Drivers of morphological diversity and distribution in the Hawaiian fern flora: Trait associations with size, growth form, and environment†
The authors thank D. Palmer for his committed scholarship, which made this work possible, and T. Ranker, K. Mehltreter, and three anonymous reviewers for helpful comments on the manuscript. This work was supported by an NSERC international postgraduate scholarship #318857 and by National Science Foundation grant #0753233.
Abstract
• Premise of the study: Hawaii is home to 238 native and 35 alien fern and lycophyte taxa distributed across steep gradients in elevation and resource availability. The fern flora spans a wide range of growth forms, with extraordinary diversity in morphology and plant size. Yet the potential factors underlying this diversity have remained enigmatic.
• Methods: We used a trait database generated from the most recent and comprehensive survey of Hawaiian ferns and lycophytes to test hypotheses of size-scaling and trait associations with environment and growth form as factors underlying this diversity. We also tested relationships among morphology, taxon abundance and distribution and identified key differences between native and alien taxa.
• Key results: Strong trait–trait relationships included geometric scaling of plant dimensions with a tendency for more divided fronds in larger ferns. Trait–environment relationships independent of size included more divided fronds at higher elevation, longer blades in shaded habitats, and fronds with shorter stipes and fewer pinnae in drier habitats. Growth forms differed in mean size with epiphytic and epipetric taxa smaller than terrestrial ferns. Plant size was independent of taxon abundance and distribution across islands, and native and alien ferns did not differ in mean size. Alien taxa were more abundant, especially at lower elevations, apparently due to human land use.
• Conclusions: These relationships point to linkages of fern form and demography with biogeography and highlight potential flora-scale physiological and morphological adaptations in ferns across contrasting environments.
The Hawaiian archipelago is the most isolated group of islands (14), a hotspot of biodiversity (64) with a flora assembled over tens of millions of years (72). The flora has extremely high endemism (87; 100): 89% for angiosperm species (99) and 77% for ferns and lycophytes (32). Given long distance traveling of spores (15; 79; 76), the evolution of fern endemism in Hawaii is especially striking because it relies on barriers to gene flow, such as polyploidy or microallopatry due to sharp island topographies (109; 58; 108), and on rare conditions such as self-fertilization, apomixis, or dispersal of multiple individuals to establish a population (29; 77). Colonization and speciation events have led to the establishment of 273 fern and lycophyte taxa that comprise one-sixth of all Hawaiian vascular plant species (32) and include keystone species in many wet forests (23, 24; 3).
Hawaiian ferns are highly variable in size and form (69; 85), but this variation has remained unexplained. Hawaiian ferns and lycophytes derive from at least 120 original colonizers from all major floristic regions (30; 31; 85; 21; 32). Given colonization from such diverse origins, morphological diversity may persist because of evolutionary obstacles that drive phylogenetic and niche conservatism, or increase through novel adaptations. A critical step forward is to test hypothesized trait associations underlying current realized patterns. We tested for (1) size-scaling of plant parts, and trait associations with (2) environment, (3) growth form, (4) abundance, (5) distribution across islands, and (6) differences between native and alien taxa.
Plant size and dimensions influence resource capture (86; 71; 26; 82). The retention of a genetic-developmental “blueprint” that maintains proportionality and function (bauplan sensu 98) may stabilize allometries among organ and body size. We hypothesized geometric scaling among frond parts (except pinnae number) based on reports for tree fern species Cibotium glaucum and Cyathea caracasana (4; 3) and that frond dissectedness would increase with size to reduce the boundary layer for heat dissipation (97) and/or to reduce investment in support tissues (e.g., vasculature, rachis) that scale disproportionately with leaf size at a cost to photosynthesis per mass (65).
Environmental pressures may have driven trait evolution or assembly of ferns and lycophytes along resource gradients, as found in several Hawaiian groups (80; 13; 37; 18; 22). Because fern species can occupy a wide variety of habitats with respect to moisture and irradiance (67; 1; 49), we expected trait differences for taxa from shaded and moist sites relative to taxa from open and dry sites (44; 24; 104; 52). Plants competing for light in shaded and moist environments would be larger with larger frond parts to increase surface area for light capture per investment in support (4), whereas plants from open and dry environments would be smaller with greater dissection to minimize the boundary layer for convective cooling (94). At higher elevation, irradiance can cause photodamage and overheating, especially given low nutrient availability (57; 56). We expected high elevation taxa to be smaller and more dissected (95, 80), with lamina pubescence to reflect light and reduce heat load (25; 104).
We hypothesized that morphology would also vary with growth form. Plants with epipetric and epiphytic habits (dwelling on rocks and other plants, respectively) may benefit from small plant and frond sizes given on average poorer resource availability and access to substrate support (103). By contrast, terrestrial taxa may be more vulnerable to mechanical damage, herbivory, and climbing plants and may benefit from protective scales (68; 67).
Morphological variation may be associated with abundance and distribution. We hypothesized that larger taxa with larger parts, and presumably faster rates of resource uptake (53), would have greater abundances and wider distributions across the Hawaiian Islands. We expected aliens to have larger plant and frond sizes, as found in angiosperms (73; 54), and to have achieved comparable abundances to natives, despite their recent introduction (10; 34). Given that many invasive species compete strongly in disturbed, resource-rich sites and have high dispersability (19; 73), we expected aliens to be common in open, low elevation areas and to be distributed across more islands than the native taxa. We expected that alien ferns would frequently exhibit the terrestrial growth form because, historically, terrestrial ferns have been cultivated more commonly in Hawaii than ferns from other habits (106, 26; 75).
As an approach complementary to the detailed comparison of sets of fern taxa (e.g., 23, 24; 2), we implemented a broad database analysis testing these predictions for 25 variables and 273 taxa on a florawide scale to determine trait associations contributing to Hawaiian fern morphological diversity and distributions.
MATERIALS AND METHODS
Creating the database
We compiled information on morphological traits, growth form, abundance, distribution, habitat, and environment of Hawaiian ferns and lycophytes from the most recent and comprehensive survey (Table 1; 69), based on detailed field observations as well as examination of specimens from 15 herbaria. We assembled this information into a trait database that we pruned to 25 variables with sufficient representation for analysis. The database contains information for 273 species, hybrid species, subspecies, varieties, and forms. Given the strong variation among these taxa and the frequently poor species resolution of Hawaiian ferns and the likelihood of taxonomic revision (32), we considered these as distinct taxa in the database. For continuous quantitative variables (i.e., plant dimensions such as frond width and length), we used mean taxon values or we averaged minimum and maximum values. We ordinalized descriptive data for abundance and plant size to create quantitative variables. Abundance classes aggregated the descriptions from 69, in which some taxa were given multiple descriptions because of variation across islands: 1 = extinct; 2 = endangered; 3 = rare; 4 = uncommon everywhere, or rare on one or more islands; 5 = common on one or more islands, uncommon elsewhere; 6 = not endangered or locally common; 7 = common. Plant size descriptions were those given by 69: 1 = small; 2 = small–medium; 3 = medium, small–large, or small–medium–large; 4 = medium–large; 5 = large.
| Ferns (families, genera, species) | Lycophytes (families, genera, species) | |||||
|---|---|---|---|---|---|---|
| Variables | Endemic | Indigenous | Alien | Endemic | Indigenous | Alien |
| Database | ||||||
| All Taxa | 173 (21, 46, 99) | 41 (15, 20, 35) | 33 (13, 23, 30) | 15 (3, 4, 9) | 8 (1, 3, 7) | 3 (1, 1, 3) |
| Growth form | ||||||
| Terrestrial | 50 (11, 20, 41) | 16 (6, 8, 16) | 15 (8, 10, 15) | 3 (2, 2, 3) | 4 (1, 2, 4) | 3 (1, 1, 3) |
| Epiphytic | 35 (6, 11, 31) | 7 (6, 6, 7) | 2 (1, 2, 2) | 4 (1, 1, 4) | 2 (1, 1, 2) | 0 (0, 0, 0) |
| Epipetric | 5 (5, 5, 5) | 6 (5, 5, 6) | 3 (2, 3, 3) | 0 (0, 0, 0) | 0 (0, 0, 0) | 0 (0, 0, 0) |
| Environment | ||||||
| Open | 21 (13, 15, 19) | 10 (7, 8, 10) | 10 (5, 8, 10) | 5 (3, 4, 4) | 4 (1, 3, 4) | 1 (1, 1, 1) |
| Shade | 109 (16, 38, 86) | 21 (12, 13, 20) | 13 (8, 12, 13) | 5 (2, 2, 4) | 3 (1, 1, 3) | 0 (0, 0, 0) |
| Wet | 115 (18, 41, 89) | 20 (12, 13, 19) | 14 (10, 13, 14) | 7 (3, 3, 6) | 6 (1, 2, 6) | 1 (1, 1, 1) |
| Dry | 11 (6, 8, 10) | 9 (6, 7, 9) | 6 (2, 5, 6) | 1 (1, 1, 1) | 1 (1, 1, 1) | 0 (0, 0, 0) |
| High elevation | 63 (15, 25, 46) | 13 (9, 9, 13) | 11 (8, 10, 11) | 3 (1, 1, 2) | 1 (1, 1, 1) | 0 (0, 0, 0) |
| Low elevation | 67 (13, 26, 51) | 18 (9, 12, 16) | 12 (8, 11, 12) | 2 (1, 1, 1) | 0 (0, 0, 0) | 0 (0, 0, 0) |
| Demography | ||||||
| Common | 68 (13, 26, 60) | 19 (10, 11, 19) | 13 (7, 11, 13) | 3 (3, 3, 2) | 3 (1, 2, 3) | 1 (1, 1, 1) |
| Uncommon | 52 (17, 26, 40) | 7 (5, 6, 7) | 2 (2, 2, 2) | 8 (2, 2, 4) | 1 (1, 1, 1) | 0 (0, 0, 0) |
| Morphology | ||||||
| Simple fronds | 4 (1, 2, 4) | 2 (2, 2, 2) | 0 (0, 0, 0) | n/a | n/a | n/a |
| Compound fronds | 92 (12, 29, 73) | 20 (7, 11, 19) | 19 (8, 14, 19) | n/a | n/a | n/a |
| Pubescent lamina | 18 (7, 12, 16) | 1 (1, 1, 1) | 3 (3, 3, 3) | 0 (0, 0, 0) | 0 (0, 0, 0) | 0 (0, 0, 0) |
| Glabrous lamina | 140 (20, 42, 93) | 37 (14, 19, 33) | 27 (12, 20, 27) | 15 (3, 4, 9) | 8 (1, 3, 7) | 3 (1, 1, 3) |
| Stipe scales | 60 (10, 20, 52) | 12 (7, 8, 12) | 10 (7, 9, 10) | n/a | n/a | n/a |
| Stipe no scales | 54 (16, 28, 49) | 13 (6, 8, 13) | 14 (7, 10, 14) | n/a | n/a | n/a |
| Rhizome scales | 16 (7, 8, 15) | 5 (3, 3, 5) | 5 (2, 5, 5) | n/a | n/a | n/a |
| Rhizome no scales | 87 (17, 34, 81) | 24 (11, 13, 23) | 15 (7, 10, 15) | n/a | n/a | n/a |
For nominal variables, we sorted taxa into categorical groups. For growth form, we compared epiphytic vs. terrestrial and epipetric vs. nonepipetric taxa (insufficient data excluded arboreal and aquatic habits from analysis of trait–habit associations); for abundance, common vs. uncommon; we categorized plants as alien or native; for habitat and elevation as wet vs. dry, open vs. shade, and high vs. low elevation range; and for morphology as simple vs. compound fronds, glabrous vs. pubescent lamina, and glabrous vs. scaly stipes and rhizomes. Taxa described as expressing more than one growth form were included as either epipetric or epiphytic to recognize the expression of these specialized habits (e.g., 103). Taxa were defined as common when their abundance class (see above) ranged from common to not endangered, and as uncommon when this ranged from uncommon to extinct. Taxa were identified as native if indigenous or endemic; in our analysis, hybrids between aliens and natives were considered as aliens.
In the categorization of site environment and distribution, we aimed to capture fundamental differences in irradiance and moisture habitats and elevation range. By necessity, our environmental variables are coarse-grained given the available information for this flora-scale analysis, and we therefore focused on patterns of species and trait variation associated with gross environmental differences, rather than variation within similar habitat types (cf. 5; 16). Thus, open sites were exposed habitats lacking forest canopy, including exposed lava flows, rock outcrops and grass meadows, and shade sites included forests, sites beneath rock wall overhangs, and shaded slopes. Wet sites were those described as wet, moist or mesic, including stream areas, waterfalls, bogs, and moist/rain-forests, and dry sites were those described as xeric, including lava flows, rocky outcrops, and dry forests. These habitat categories would likely also differ in soil composition and biota. Taxa that ranged across both open and shade or wet and dry environments were listed within the less common (and presumably more specialized) open and dry environment categories. For mean elevation range, high and low referred respectively to above and below the all-taxon mean elevation of 1012 m a.s.l. Taxa with undivided (entire) leaves were categorized as simple and those with any number of pinnate divisions as compound. Taxa with and without lamina hairs were considered pubescent and glabrous, respectively. Similarly, taxa with stipe or rhizome scales were included in the scales category, and those without as having no scales.
Data sets, statistical analyses, and meta-hypothesis testing
Trait information was derived from a database that included the most information available on Hawaiian ferns to highlight strong patterns in a highly diverse flora. However, we acknowledge several provisos and emphasize that limitations in this approach will require future study on smaller groups to fully confirm hypotheses. Most importantly, our analysis relied on taxon descriptions, and these were sometimes incomplete. A key limitation of using a precompiled trait database (such as a flora) for comparative analysis is that not all taxa are represented across all variables, leading to unequal representation of groups in some comparisons, analogous to phylogenetic analyses in which larger data sets are inherently more sparse (83). However, the “total evidence” approach (55) provides the best present knowledge by maximizing the number of taxa and characters to increase biological information without necessarily reducing resolution (e.g., 84). To minimize type I and type II errors, we omitted traits for which data were available for only few taxa (e.g., caudex dimensions, which only applied to tree ferns). Our minimum taxon representation for a given trait was 18; for traits that involved multiple character states, we required this level of taxon representation for each character state. We then generated trait dataframes for three groups: (1) ferns and lycophytes including natives and aliens, (2) only native ferns and lycophytes, and (3) only native ferns. Using these three dataframes, we tested trait–trait (T×T), trait–environment (T×E), trait–demography/distribution (T×D), environment–environment (E×E), environment–demography/distribution (E×D), and demography–demography (D×D) relationships. Our presentation of results focuses only on relationships for which we had a priori hypotheses. Additionally, we determined all possible variable relationships to investigate the percipience of our analyses (see below). We tested trait associations using Pearson and Spearman correlation coefficients on both untransformed and log-transformed data to accommodate linear and nonlinear relationships for the three dataframes (all ferns and lycophytes; native ferns and lycophytes; and native ferns). We considered a correlation significant if Pearson and Spearman correlations were significant at P ≤ 0.05.
We used analyses of variance (ANOVAs) to test for variation in quantitative variables between categorical groups using dataframes 1 and 3 (all ferns and lycophytes and just native ferns, respectively). Relationships among categorical variables were tested using χ2 contingency tables and Fisher's exact tests for all taxa and only native ferns, respectively. Statistics were calculated using the program SPSS 15.0 (SPSS, Chicago, Illinois, USA).
For allometries between size traits, we fitted standardized major axes (SMA using the program SMATR; 102) to log-transformed data, equivalent to fitting power law curves; log y = α + β × log x (88; 81). The use of SMA is appropriate for determining functional relationships among two independent variables (88; 81; 102). We tested for deviation of SMA slopes from those expected by geometric scaling, i.e., the retention of proportional dimensions with increasing size (66; Sack et al., 2003, 81; 102); for example, linear dimensions should scale with a β of 1, while a linear dimension should scale with an area dimension with a β of 0.5.
To quantify our ability to currently explain trait variation among native Hawaiian ferns, we evaluated the percipience (i.e., the power to perceive differences) of our analysis and the predictive and explanatory success of our hypotheses, comparing results to those expected from random chance alone. We used a novel, simple multivariate approach based on probability theory (89; 70) to assess the strength of our overall analysis given multiple tests relative to random associations. We quantified overall percipience as the number of significant results out of the total number of tests: a value of 1 would indicate that hypotheses were targeting real variation in the data. This proportion was compared against the 0.05 expected from random chance. The predictive power of our hypotheses, i.e., the ability of the hypotheses to anticipate significant differences and their direction a priori, was calculated as the number of supported predictions out of the total number of tests and compared to the random chance model of 0.025 (the 0.05 probability of a significant result due to chance alone multiplied by the 0.5 probability of a correlation in the direction hypothesized). Finally, the explanatory power of hypotheses, i.e., the ability of a priori hypotheses to explain the direction of observed significant differences post hoc, was determined as the number of supported predictions divided by the total number of significant findings and compared to the random chance model of 0.5 (the probability that significant differences were in the expected direction by chance alone). A high explanatory power indicates that proposed mechanisms are sufficient to explain the significant trait associations.
Phylogenetic analyses
Standard trait correlations can be used as evidence for true and realized association among plant traits, but are weak evidence of correlated evolution because they do not distinguish cases of preadaptation and ecological sorting from in situ evolution (28). For instance, a strong bivariate trait correlation may indicate an evolutionary event and subsequent trait conservatism during diversification or repeated trait coevolution. Thus, to ensure evolutionary correlations the common practice is to retest correlations using phylogenetically independent contrasts (PICs) that account for species’ shared history (40; 60). Using the best available phylogenetic data for Hawaiian fern lineages, the phylogenies for the endemic genus Adenophorus (78) and Dryopteris (31), we calculated independent contrast correlations for traits represented by seven or more taxa using the AOTF function in the program PHYLOCOM (105). We note that in the absence of more phylogenetic data, ahistorical analyses provide strong insight toward mechanistically grounded and ecologically important trait correlations but that further phylogenetic resolution and analysis are needed to tease apart patterns of sorting vs. in situ adaptation.
RESULTS
Trait–trait relationships: Scaling of size and morphology
Hawaiian ferns exhibit tremendous variation in plant size, from the small grammitids and filmy ferns (Polypodiales; Hymenophyllales) and aquatic ferns (Salviniales) to the large arborescent ferns (Cyatheales). Taxa also varied widely in frond size and shape (Table 2). Fern fronds ranged nearly 500-fold in length among both native (from Gonocormus minutus to Cibotium glaucum) and alien taxa (from Azolla filiculoides to Angiopteris evecta). Native fern fronds also ranged from simple leaf types (e.g., Adenophorus oahuensis) up to 5-pinnate divisions (e.g., Asplenium schizophyllum), and alien fronds from 2 to 8 pinnate divisions (e.g., Diplazium esculentum and Asplenium hispidulum).
| Native ferns | All taxa | |
|---|---|---|
| Trait | Min, Mean, Max (N) | Min, Mean, Max (N) |
| Size-related | ||
| Species plant size* | 1, 2.73, 5 (123) | 1, 2.76, 5 (148) |
| Rhizome diam. (mm) | 0.3, 6.59, 37.5 (37) | 0.3, 6.53, 37.5 (40) |
| Stipe diam. (mm) | 0.3, 1.79, 4.5 (24) | 0.3, 1.79, 4.5 (24) |
| Stipe length (cm) | 0.1, 17.27, 100 (17) | 0.1, 19.09, 100 (18) |
| Frond width (cm) | 0.25, 9.69, 40 (55) | 0.25, 15.66, 300 (58) |
| Frond length (cm) | 1.5, 77.98, 700 (136) | 1.5, 81.66, 700 (157) |
| Blade width (cm) | 0.3, 11.87, 57.5 (16) | 0.3, 11.77, 57.5 (18) |
| Blade length (cm) | 0.75, 30.53, 110 (29) | 0.75, 30.01, 110 (31) |
| USEG width (mm) | 0.2, 3.63, 10 (34) | 0.2, 4.62, 20 (37) |
| USEG length (mm) | 0.5, 11.6, 37.5 (20) | 0.5, 13.22, 55 (23) |
| Pinna width (mm) | 0.95, 30.32, 180 (29) | 0.95, 36.96, 260 (31) |
| Pinna length (mm) | 1.5, 127.6, 850 (43) | 1, 168.62, 1500 (54) |
| No. pinnae pairs | 1.5, 20.099, 40 (71) | 1.5, 19, 40 (80) |
| Shape | ||
| No. pinnate divisions | 0, 1.78, 4.5 (118) | 0, 1.81, 5 (136) |
| No. pinnate variations | 1, 1.52, 4 (118) | 1, 1.55, 7 (136) |
| Demography | ||
| Abundance* | 1, 5.29, 7 (146) | 1, 5.36, 7 (179) |
| Distribution | 0, 5.06, 8 (211) | 0, 4.99, 8 (273) |
| (No. islands) | ||
| Environment | ||
| Elevation (m a.s.l.) | 81, 1042.33, 2470 (170) | 81, 1012.69, 2470 (208) |
- Notes: *, designates ordinal scale; a.s.l., above sea level; diam., diameter; Min, minimum; Max, maximum; USEG, ultimate segment
We found strong scaling of the dimensions of fronds, frond components, and rhizomes with plant size (Fig. 1; r = 0.45–0.86, P < 0.05, N = 11–121; Appendices S1 to S3, see Supplemental Data with the online version of this article; relationships were significant for all dataframes unless otherwise specified, and both rp and rs values are presented in Appendices S1–S3, with the highest value reported in the text). We also found tight coordination of length and width for fronds (r = 0.98, P < 0.001, N = 58), blades (r = 0.61, P < 0.001, N = 7), pinnae (r = 0.95, P < 0.001, N = 29) and ultimate segments (r = 0.65, P = 0.012, N = 14; see Fig. 1). As expected, pinnae number was independent of plant size (r = −0.008, P = 0.945, N = 72). Frond component dimensions were also positively correlated with one another: frond length was correlated with length and width of the blade (r = 0.90-0.98, P < 0.001, N = 7–10), the pinna (r = 0.94–0.95, P < 0.001, N = 28–47), and the ultimate segment length (r = 0.68, P = 0.002, N = 18). Wider fronds had larger pinnae (r = 0.91–0.99, P < 0.001, N = 15–18) but fewer pinnae pairs (r = −0.61 P < 0.001, N = 36), and blade and pinna lengths were correlated (r = 0.99, P < 0.001, N = 6). We also found scaling of stipe diameter with laminar components, i.e., frond, blade, and pinna lengths and ultimate segment width (r = 0.61–0.82, P = 0.005–0.041, N = 9–20), and greater rhizome diameter for taxa with longer fronds (r = 0.76, P < 0.001, N = 28).
Left: Size scaling in Hawaiian ferns as seen by positive correlations (Spearman, rS; Pearson, rP) between frond length and (A) frond width, (B) pinna width (inset: pinna length), (C) ultimate segment length, and (D) rhizome diameter. Open symbols, native fern taxa; filled symbols, aliens. Lycophytes were excluded because they lack the frond morphology of ferns. * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001. Right: Schematic of fern indicating the frond aboveground and rhizome belowground. The frond is separated into a stipe (stem-like) and blade (lamina). The first order division of the blade is termed pinna, and the pinnae may be further divided, with the highest order division termed an ultimate segment (useg).
Across the flora, size differences were linked with other morphological characters. Larger taxa were more commonly glabrous, and smaller taxa were pubescent (P = 0.007, N = 136, 12; online Appendix S4), with both size and pubescence associated with habit (see following section). Supporting our hypothesis for greater frond dissection in larger ferns, the fronds of larger taxa tended to have compound blades (P < 0.001, N = 101, 6; Appendix S4), with more numerous pinnate divisions (r = 0.48, P < 0.001, N = 95; Appendices S1–S3) and greater variation in the number of these divisions (r = 0.31, P = 0.002, N = 95). Greater leaf dissection was also positively correlated with stipe diameter (r = 0.54, P = 0.014, N = 20), frond length (for native ferns only; r = 0.270, P = 0.005, N = 109; Appendix S3), frond and blade widths (r = 0.65–0.88, P ≤ 0.02, N = 6–48), blade length (r = 0.704, P = 0.023, N = 10), pinna size (r = 0.48–0.52, P ≤ 0.006 N = 27–44), and the number of pinnate variations (r = 0.63, P < 0.001, N = 136; Appendices S1–S3). Taxa that expressed more variability in pinnate divisions also had, on average, wider fronds (r = 0.30, P = 0.039, N = 38) because of their longer pinnae (r = 0.48, P = 0.004, N = 35).
The significant relationships among size traits did not deviate from geometric scaling (95% confidence interval tests). The mean standard major axis slope (β) determined for 15 relationships among continuous (non-ordinal) traits was 1.09 ± 0.077 (N = 6–55; online Appendix S6).
Trait–environment relationships: Trends with elevation
Fern form varied with elevation. At higher elevations, ferns exhibited a greater number of pinnate divisions (r = 0.20–0.21, P = 0.02–0.03, N = 113–127) and pinnae pairs (r = 0.34, P = 0.003, N = 76) and reduced pinna length (r = −0.38, P = 0.007, N = 48), but no significant greater tendency for pubescent over glabrous laminae (P > 0.05, N = 21, 187; Appendix S4). These trends for number and length of pinnae with elevation were apparently driven by the differences among native and alien taxa; the relationships were not supported when considering only native ferns (r = −0.16–0.20, P = 0.1–0.33, N = 40–68, Appendices S1–S3). The trends were independent of plant size, which did not correlate with elevation (r = 0.011, P = 0.90, N = 135; Appendices S1–S3).
Trait–environment relationships: Open vs. shade and wet vs. dry habitats
Plant traits correlated with irradiance and moisture habitat independently of plant size, which did not differ significantly between shade and open-habitat taxa (P = 0.39, N = 28, 134; Appendix S4) or between wet and dry habitats (P = 0.60, N = 19, 129; Appendix S4). Open fern sites also tended to be dry relative to shaded sites (P < 0.001, N = 194; online Appendix S5). Ferns that occupied open habitats tended to have shorter blades (P = 0.034, N = 8, 21; Fig. 2). Taxa from wetter sites tended to have blades with more pinnae pairs (P = 0.057, N = 57, 15 for all ferns; P = 0.046, N = 53, 11 for native ferns; Fig. 2).
Association of frond form with habitat moisture and irradiance in Hawaiian ferns (ANOVAs, log-transformed data). (A) Mean number of pinnae pairs is greater for wet-site taxa (WET). (B) Mean blade length is shorter in open sites (OPEN). * P ≤ 0.05.
Trait–demography relationships: Associations with abundance and distribution
Across all native and alien taxa and within the terrestrial, epiphytic, and epipetric growth forms considered individually, abundance and distribution were independent of plant size (r = 0.079–0.092, P = 0.270–0.401, N = 116–146; Appendices S1–S3). However, frond traits varied with taxon abundance and island distribution. Abundant taxa tended to have larger pinnae (r = 0.36–0.48, P = 0.016, N = 25–44) with wider ultimate segments (r = 0.70, P < 0.001, N = 23) and were reported on a greater number of the Hawaiian Islands (r = 0.54, P < 0.001, N = 179, Appendices S1–S3). Taxa from open sites also tended to be more abundant (P = 0.054, N = 37, 122; Appendix S5), an association apparently driven by alien ferns, because it was not apparent when only native ferns were considered (P = 0.21, N = 25, 107).
Taxa distributed across many islands tended to have longer fronds (r = 0.20, P = 0.01, N = 157) composed of smaller blades (r = −0.44 to −0.47, P = 0.01–0.05, N = 18–31; Appendices S1–S3) and longer ultimate segments (r = 0.43, P = 0.040, N = 23). When only native ferns were considered, taxa distributed across many islands tended to have greater frond widths (r = 0.28, P = 0.04, N = 55; Appendix S3) and fewer pinnae pairs (r = −0.27, P = 0.02, N = 71), and were more common in dry sites (P = 0.043, N = 20, 136; Appendix S4).
Differences in morphology and distribution across growth forms
The flora included terrestrial (e.g., Cibotium glaucum and Ophioglossum nudicaule), epiphytic (e.g., Asplenium nidum and Grammitis tenella) and epipetric taxa (e.g., Christella wailele and Diellia erecta), with the terrestrial growth form most common (see Table 1; 55.5% of taxa were terrestrial, 35.3% epiphytic and 9.2% epipetric taxa). Epiphytes tended to be smaller plants with smaller parts (Fig. 3; Appendix S4); terrestrial taxa were larger in plant size (P = 0.008, N = 75, 37; Fig. 3; Appendix S4), rhizome diameter (P = 0.003, N = 33), frond length and width (P < 0.001, N = 71, 29 and 28, 17, respectively; Appendix S5), and number of pinnate divisions (P = 0.044, N = 66, 22). Considering only the native ferns, terrestrial taxa also exhibited greater pinna widths than did epiphytes (P < 0.05, N = 4, 19; Appendix S4). Epiphytes more frequently occurred in shaded sites (P < 0.001; Appendix S5) and possessed simple leaves (P < 0.001, N = 22, 67) with lamina pubescence (P = 0.007, N = 51, 93) and lower incidence of stipe scales (P = 0.003, N = 30, 75). Epipetric taxa also tended to be smaller in stature (P < 0.05, N = 10, 102; Appendix S5; Fig. 3). Among native ferns, epipetric taxa had a lower mean elevation range than did nonlithophytes (P < 0.05, N = 11, 95; the relationship was nonsignificant when including aliens and lycophytes; P < 0.1, N = 14, 117).
Association of plant size with growth form in Hawaiian ferns (ANOVAs, log-transformed data). (A) Epipetric taxa (EPE) have a smaller mean plant size than nonepipetric taxa (NOT_EPE). (B) Epiphytic taxa (EPH) have a smaller mean plant size than terrestrial taxa (TER). * P ≤ 0.05, ** P ≤ 0.01.
Morphology, growth form, and demography for aliens vs. natives
Alien and native ferns did not differ in mean plant size (P = 0.28, N = 19, 129; Appendix S4), but contrasted in frequency of habit and demography. As for native ferns, aliens were more commonly terrestrial than epiphytic (P = 0.010, N = 51, 93; Appendix S5) with 80% terrestrial species and the rest divided across the epiphytic, epipetric, and aquatic forms (data not shown). Aliens had a lower mean elevation range (P < 0.001, N = 18, 189; Fig. 4) and a higher proportional distribution in open (P = 0.011; Appendix S5) and dry environments (P = 0.051) than did native taxa. Although on average aliens were not more widely distributed across the Hawaiian Islands (P > 0.1, N = 36, 237; Appendix S4), they were on average more abundant than native taxa (P = 0.007, N = 16, 162, Appendix S5; P = 0.020; Fig. 4).
Contrasting abundance and distribution of alien and native Hawaiian ferns (ANOVAs, log-transformed data). (A) Aliens are on average more abundant than native taxa and (B) occur at lower elevations. * P ≤ 0.05, ** P ≤ 0.01, *** P ≤ 0.001.
Percipience and predictive and explanatory power
For native ferns, the percipience and the predictive and explanatory power of our hypotheses all exceeded expectations from chance (online Appendix S7). Percipience and predictive power were highest for tests relating to size-scaling (50%), followed by those related to growth form (30%), demography (20–40%) and environmental variables (4–20%). The explanatory power was lowest for hypotheses of distribution (57%), but very high for those based on size-scaling (91%) and other hypotheses for trait-trait and trait-environment correlations (100%).
Phylogenetic analyses
We hypothesized that larger plants with larger parts would also tend to be more abundant and widely distributed. However, in Adenophorus, frond length negatively correlated with abundance (r = −0.65, P < 0.05; N = 10; online Appendix S8). In Dryopteris, we found a strong scaling of frond length and plant size (r = 0.86, P < 0.05; N = 11) and an association of taxon abundance with a wide distribution across islands (r = 0.72, P < 0.05; N = 8; Appendix S8).
DISCUSSION
This study highlighted novel, flora-scale relationships between structure and habitat and important differences between native and alien Hawaiian ferns. The trait linkages with plant size, growth form, demography, and biogeography presented here are a first indicator of potential morphological and physiological adaptations to contrasting environments across the Hawaiian fern flora.
Size-scaling of Hawaiian fern form
We found a strong coordination of plant size and the dimensions of fronds, frond parts, and rhizomes. These patterns indicated the maintenance of plant proportions with increasing size in Hawaiian ferns. This geometric scaling of plant parts extended to the flora scale, a trend previously documented for fronds within one species of tree fern (Cibotium glaucum; 3). The strength of size-scaling relationships is notable especially given that it applies across the very diverse species of the Hawaiian fern flora. Mechanisms for this pattern of scaling within a species and across an entire fern flora include genetic linkages and/or developmental constraints, which may result in a bauplan constraint on variation in gross morphology (e.g., the proportionality of organ dimensions). One genetic mechanism might be pleiotropy for size traits such that selection for increased plant size would affect all parts equally and lead to linear allometries (6), similar to ontogenetic constraints when matched relative growth rates of organs cause geometric scaling of their dimensions across a range of plant sizes (71; 41). Notably, pinnae number appeared independent of plant and frond dimensions in our study and in a previous study of Cyathea caracasana tree ferns across light gradients where pinnae number was unrelated to stipe length (4). As expected, frond dissectedness increased with plant size. This pattern may arise as pinnae become increasingly separated with increasing frond size as space becomes available for the addition of higher-order divisions. Further, there may be a functional advantage for larger plants with greater leaf sizes that avoid high temperatures by dividing the leaf into smaller parts to increase convective cooling (95, 80). However, producing divided fronds may entail greater mass allocation to rachis support tissue instead of photosynthetic tissue (65).
Environmental associations of Hawaiian fern form
Our study is the first to our knowledge to show a linkage of frond form with light and water habitat and elevation independently of plant size. These patterns appear fundamental, because they are consistent with trends shown in other lineages of vascular plants and mosses, which evolved leaves independently (35, 36; 22; 101). Dry environments and open environments with higher irradiance appear to have selected or assembled taxa with smaller fronds and fewer and shorter pinnae because a thinner boundary layer allows greater heat dissipation (33; 81). Additionally, small leaves may possess a hydraulic advantage for water transport from veins to mesophyll, all else being equal (62). The tendency for longer blades in shaded environments would facilitate light capture relative to mass investment (35).
The variation among taxa in frond traits was also associated with elevation. Our analysis showed a greater abundance of alien taxa with simple leaves at low elevation. The positive association of frond dividedness with elevation may additionally be influenced by water availability because higher elevation taxa tended to occur in wetter sites and wet-site taxa had more divided fronds. This finding contrasted with the trend toward greater representation of simple leaves at high elevation in Bolivia and Borneo (51) and suggests a potential for distinct but functionally equivalent types of adaptation to high elevation. By reducing overall leaf size, simple leaves may reduce their boundary layer, like dissected leaves, to prevent overheating under high irradiance. Across floras, differences in the relationship between leaf dividedness and elevation may be driven by unique phylogenetic history and assembly processes. We did not find support for greater frond pubescence in high elevation ferns, though such a mechanism to reduce photodamage and heat load under higher irradiance has been shown in angiosperms (e.g., 25). Notably, a lack of pubescence at higher elevation was also found in ferns of the Bolivian Andes where greater precipitation at high elevation was believed to ameliorate the need for greater reflectance (53). A strength of this flora-wide analysis is its canonical perspective on trait variation across environments, showing general trends by aggregating communities with different species that are assembled across many local resource and climate gradients. There is a clear need in future studies to consider specific elevation gradients and how traits are optimized to the particular ways that elevation and light and water availability covary along those gradients, to improve understanding of variation in trait patterns within and across floras.
Trait associations with growth form
Plant size and morphological heterogeneity were strongly associated with growth form. Epiphytic and epipetric taxa were on average smaller, with smaller component parts than terrestrial ferns that were larger and had more divided fronds. Epiphytes exhibited simple leaves with lamina pubescence but often lacked stipe scales and were more likely to occur in shaded sites and at lower elevations. This finding suggests that scales may help to protect developing crosiers and stave off climbing plant competitors (68, p. 2002). In taxa from exposed habitats, scales may protect the photosynthetic apparatus from photodamage during desiccation (27); however, we did not find a trend for greater scale representation in dry or open habitats, or among epipetric taxa. As found for angiosperms, the trend for greater lamina pubescence in epiphytes may reflect allocation to mechanical and herbivore defense (39) or to longer lifespan in low resource environments (46; 17; 45), given that fern epiphytes persist in water and nutrient-limited aerial habitats (12; 103). Pubescence might also reduce transpiration rates through effects on boundary layer and/or by reflecting radiation, improving water-use efficiency (103).
Morphologies conducive to greater abundance and distribution across islands and habitats
With the ability to proliferate and disperse very long distances via spores, ferns may be limited in their abundance and distribution by environmental filters on acceptable forms and physiologies in given habitats (92, 93; 47). Across vascular plant floras, larger plants have been found to be more competitive and to have greater abundances and wider distributions (11; 74; 73). Notably, our analysis indicated that across Hawaiian ferns, abundance and distribution were independent of plant size, whether considered across all taxa or within the terrestrial, epiphytic, and epipetric growth forms. Additionally, abundant taxa were distributed across more islands. One reason for this linkage may be that greater abundance provides greater dispersal source strength. This trend for Hawaiian ferns contrasts with the weak negative correlation of range size and local abundance found for pteridophytes along elevation gradients in Bolivia (50); however, in that study species abundances were reduced for wider ranging species sampled at their geographic range limits. The trend found here for Hawaiian ferns is consistent with the more commonly reported positive correlation of species’ abundance and range size for angiosperms and animals (42; 59; 9). Whether the trend in Hawaiian ferns arose due to preadapation and sorting—i.e., the simultaneous success of colonizers in dispersal and competition—or due to the evolution of such well-dispersing competitors in multiple colonizing lineages, requires further investigation when more phylogenetic information becomes available.
We expected that shade and wet site taxa would be most abundant because the literature has emphasized that many fern species are commonly associated with moist, shady environments (67; 1; 49). However, we found that open-site taxa were typically more abundant when aliens were included, but among native ferns considered alone, taxa of all habitats were similar in their mean abundance. Also in contrast to expectation, dry-site taxa were more widely distributed across the Hawaiian Islands. One possible explanation may be the greater availability of dry site habitat across the islands before human settlement (48). Biological mechanisms may also play a role, e.g., the presence of apogamy (development of embryo without fertilization) may facilitate wider establishment of xeric-adapted ferns by reducing reliance on external water availability for reproduction, and by producing, on average, faster rates of gametophyte and sporophyte growth and maturation (91; 63; 90).
Key differences in alien and native fern growth form, abundance and distribution
Aliens differed significantly from natives in their typical growth form, abundances, and distributions. Contrary to the expectation that aliens would be larger than natives, we found similar ranges in the sizes of plants and organs. Whereas native ferns exhibited a more even distribution of growth forms, 80% of alien fern species were terrestrial. This pattern may result from the more common cultivation of terrestrial ferns (106, 26; 75) or reflect a greater ability of terrestrial ferns to naturalize. Future studies are needed to test for whether given traits (such as growth form) correlate with rates of escape for cultivated ferns to better inform weed assessment programs (20).
Overall, given spore dispersal, we expected that alien ferns would also be widely distributed and abundant in Hawaii, in spite of their relatively recent introductions. Indeed, on average alien and native fern taxa occurred across a similar number of islands. However, whereas among native ferns those common on dry sites tended to have the widest distributions, among aliens, distribution was independent of habitat. We were surprised to find that aliens were more abundant on average than native ferns, due to 34 of 35 alien taxa being common, whereas nearly 25% of the 163 native taxa were rare, endangered, or extinct. Although we note that this trend may not account for some additional recently escaped alien species with low abundances, the finding of so many alien species more abundant than natives draws attention to an important phenomenon. As expected, ranges for alien taxa were centered at lower elevation than natives, and aliens were commonly found in open sites where concentrated human activity may have generated disturbed habitats preferred by alien ferns and lycophytes.
Success of hypotheses and future directions
The power of our hypotheses for trait associations was very high for explaining morphological variation in Hawaiian ferns (near 100% of hypotheses were successful in explaining differences in all cases except for island distribution, 57%). This finding demonstrates the strength of fundamental premises to explain key florawide patterns. Explanatory power was in our definition the success of the hypotheses to point the direction a significant difference was found. The predictive power is lower given its definition as the success of hypotheses to anticipate all differences correctly and in the right direction; thus, predictive power decreases when no significant differences are found. Still, the predictive power of our hypotheses was substantially greater than that expected from random chance (45% for size-scaling, 27–33% for growth form, 4–20% for environment, and 21–22% for demography). Despite these substantial proportions, that only a minority of our hypotheses were predictive points to the need for further research to elucidate additional factors determining variation in form and distribution. Additional trait relationships might have been supported given finer scale information of trait values and resource gradients and with data for all taxa, which would enable identification of trends with greater robustness, as well as interesting outliers to general trends. Studies of ferns in common garden and field conditions would also allow assessment of plastic responses to environmental variation and of the extent to which intraspecific plasticity might contribute to the observed trends. Further, we note that linking single traits with habitat and distributions does not account for multiple “optimal designs” for adaptation to particular habitats (61). Thus, different clusters of traits may convey functional equivalence and a similar ability to persist, disperse, and become abundant in given environments. As more data become available, future studies may consider the linkage among clustered traits with environment simultaneously.
The trait relationships elucidated here suggest potential adaptation or ecological sorting of preadapted taxa (38) and indicate that morphological traits would be important in resource capture and acquisition and in determining differences in plant fitness across environments. Still, we note that the traits available in the database, though valuable, are limited. Many other features of form and physiology are highly diverse across ferns and would be critical to determine performance and habitat specialization. For example, further work is needed to resolve how variation in nutrient composition and stoichiometry, leaf venation, and allocation patterns for structural support tissue influence water and gas flux rates, and consequently, growth and reproduction in ferns (8, 7; 2). Additionally, the identification of ancestral colonists will enable determination of the persistence of traits and ecology. For example, the three Hawaiian Polystichum species, all typically of high elevations, radiated from an Asian ancestor of high elevation (e.g., 21). Further resolution of phylogenetic relationships among fern taxa will also clarify the historical patterns that influenced trait evolution and current distributions (43), the degree to which variation evolved in situ, and the emergence of adaptive trait complexes (cf. 22). Future work on flora-scale databases should open up exciting avenues for such important investigations.
