical light gradient was measured (Fig. S1). Boxplots: height distribution of the major tax-
onomic groups of epiphytes at the study site in Panama. Height distributions are based on either the height of all individuals or the mean
height of each species. Boxplots depict median heights (horizontal line), interquartile ranges (boxes) and approximate 95% condence
intervals (whiskers). Outliers are not shown. Arrows on right side: signicant vertical leaf trait gradients at the study site (trait abbrevia-
tions as in Fig. 1). Leaf traits showing pronounced changes in community trait means with height are marked by *.
environmental changes along the height gradient act as a photosynthetic nitrogen-use eciency (PNUE) were
particularly strong lter on SLA. observed among epiphyte taxa, with aroids having the low-
The most striking among-group dierences were est PNUE, and bromeliads having the highest (Stuntz &
observed for LDMC, with LDMC of ferns being twofold Zotz 2001). Thus, for a given nitrogen concentration, the
higher, on average, than in all other groups. LDMC values photosynthetic capacity was higher in bromeliads. This
have not been reported for many epiphyte species, but suggests that the observed among-taxa dierences in leaf
Woods (2013) also found high LDMC values in two El- nitrogen cannot be used to infer similar dierences in pho-
aphoglossum species and low values in one Microgramma tosynthetic capacity.
species. This agrees with our results and shows that the We did not observe signicant among-group dierences
LDMC of fern species can dier substantially (Table S1). in d13C and d15N values. Our results largely agree with
However, the large number of fern species sampled in our observations along an elevational gradient in Costa Rica
study (n = 24) suggests that high LDMC values are more (Cardelus & Mack 2010). In contrast, Hietz, Wanek &
common in ferns. Popp (1999) observed signicantly depleted d15N values in
Aroids had the highest leaf nitrogen and chlorophyll bromeliads, but these were mainly of atmospheric habit.
concentrations, whereas bromeliads consistently had the All these studies found high variability in isotope ratios of
lowest. In fact, both traits were correlated (r = 067; Table species within taxonomic groups, suggesting that the envi-
S5). Chlorophyll concentrations have not yet been com- ronmental conditions and species-specic characteristics
pared among major epiphyte taxa, but our results agree are more important in determining isotope ratios in leaf tis-
with reported leaf nitrogen values. For example, Stuntz & sue of individual epiphytes than their taxonomic aliation.
Zotz (2001) also found the highest nitrogen concentrations Compared to global trait means of non-epiphytic taxa
in aroids. The lowest nitrogen concentrations, in turn, (TRY; Kattge et al. 2011), both low nitrogen concentra-
were consistently observed in bromeliads (Hietz, Wanek & tions and thick leaves are particularly noticeable dier-
Popp 1999; Stuntz & Zotz 2001; Cardel us & Mack 2010). ences (see Table S1 for details). These trait dierences can
An increase in leaf nitrogen content is usually associated be regarded as an adaptation of epiphytes to an environ-
with an increase in photosynthetic capacity (Stuntz & Zotz ment in which water and nutrients are only intermittently
2001; Wright et al. 2004). Interestingly, dierences in available.
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
10 G. Petter et al.
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology
Functional traits of vascular epiphytes 11
Rijkers, T., Pons, T.L. & Bongers, F. (2000) The eect of tree height and Wright, I.J., Reich, P.B., Cornelissen, J.H.C., Falster, D.S., Groom,
light availability on photosynthetic leaf traits of four neotropical species P.K., Hikosaka, K. et al. (2005) Modulation of leaf economic traits
diering in shade tolerance. Functional Ecology, 14, 7786. and trait relationships by climate. Global Ecology and Biogeography,
Rozendaal, D.M.A., Hurtado, V.H. & Poorter, L. (2006) Plasticity in leaf 14, 411421.
traits of 38 tropical tree species in response to light; relationships with Zotz, G. (2000) Size-related intraspecic variability in physiological traits
light demand and adult stature. Functional Ecology, 20, 207216. of vascular epiphytes and its importance for plant physiological ecology.
Schimper, A.F.W. (1888) Die Epiphytische Vegetation Amerikas. Gustav Fi- Perspectives in Plant Ecology, Evolution and Systematics, 3, 1928.
scher, Jena. Zotz, G. (2004) How prevalent is crassulacean acid metabolism among vas-
Shipley, B., Lechowicz, M.J., Wright, I. & Reich, P.B. (2006) Fundamental cular epiphytes? Oecologia, 138, 184192.
trade-os generating the worldwide leaf economics spectrum. Ecology, Zotz, G. (2007) Johansson revisited: the spatial structure of epiphyte assem-
87, 535541. blages. Journal of Vegetation Science, 18, 123130.
Sides, C.B., Enquist, B.J., Ebersole, J.J., Smith, M.N., Henderson, A.N. & Zotz, G. (2013) The systematic distribution of vascular epiphytes a critical
Sloat, L.L. (2014) Revisiting Darwins hypothesis: does greater intraspe- update. Botanical Journal of the Linnean Society, 171, 453481.
cic variability increase species ecological breadth? American Journal of Zotz, G. & Assho, R. (2010) Growth in epiphytic bromeliads: response to
Botany, 101, 5662. the relative supply of phosphorus and nitrogen. Plant Biology, 12, 108
Sims, D.A., Gebauer, R.L.E. & Pearcy, R.W. (1994) Scaling sun and shade 113.
photosynthetic acclimation of Alocasia macrorrhiza to whole plant per- Zotz, G. & Schultz, S. (2008) The vascular epiphytes of a lowland forest in
formance - II. Simulation of carbon balance and growth at dierent pho- Panama - species composition and spatial structure. Plant Ecology, 195,
ton ux densities. Plant, Cell & Environment, 17, 889900. 131141.
Stuntz, S. & Zotz, G. (2001) Photosynthesis in vascular epiphytes: a survey
of 27 species of diverse taxonomic origin. Flora, 196, 132141.
Van Valen, L. (1965) Morphological variation and width of ecological Received 28 May 2014; accepted 26 April 2015
niche. American Naturalist, 99, 377390. Handling Editor: Jennifer Baltzer
Violle, C., Navas, M.-L., Vile, D., Kazakou, E., Fortunel, C., Hummel, I.
et al. (2007) Let the concept of trait be functional!. Oikos, 116, 882892.
Wanek, W., Arndt, S.K., Huber, W. & Popp, M. (2002) Nitrogen nutrition Supporting Information
during ontogeny of hemiepiphytic Clusia species. Functional Plant Biol-
ogy, 29, 733740. Additional Supporting information may be found in the online
Wania, R., Hietz, P. & Wanek, W. (2002) Natural 15N abundance of version of this article:
epiphytes depends on the position within the forest canopy: source
signals and isotope fractionation. Plant, Cell and Environment, 25, Fig. S1. Vertical light gradient at the study site.
581589. Fig. S2. Trait-height relationships at community level (census
Wilson, P., Thompson, K. & Hodgson, J. (1999) Specic leaf area and leaf
data).
dry matter content as alternative predictors of plant strategies. New Phy-
tologist, 143, 155162. Fig. S3. Trait-height relationships of taxonomic groups.
Woods, C. (2013) Factors Influencing the Distribution and Structure of Trop- Figs S4-S9. Intraspecic trait-height relationships.
ical Vascular Epiphyte Communities at Multiple Scales. PhD thesis, Fig. S10. Variance partitioning.
Clemson University, South Carolina.
Fig. S11. Relationship between trait variability and vertical range.
Wright, S., Horlyck, V., Basset, Y., Barrios, H., Bethancourt, A., Bohlman,
S. et al. (2003) Tropical Canopy Biology Program, Republic of Panama. Table S1. Detailed trait information on all sampled species.
Studying Forest Canopies From Above: The International Canopy Crane Table S2. AIC-based comparisons of GLMMs testing H4.
Network (eds S.J. Wright, V. Horlyck & Y. Basset), pp. 137155. United Tables S3S4. Summary statistics of LMs (community trends).
Nations Environmental Programme & Smithsonian Tropical Research Table S5. Pairwise correlations between traits.
Institute, New York.
Table S6. AIC-based comparisons of LMs testing H2.
Wright, I.J., Reich, P.B., Westoby, M., Ackerly, D.D., Baruch, Z., Bongers,
F. et al. (2004) The worldwide leaf economics spectrum. Nature, 428, Table S7. Proportion of species with signicant traitheight rela-
821827. tionships.
2015 The Authors. Functional Ecology 2015 British Ecological Society, Functional Ecology