Understanding species' ability to colonize new habitats is a key knowledge allowing us to predict species' survival in the changing landscapes. However, most studies exploring this topic observe distribution of species in landscapes which are under strong human influence being fragmented only recently and ignore the fact that the species distribution in these landscapes is far from equilibrium. Oceanic islands seem more appropriate systems for studying the relationship between species traits and its distribution as they are fragmented without human contribution and as they remained unchanged for a long evolutionary time. In our study we compared the values of dispersal as well as persistence traits among 18 species pairs from the Canary Islands differing in their distribution within the archipelago. The data were analyzed both with and without phylogenetic correction. The results demonstrate that no dispersal trait alone can explain the distribution of the species in the system. They, however, also suggest that species with better dispersal compared to their close relatives are better colonizers. Similarly, abundance of species in the archipelago seems to be an important predictor of species colonization ability only when comparing closely related species. This implies that analyses including phylogenetic correction may provide different insights than analyses without such a correction and both types of analyses should be combined to understand the importance of various plant traits for species colonization ability.
Citation: Vazačová K, Münzbergová Z (2014) The Importance of Species Traits for Species Distribution on Oceanic Islands. PLoS ONE9(7): e101046. https://doi.org/10.1371/journal.pone.0101046
Editor: Sylvain Delzon, INRA - University of Bordeaux, France
Received: February 14, 2014; Accepted: June 3, 2014; Published: July 8, 2014
Copyright: © 2014 Vazačová, Münzbergová. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This project was supported by GAČR P505/10/0593, GAUK 48807, Mobility Fund of the Charles University in Prague and partly by MSMT and RVO 67985939. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Species ability to disperse and colonize new habitats is a key prerequisite for their response to ongoing landscape and climate changes , . Understanding, which are the main traits responsible for this ability, is thus fundamental for prediction of future fates of different species , . Many recent studies are attempting to understand the importance of species traits for species ability to colonize habitats of different size and isolation (e.g. , ). Most of these studies are done in various fragmented landscapes, predominantly in grasslands and forests. Often these studies demonstrate that species distribution is not only determined by current landscape structure, but is largely a result of landscape structure in the past (e.g. , ).
Strong species response to past landscape structure can be attributed to slow growth dynamics of many perennial species in combination with relatively fast changes in the current landscapes , . Due to dispersal limitation ,  and extinction debt ,  the distribution of these species may reflect historical habitat configuration. Species distribution in the landscape may then not reflect species long-term ability to successfully colonize habitats and to survive there. Thus the traits driving species distribution on young habitat fragments in a changing landscape can be different from those in the landscapes fragmented for longer evolutionary time , .
Due to intensive human activity all over the world, it is rather difficult to identify fragmented habitats which remained unchanged for a long time period, on which we could study species ability to colonize new habitats on long time scales. Oceanic islands seem to be suitable candidates of such systems , . In contrast to continental landscape, oceanic fragments are not a result of human activity and remained almost unchanged in size and number since their origin. Thus the islands are generally thought to be more stable in time as they are fragmented and isolated for much longer time periods. For these reasons they are suitable systems for studying the importance of dispersal traits for species occurrence on isolated patches. Similarly to the studies on habitat fragments on the mainland (e.g. , ), we can predict that species occurring on the youngest and the most isolated islands will have higher dispersal ability than species present on older and more connected islands.
In this study, we analyzed species traits determining distribution of selected native species on the Canary Islands. The Canary archipelago is a suitable model system as it consists of islands differing in their age, size and isolation as well as in species composition. Specifically, we attempted to understand the determinants of species presence on the newest, smallest and most isolated island (El Hierro).
Because closely related species often share a wide range of biological traits, distribution of a species may be related to the traits under study or to other traits correlated with these traits that are characteristic for the whole clade to which the species belongs , . Comparison of results of analyses with and without phylogenetic correction can help in distinguishing between the traits that are really responsible for a pattern and traits correlated with these within larger species groups. The necessity of phylogenetic correction is a highly debated issue (e.g., , ) and it has been suggested that the phylogenetic and ecological explanations for species distribution in a landscape are not mutually exclusive (see also ). Separating the phylogenetic and ecological explanations for species distribution is thus difficult. It is, however, generally recognized that both of these types of analyzes should be considered when trying to explain the effect of species traits on species distribution (e.g. ).
To consider species phylogeny in this study, we compared dispersal values of 18 pairs of closely related species differing in their presence on El Hierro. In addition, we used the same species to test the relationship between species traits and number of occupied islands in the archipelago. For each species, we assessed the dispersal ability by all possible dispersal vectors acting on islands, i.e. wind, water and animals (anemo-, hydro-, exo- and endozoochory). We also used published sources complemented with our field experience to identify the most likely dispersal mode for each species pair.
Even though nowadays some parts of the islands are quiet heavily inhabited, we suppose that the main dispersal events happened before human's strong influence. Also none of the studied species is purely ruderal. All the species occur in some (semi)-natural habitats such as laurel forests and canary pine woodlands. Such communities obviously suffer from human destructive activities being fragmented and reduced in area, but species extinctions on single islands occur only rarely and were not reported for any of our model species .
Although dispersal ability is widely considered as a major determinant of species distribution on islands due to their isolation, other traits, especially those related to species persistence on habitats should not be overlooked as was shown in studies e.g. by Maurer et al.  and Saar et al. . For this reason we also tested traits related to species survival and persistence on the islands (i.e. species longevity and woodiness) and traits characterizing species distribution serving as a proxy for amount of seeds available (number of vegetation zones and number of islands occupied by a species). As a number of occupied islands itself can be a function of plant traits, we also explored the life history traits associated with number of occupied islands.
Specifically, we asked the following questions: 1) Which life history traits explain species presence on El Hierro? 2) Which life history traits explain the number of islands occupied by the species? 3) How do the conclusions change when applying phylogenetic correction?
We predict that species occurring on El Hierro will have better dispersal ability and will occupy more islands than species not occurring there. We also expect that species occupying more islands will be more likely r-strategists possessing traits, which enable rapid colonization of free space on islands (i.e. non-woody annuals occupying more vegetation zones).
Materials and Methods
To test exozoochorous dispersal, we used a pigeon of the King breed, purchased from a local breeder. To minimize subjection to stress during the experiment, the animal was caged in its home aviary (2×1.5×1 m) and had free access to commercial diet and water. The bird was not subjected to any invasive intervention which could cause him suffering. As he was tamed since his youth, his manipulation during seed incorporation into feathers did not cause him extreme stress. The manipulation with pigeon was approved by Ministry of Education, youth and sport of the Czech Republic (permission no. 24773/2008-10001) and complied with the relevant legislation of the Czech Republic (article 11, regulation no. 207/2004).
The Canary Islands are part of the Macaronesian archipelago situated between 27°45′ and 29°2′N and between 18°00′ and 13°37′W. They consist of 7 main volcanic islands differing in age and size (Figure 1). The age of the islands decreases with increasing distance from the closest mainland (Africa) and from east to west; the easternmost islands are the oldest, while the westernmost are the youngest. Vegetation composition and habitat diversity on islands is highly influenced by altitudinal gradients in combination with predominant north-eastern trade winds . The oldest and most eroded islands Lanzarote and Fuerteventura lack forests, other, steeper and roughed islands (Gran Canaria, Tenerife, La Palma, La Gomera, El Hierro) are covered by thermo-sclerophyllous woodlands, evergreen laurel forests and canary pine woodlands. The highest parts of Tenerife and La Palma host meso-oromediterranean summit broom scrubs .
We selected 36 species belonging to 22 genera and 15 families, all native to the Canary Islands . The species were grouped into pairs (Table 1). The species within the pair usually belong to the same genera. In three pairs, the two species in the pair represented closely related genera from the same family. Within each pair, the species differed in occurrence on El Hierro, on the youngest Canary Island, but they both were present on the adjacent islands (at least on Tenerife and La Palma or Tenerife and La Gomera). We chose Tenerife as it is considered as a centre of biodiversity of the area and thus can play a key role as a source for species dispersal to the westernmost islands (, but see ). La Gomera and La Palma were chosen because of their relative proximity to El Hierro and due to their similar size. All the three islands are also similar to El Hierro in the main vegetation zones including Euphorbia scrubs, thermo-sclerophyllous woodlands, evergreen laurel forests and canary pine woodlands. Due to these similarities we can suppose that species present on Tenerife and La Palma or Tenerife and La Gomera and not on El Hierro are those which have not been able to reach El Hierro due to dispersal limitation and not due to ecological barriers related to the absence of habitat .
We are aware that species presence/absence on El Hierro could be potentially mediated also by human activities. However, this island is less inhabited than the other Canary Islands. While some of the selected genera may occur in ruderal habitats (e.g. the genus of Reseda, Senecio, Trifolium), all of these occur also in some (semi)-natural habitats such as laurel forests and canary pine woodlands and could thus be distributed on the islands prior to increased human activities. We thus suppose that the main dispersal events happened before human's strong influence.
Species selection was further limited to species for which sufficient seed samples could be obtained. For this reason we had to exclude all the previously considered species pairs having fleshy fruits.
Diaspores (fruits or seeds representing the most probable dispersal units, see Table 1) for each species were collected in natural populations on the islands except for Limonium species. Diaspores of Limonium were obtained from the populations in the Botanical Garden “Jardín Canario Viera y Clavijo”, Gran Canaria. The garden populations originally come from the island populations.
The collection from protected areas was done in cooperation with the Botanical Garden “Jardín Canario Viera y Clavijo”, Gran Canaria which obtained appropriate permission for collecting seeds for scientific purposes. The permission was issued by Consejería de Medio Ambiente y Aguas, Islas Canarias. The permission for seed collection from unprotected areas was not required.
In the field we preferably sampled 3 populations per species. For each population, we aimed to collect diaspores from at least 8 individuals. Each population was then tested for dispersal abilities separately. Garden collection was considered as one population and we sampled seeds from 8 individuals in the garden. To have the same number of measurements for the species with seeds collected from the field and from the garden, we had 3 replicates for each dispersal experiment for diaspores collected in the garden.
We used 20 diaspores per species and population for experiments with anemochory, hydrochory and exozoochory and 30 diaspores for testing endozoochory, i.e. 60 and 90 diaspores, respectively. Such number was a compromise between a large amount of species tested and number of seeds used in the literature (c.f. ).
For testing other traits related to dispersal (i.e. seed mass and seed viability) we used simple seeds, not fruits. In dispersal modes, where we used fruits as dispersal units, but accounted also for seed viability (i.e. hydrochory and endozoochory), the number of all seeds extracted from the fruits was used as a baseline number of seeds.
Traits related to dispersal
The ability of diaspores to disperse by wind was estimated as terminal velocity defined as the maximum rate of seed falling in still air . It was measured as the flight time of a diaspore from predefined height (270 cm ). Mean dispersal distance D was expressed as:where w is the wind speed (being constant for all species), h is the average plant height and t is the terminal velocity. Values of average plant height were obtained from the literature , , , , 
We are aware that our dispersal model is simplified. Nevertheless, it has been successfully used in other studies to characterize mean dispersal distance of diaspores (e.g. , ) and is the easiest way to combine the three key variables affecting wind dispersal. We thus suggest that it is a useful proxy of potential wind dispersal distances for comparison among species.
In the analyses, we used both terminal velocity (m/s) and mean dispersal distance (m). In addition, we tested for the difference in plant height between species present on El Hierro and species absent from El Hierro to see to what extent are the differences in dispersal distance affected by differences in plant height.
The potential of diaspores to disperse in salt water (buoyancy) was measured as the proportion of diasporess still floating after a defined time period. Diaspores were gently put into beakers filled with salt water having 3.7% salinity (i.e. average salinity of the Atlantic Ocean along the Canary Islands coast). The size of beakers was proportional to the size of diaspores. Sea waves were simulated by continual shaking in electric orbital shaker with frequency of 100 shakes per min. The number of diaspores floating on water surface was checked immediately after putting them into bins and then after 5 minutes of shaking, 1, 2, 6, 24 hours and 7 days of shaking . The experiment was finished after 1 week of diaspores shaking as it is the minimal time a diaspore needs for reaching the Canary islands from mainland when taking into account average speed of water currents in the Atlantic Ocean (60–90 km per week ) and the distance between mainland and the closest island (Africa to Fuerteventura, 96 km).
At the end of the experiment, the number of floating and number of sunk diaspores was counted and the two groups of diaspores were then tested for seed viability.
In the analyses, we used the proportion of viable seeds which kept floating until the end of the experiment from the total number of viable seeds before the experiment.
The diaspore buoyancy was also expressed as T50, the number of minutes, after which 50 percent of diaspores was still floating. This parameter is commonly used in other studies assessing hydrochory ,  however it does not take into account seed viability.
We also used the information on effect of salt water on viability of seeds expressed as the proportion of viable seeds after the experiment (both floating and sunk)/seed viability before the experiment. Viability of seeds was tested by dying the seeds with 0.1% solution of 2,3,5-triphenyl-2H-tetrazolium chloride . In contrast to germination tests, it is not dependent on selection of the right conditions for germination for each individual species and it is thus in fact more reliable for between species comparisons.
Birds are the most important long-distance island dispersers transporting diaspores both externally and internally. The main bird dispersers acting on the Canary Islands are blackbirds (Turdus merula), robins (Erithacus rubecula), blackcaps (Sylvia atricapilla and S. melanocarpa, ), common ravens (Corvus corax, ), gulls (Larus cachinnans, ) and pigeons (Columba livia, C. junoniae and C. bolli).
Bird exozoochory (Epizoochory).
Bird exozoochory was tested as diaspore adhesion to bird feathers. As a model species we used a pigeon of the King breed, a utility breed with poor flight ability that is amenable to our experiments.
Although this species is clearly not native to Canary Islands, the functionality of its feathers for diaspore dispersal is readily comparable with native insular pigeon species.
As the seed coat of some species (e.g. Plantago, Arabis) contains mucilaginous substances which become sticky when wet, all the diaspores were moistened before the application into pigeon feathers. Moistened diaspores were gently incorporated into feathers on 4 different body parts (on bust, neck and back, under wing). After 1 hour of pigeon free movement in an aviary we checked the numbers of diaspores still attached to feathers. Taking into account the average flight speed of a trained pigeon (80 km/h ) and the shortest distance between mainland and the closest island (96 km), diaspores which remained attached to feathers after 1 hour are potentially able to get to the islands by this type of dispersal.
In the analyses we tested the proportion of diaspores which kept attached to feathers after 1 hour (we refer to this value as seed adhesion). This parameter lacks the effect of real bird flight as we do not take into account the air movement around feathers during the flight that can dry out diaspores and cause them to drop earlier than in our simulation. However some behavior of our pigeon during seed testing such as cleaning of feathers was similar to behavior of wild birds. Thus, we still think that our data are sufficient for the purpose to differentiate among diaspores with different ability to disperse by exozoochory.
Bird endozoochory was tested by simulating diaspore gut passage through pigeon digestive tract. Plastic flasks filled up with diaspores were shaken with wet grit (small stones eaten by birds to enhance digestion, commercial mixture for pigeons) for 24 hours in electric orbital shaker (200 shakes per minute ). Then diaspores were separated from the grit, rinsed and immersed in 5 ml of 1 M H2SO4 (pH≈0.3 ) for 4 hours. Intact seeds were retrieved, counted and tested for viability. The proportion of number of viable seeds which survived the simulation to the number of seeds viability before the experiment was used in the analysis. Seed viability after simulation was tested as described above.
Altogether 90 seeds per species were weighted. For this purpose, they were divided to groups by 10 to 30 seeds per group (10 seeds in the group for the largest and 30 for the smallest seeds, to get reasonable size estimates given by the precision of the balance, 0.0001 g). Seed mass is generally recognized as a rough proxy of seed dispersal ability and germination ability (e.g. , ). The same amount of seeds was used for viability testing of intact seeds.
Most likely dispersal mode
For all species pairs the most likely dispersal mode was estimated from available literature (Table 1). Where such data were missing, we estimated the dispersal mode according to our experience with dispersal and diaspore morphology of the species.
Traits related to persistence and distribution
Data on species longevity (short-lived vs. perennial), woodiness (woody vs. not woody) and the number of vegetation zones with species occurrence were gained from Bramwell and Bramwell  and Schönfelder and Schönfelder , .
Species distribution was expressed as a number of occupied islands, according to Arechavaleta et al. .
To test the importance of life history traits for species presence on El Hierro, we used a generalized linear model with binomial distribution. Species category (present on El Hierro vs. absent from El Hierro) was used as dependent variable and species traits as independent variables. In this analysis the number of islands occupied by a species was counted excluding El Hierro as the effect of El Hierro is already included in the dependent variable.
The importance of traits for species distribution among islands was tested by log-normal regression. Number of islands occupied by a species was used as dependent variable and species traits as independent variables.
The analyses were also performed with phylogenetic correction. Because the exact phylogenetic relationships between the studied species are unknown, we used the simplest version of phylogenetic correction based on comparison of species within the pairs (e.g. ). The corrected trait values PC were calculated by applying the formula:where S is the trait value of a single species (either present or absent from El Hierro) and MP is the mean of the trait value for each species pair. The phylogenetically corrected trait values were used in the tests as described above.
All the tests were done using two different approaches. First, we tested the effect of each trait separately. Afterwards, we combined all the traits in a single model and used forward step wise regression to select an optimal model.
To visualize the similarity between different species in their traits we used principal component analysis (PCA). The data on single species traits were treated as “species”, and data on each species represented “samples.” The analysis was centered and standardized by “species”; in this way all the traits were expressed in the same, relative, units.
Species present on El Hierro and species absent from El Hierro did not differ in any studied dispersal traits. There was, however, marginally significant effect on number of occupied islands (without El Hierro) (Table 2) with species present on El Hierro occupying more islands than species absent from El Hierro (Figure 2A). The results changed dramatically after incorporating phylogenetic correction into analyses. After phylogenetic correction, species presence on El Hierro was significantly influenced by dispersal distance, seed mass, species longevity and by the number of islands occupied by the species (Table 2). Species present on El Hierro dispersed further by wind, had smaller seeds, shorter life-span and occupied more islands than species absent from El Hierro (Figure 2B–D).
(PC, A, p = 0.062) and with PC (B, p<0.001), dispersal distance with PC (C, p = 0.05) and seed mass with PC (D, p = 0.007).
Number of islands occupied by a species was significantly influenced only by species longevity (Table 3). Species occupying more islands were more likely annuals than species occupying fewer islands. This trend remained the same even after phylogenetic correction. All the significant variables also remained in the model after stepwise regression showing that the traits are largely independent of each other (Tables 2 and 3).
Principal component analysis of dispersal traits showed that species within a pair are rather dissimilar in their traits (Figure 3). As seen in Figure 3, species are partly grouped according to the most likely dispersal mode.
The first axis explained 27.7% of variability, the second axis explained 26.1%. Different symbols indicate species most likely dispersal modes (according to literature): species with solid black circles are most likely dispersed by endozoochory, species with solid grey circles are dispersed by hydrochory, species with opened symbols are dispersed by anemochory and species with solid black triangles are dispersed by exozoochory. Species pairs are connected by lines.
The results of the study indicated that species presence on El Hierro, the smallest, youngest and the most remote island is influenced by both dispersal and persistence traits as well as by the number of other Canary Islands occupied by a species. This result was, however, found only after applying phylogenetic corrections. This suggests that the advantage of these traits is relative, and the traits thus play a role only after accounting for other possible differences between closely related species.
Contrasting results with and without phylogenetic correction were found previously also by e.g. Tremlová and Münzbergová  for dispersal traits, by Lanta et al.  for traits related to plant growth and by Stratton  for flower longevity pointing out the necessity for considering phylogenetic information in the analyses. The strong discrepancy between the two types of results is related to the stability of these traits within species phylogenies (e.g. , , ). The results obtained in this study should thus be interpreted not as the main effects of the given dispersal mode. In contrast, they e.g. suggest that within a given species group (sharing a wide range of biological traits) the species with relatively better dispersal are better colonizers.
Our expectation that species present on El Hierro disperse better than species absent from El Hierro holds only for wind dispersal mode. The importance of anemochory in dispersal among oceanic islands has been mentioned in classical islands studies , . Regarding the Canary archipelago, seed transport from the eastern to the western islands (including El Hierro) can be mediated by northeasterly trade winds (which blew during arid Quaternary episodes ) as it was reported by e.g. Percy and Cronk  or Allan et al.. However, when estimating dispersal distance using simply the data on terminal velocity, plant height and mean wind speed on islands (6.55 m/s ) and the nearest distance from El Hierro to neighboring island (La Gomera, 50 km), no species would be able to reach the island by wind. While such simple dispersal model is commonly used to approximate wind dispersal ability of species, such a model is rather simplified . To estimate realistic dispersal distances of species we need to know also other parameters related to wind activity (mainly turbulence and updrafts) and island topography. Considering these types of data in the model is, however, beyond the scope of this study. Another indirect evidence for the importance of wind as an important dispersal mechanism on islands is that species present on El Hierro have smaller seeds (and thus more suitable for flying in the air) than species not present there. Generally, according to Lindborg et al. , species with smaller seeds are better dispersers, whereas those with large seeds are better recruiters and tend to have improved establishment in a wider range of habitats ,  or when competing with neighbors (see ). However, the good competitive ability is not necessarily important for habitats on young volcanic islands arising de novo such as El Hierro. Additionally, the vegetation on El Hierro was repeatedly disturbed by volcanic activities causing extensive landslides further favoring good colonizers over good competitors.
No significant differences in other dispersal traits between species differing in the presence on El Hierro can signify that these species do not disperse by the tested dispersal modes in reality. For this reason we also tested the most likely dispersal mode, which was based on the selection of the most likely dispersal mode within species pair according to the literature. However, using the most-likely dispersal mode did not show significant differences between species present on El Hierro and absent from El Hierro. The use of such type of dispersal information from a variety of literary sources based on heterogeneous methodology for determining the most likely dispersal mode is questionable, but frequently practiced , . As a result, the most likely dispersal mode differs according to different authors for some species (e.g. for Euphorbia hydrochory in Wald et al. 2005  and endozoochory in Carlquist 1967 ). However, even after changing the most likely dispersal mode of some species there were no significant differences between species present and absent on El Hierro in their dispersal ability. Moreover, the only significant wind dispersal in our study was the most frequently chosen most likely dispersal mode. This suggests that the selection of the most likely dispersal mode is not so far from the reality.
According to our results, species present on El Hierro are distributed on more islands (excluding El Hierro) than species absent from El Hierro. This could be due to better wind dispersal ability of species on El Hierro. However, no dispersal trait significantly predicted number of islands occupied by the species. This suggests that better dispersal ability is not generally related to distribution on more islands as we could suppose. No relationship between dispersal and range size was shown e.g. by Kelly and Woodward , Goodwin et al.  and Lester et al. . Lester et al.  assumed that dispersal may only influence species' geographical distributions at certain spatial scales or in particular habitats or environment and/or within certain taxonomic groups, depending on how the mechanisms by which dispersal and range size are related.
The reason why the number of occupied islands is a good predictor of species' presence on El Hierro could be that the number of occupied islands represents a measure of the amount of available sources (i.e. a proxy of number of plant populations or species abundance) for species' colonization (e.g. ). Indeed, to properly measure the amount of available sources we should also know the species local abundances and seed production. Obtaining good information on these two characteristics is, however, rather complicated and such data are not available. Alternatively, number of occupied islands could also be linked to niche width as species with wider distribution range tend to have wider niches and thus more likely occupy a novel habitat (Knappová unpubl.).
Species longevity was another trait influencing species presence on El Hierro.
Species on El Hierro were mainly short-lived (annuals and biennials) showing that short life span enabling rapid production of offspring can be an advantage for colonizing this westernmost island. Due to their ruderal strategy, short-lived species are usually able to grow on newly emerged or disturbed habitats indicating that the island vegetation is still developing. According to Kelly  and Kelly and Woodward  short-lived species are expected to have smaller ranges than perennials, which is in contrast to our results. We showed the opposite pattern; short-lived species have wider distribution among islands.
There are other possible traits such as seed bank longevity, seed production, pollination mode or detailed characteristics of species habitat requirements (e.g. in the form of indicator values) or species local abundance, which can influence species distribution as was shown e.g. by Pocock  and Gabrielová et al. . These studies are mainly done on European species, where most of these data are available as a part of databases , . No such complete data is, however, available for the whole flora of the Canary Islands.
Possible limitations of the study
Despite the above arguments explaining limited role of dispersal traits in species distribution we cannot exclude the possibility that the importance of species dispersal is undervalued due to our species selection, especially by excluding species with fleshy fruits. We excluded species with fleshy fruits primarily for practical reasons as we were not able to collect sufficient number of fruits due to scarcity and the protection status of some of the potential species (e.g Sambucus palmensis, Pleiomeris canariensis, Heberdenia excelsa). However, as our species list involves mainly anemo- and exozoochorous species, addition of only few pairs of species with fleshy fruits would generate uneven distribution of dispersal modes resulting in few strong outliers. Such data could maybe lead to conclusion that dispersal is more important than we are suggesting based on the current results. On the other hand such a conclusion based on few outliers would not be very robust. We thus suggest that the limited species selection used in this study can also be viewed as an advantage as our study provides relatively robust conclusions for a wide range of anemo- and exozoochorous species.
Another possible critique of our study is that we are working with only 18 pairs of species. Species number was mainly limited due to the approach used to study dispersal, which was dependent on large number of seeds available for each species. Thanks to this approach, we were, however, able to obtain really detailed information on species dispersal by the main dispersal vectors acting among islands. In contrast, other dispersal studies dealing with more species are often based only on categorization of dispersal abilities inferred from the combination of seed visual observation and field experience ,  or assessing dispersal by one dispersal mode only (). Such approach enables to cover larger number of species, but species traits are only roughly assessed. As a result, the insights obtained in these studies are more general on one hand, but very rough on the other, not allowing to explore the importance of smaller differences in dispersal ability between different species. We suggest that the results obtained in our study are more likely to indicate possible long-term fates of species in fragmented systems within sets of species of similar growth forms dispersing in similar ways.
The results demonstrated that the relationship between species distribution and species traits depends on the approach we use. Different results were obtained after incorporating phylogenetic relationship between species than when such correction was not used. Thus we suggest to combine both approaches when analyzing closely related species to understand the importance of various plant traits for species distribution.
Values of dispersal traits of 18 species pairs used in the study (the first mentioned is species absent from El Hierro).
Values of persistence traits, traits related to distribution and other traits of 18 species pairs used in the study (the first mentioned is species absent from El Hierro).
We thank Dr. D. Bramwell, the director of the Botanical Garden “Jardín Canario Viera y Clavijo” and Dr. J. Caujapé Castells, the Head of Molecular Biodiversity Labs and DNA Bank for enabling collection of seeds from the Botanical Garden and the possibility to participate in seed collection from the field. We also thank other people from these institutes, namely R. Jaén Molina, M. Soto Medina and M. Olangua-Corral for all the kind help. Many thanks belong also to the staff of the Institute of Botany ASCR in Průhonice for helping with all the experiments and William K. Morris and two anonymous reviewers for useful comments to the manuscript.
Conceived and designed the experiments: KV ZM. Performed the experiments: KV. Analyzed the data: KV ZM. Wrote the paper: KV ZM.
- 1. Schwartz MW, Iverson LR, Prasad AM, Matthews SN, O'Connor RJ (2006) Predicting extinctions as a result of climate change. Ecology 87: 1611–1615.
- 2. Gallagher RV, Hughes L, Leishman MR (2013) Species loss and gain in communities under future climate change: consequences for functional diversity. Ecography 36: 531–540.
- 3. Marini L, Bruun HH, Heikkinen RK, Helm A, Honnay O, et al. (2012) Traits related to species persistence and dispersal explain changes in plant communities subjected to habitat loss. Divers Distrib 18: 898–908.
- 4. Stevens VM, Trochet A, Blanchet S, Moulherat S, Clobert J, et al. (2013) Dispersal syndromes and the use of life-histories to predict dispersal. Evol Appl 6: 630–642.
- 5. Murray BR, Thrall PH, Gill AM, Nicotra AB (2002) How plant life-history and ecological traits relate to species rarity and commonness at varying spatial scales. Austral Ecol 27: 291–310.
- 6. Tremlová K, Münzbergová Z (2007) Importance of species traits for species distribution in fragmented landscapes. Ecology 88: 965–977.
- 7. Chýlová T, Münzbergová Z (2008) Past land use co-determines the present distribution of dry grassland plant species. Preslia 80: 183–198.
- 8. Saar L, Takkis K, Partel M, Helm A (2012) Which plant traits predict species loss in calcareous grasslands with extinction debt? Divers Distrib 18: 808–817.
- 9. Lindborg R (2007) Evaluating the distribution of plant life-history traits in relation to current and historical landscape configurations. J Ecol 95: 555–564.
- 10. Purschke O, Sykes MT, Reitalu T, Poschlod P, Prentice HC (2012) Linking landscape history and dispersal traits in grassland plant communities. Oecologia 168: 773–783.
- 11. Herben T, Münzbergová Z, Mildén M, Ehrlén J, Cousins SAO, et al. (2006) Long-term spatial dynamics of Succisa pratensis in a changing rural landscape: linking dynamical modelling with historical maps. J Ecol 94: 131–143.
- 12. Münzbergová Z, Cousins SAO, Herben T, Plačková I, Mildén M, et al. (2013) Historical habitat connectivity affects current genetic structure in a grassland species. Plant Biol 15: 195–202.
- 13. Piqueray J, Cristofoli S, Bisteau E, Palm R, Mahy G (2011) Testing coexistence of extinction debt and colonization credit in fragmented calcareous grasslands with complex historical dynamics. Landscape Ecol 26: 823–836.
- 14. Hylander K, Ehrlén J (2013) The mechanisms causing extinction debts. Trends Ecol Evol 28: 341–346.
- 15. Parisod C, Bonvin G (2008) Fine-scale genetic structure and marginal processes in an expanding population of Biscutella laevigata L. (Brassicaceae). Heredity 101: 536–542.
- 16. Parisod C, Christin PA (2008) Genome-wide association to fine-scale ecological heterogeneity within a continuous population of Biscutella laevigata (Brassicaceae). New Phytol 178: 436–447.
- 17. Duarte MC, Rego F, Romeiras MM, Moreira I (2008) Plant species richness in the Cape Verde islands - eco-geographical determinants. Biodivers Conserv 17: 453–466.
- 18. Hansen DM, Traveset A (2012) An overview and introduction to the special issue on seed dispersal on islands. J Biogeogr 39: 1935–1937.
- 19. Darling E, Samis KE, Eckert CG (2008) Increased seed dispersal potential towards geographic range limits in a Pacific coast dune plant. New Phytol 178: 424–435.
- 20. Riba M, Mayol M, Giles BE, Ronce O, Imbert E, et al. (2009) Darwin's wind hypothesis: does it work for plant dispersal in fragmented habitats? New Phytol 183: 667–677.
- 21. Westoby M, Leishman M, Lord J (1995a) Further remarks on phylogenetic correction. J Ecol 83: 727–729.
- 22. Westoby M, Leishman M, Lord J (1995b) Issues of interpretation after relating comparative datasets to phylogeny. J Ecol 83: 892–893.
- 23. Pocock MJO, Hartley S, Telfer MG, Preston CD, Kunin WE (2006) Ecological correlates of range structure in rare and scarce British plants. J Ecol 94: 581–596.
- 24. Grime JP, Hodgson JG (1987) Botanical contributions to contemporary ecological theory. New Phytol 106: 283–295.
- 25. Lindborg R, Helm A, Bommarco R, Heikkinen RK, Kuhn I, et al. (2012) Effect of habitat area and isolation on plant trait distribution in European forests and grasslands. Ecography 35: 356–363.
- 26. Caujapé-Castells J, Tye A, Crawford DJ, Santos-Guerra A, Sakai A, et al. (2010) Conservation of oceanic island floras: present and future global challenges. Perspect Plant Ecol Evol Syst 12: 107–129.
- 27. Maurer K, Durka W, Stocklin J (2003) Frequency of plant species in remnants of calcareous grassland and their dispersal and persistence characteristics. Basic Appl Ecol 4: 307–316.
- 28. Fernandéz-Palacios JM, Andersson C (2000) Geographical determinants of the biological richness in the Macaronesian region. Acta Phytogeographica Suecica 85: 41–50.
- 29. del Arco Aguilar M-J, González-González R, Garzón-Machado V, Pizarro-Hernández B (2010) Actual and potential natural vegetation on the Canary Islands and its conservation status. Biodivers Conserv 19: 3089–3140.
- 30. Arechavaleta M, Rodríguez S, Zurita N, García A (2010) Lista de especies silvestres de Canarias. Hongos, plantas y animales terrestres. Gobierno de Canarias. 579 p.
- 31. Sanmartín I, Van der Mark P, Ronquist F (2008) Inferring dispersal: a Bayesian approach to phylogeny-based island biogeography, with special reference to the Canary Islands. J Biogeogr 35: 428–449.
- 32. Caujapé-Castells J (2011) Jesters, red queens, boomerangs and surfers: a molecular outlook on the diversity of the Canarian endemic flora. In: Bramwell D, Caujapé-Castells J, editors. The biology of island floras. Cambridge University Press London. pp. 284–324.
- 33. Knevel I, Bekker R, Kunzmann D, Stadler M, Thompson K (2005) The LEDA traitbase collecting and measuring standards of life history traits of the Northwest European flora. Bedum (The Netherlands): Scholma Druk B.V.
- 34. Thompson K (2005) Terminal velocity. In: Knevel I, Bekker R, Kunzmann D, Stadler M, and Thompson K, editors. The LEDA traitbase collecting and measuring standarts of life history traits of the Northwest European flora. Scholma Druk, B.V., Bedum (The Netherlands). pp. 122–124.
- 35. Münzbergová Z (2004) Effect of spatial scale on factors limiting species distributions in dry grassland fragments. J Ecol 92: 854–867.
- 36. Tutin T, Heywood V, Burges N, Moore D, Valentine D, et al.. (1964–1980) Flora Europaea. Cambridge University Press.
- 37. Castroviejo S, Laínz M, López González G, Montserrat P, Muñoz Garmendia F, et al.. (1986–2012) Flora Iberica, Plantas Vasculares de la Península Ibérica e Islas Baleares. Real Jardín Botánico, CSIC, Madrid.
- 38. Bramwell D, Bramwell Z (2001) Wild flowers of the Canary Islands.
- 39. Schönfelder I, Schönfelder P (2002) Kosmos Atlas Mittelmeer- und Kanarenflora. Kosmos, Stuttgart.
- 40. Schönfelder P, Schönfelder I (2002) Květena Kanárských ostrovů. Academia, Praha.
- 41. Römermann C, Tackenberg O, Poschlod P (2005) Buoyancy. In: Knevel I, Bekker R, Kunzmann D, Stadler M, Thompson K, editors. The LEDA traitbase collecting and measuring standards. Scholma Druk, B.V., Bedum (The Netherlands). pp. 124–127.
- 42. Zhou M, Paduan JD, Niiler PP (2000) Surface currents in the Canary Basin from drifter observations. J Geophys Res 105: 21893–21911.
- 43. Boedeltje G, Bakker JP, Bekker RM, Van Groenendael JM, Soesbergen M (2003) Plant dispersal in a lowland stream in relation to occurrence and three specific life-history traits of the species in the species pool. J Ecol 91: 855–866.
- 44. Cottrell H (1947) Tetrazolium salt as a seed germination indicator. Nature 159: 748.
- 45. Olesen J, Valido Amador A (2004) Lizards and birds as generalized pollinators and seed dispersers of island plants. Ecología insular = Island Ecology: recopilación de las ponencias presentadas en el Symposium de Ecología Insular. Asociación española de ecología terrestre, AEET. pp.229–249.
- 46. Nogales M, Hernández E, Valdés F (1999) Seed dispersal by common ravens Corvus corax among island habitats (Canarian Archipelago). Ecoscience 6: 56–61.
- 47. Nogales M, Medina FM, Quilis V, González-Rodríguez M (2001) Ecological and biogeographical implications of Yellow-Legged Gulls (Larus cachinnans Pallas) as seed dispersers of Rubia fruticosa Ait. (Rubiaceae) in the Canary Islands. J Biogeogr 28: 1137–1145.
- 48. Gessaman JA, Nagy KA (1988) Transmitter loads affect the flight speed and metabolism of homing pigeons. Condor: 662–668.
- 49. Vazačová K, Münzbergová Z (2013) Simulation of seed digestion by birds: How does it reflect the real passage through a pigeon's gut? Folia Geobot: 1–13.
- 50. Santamaría L, Charalambidou I, Figuerola J, Green AJ (2002) Effect of passage through duck gut on germination of fennel pondweed seeds. Arch Hydrobiol 156: 11–22.
- 51. Hewitt N, Kellman M (2002) Tree seed dispersal among forest fragments: II. Dispersal abilities and biogeographical controls. J Biogeogr 29: 351–363.
- 52. Statsoft I (2013) Statistica 12.0.
- 53. ter Braak C, Šmilauer P (2002) CANOCO reference manual and CanoDraw for Windows User's Guide: Software for canonical community ordination (version 4.5). Microcomputer Power Ithaca, NY.
- 54. S-plus 2000 Professional Edition for Windows. Release 2. MathSoft, Inc.
- 55. Lanta V, Klimešová J, Martincová K, Janeček Š, Doležal J, et al. (2011) A test of the explanatory power of plant functional traits on the individual and population levels. Perspect Plant Ecol 13: 189–199.
- 56. Stratton DA (1989) Longevity of individual flowers in a Costa Rican cloud forest - ecological correlates and phylogenetic constraints. Biotropica 21: 308–318.
- 57. Felsenstein J (1985) Phylogenies and the comparative method. Am Nat: 1–15.
- 58. Harvey PH, Pagel MD (1991) The comparative method in evolutionary biology. Oxford: Oxford University Press
- 59. Van der Veken S, Bellemare J, Verheyen K, Hermy M (2007) Life-history traits are correlated with geographical distribution patterns of western European forest herb species. J Biogeogr 34: 1723–1735.
- 60. Carlquist SJ (1965) Island life: A natural history of the islands of the world. New York: Natural History Press.
- 61. Bramwell D (1979) Plants and islands. London, New York: Academic Press.
- 62. Ortiz JE, Torres T, Yanes Y, Castillo C, De la Nuez J, et al. (2006) Climatic cycles inferred from the aminostratigraphy and aminochronology of Quaternary dunes and palaeosols from the eastern islands of the Canary Archipelago. J Quaternary Sci 21: 287–306.
- 63. Percy DM, Cronk QCB (2002) Different fates of island brooms: Contrasting evolution in Adenocarpus, Genista, and Teline (Genisteae, Fabaceae) in the Canary Islands and Madeira. Am J Bot 89: 854–864.
- 64. Allan G, Francisco-Ortega J, Santos-Guerra A, Boerner E, Zimmer EA (2004) Molecular phylogenetic evidence for the geographic origin and classification of Canary Island Lotus (Fabaceae: Loteae). Mol Phylogen Evol 32: 123–138.
- 65. Hill G (2003) Wind prospecting on the. Canary Islands. Department of Physical Geography. Goteborg.
- 66. Tackenberg O, Poschlod P, Bonn S (2003) Assessment of wind dispersal potential in plant species. Ecol Monogr 73: 191–205.
- 67. Aizen MA, Patterson WA (1990) Acorn size and geographical range in the North-American oaks (Quercus L.). J Biogeogr 17: 327–332.
- 68. Westoby M, Jurado E, Leishman M (1992) Comparative evolutionary ecology of seed size. Trends Ecol Evol 7: 368–372.
- 69. Weiher E, van der Werf A, Thompson K, Roderick M, Garnier E, et al. (1999) Challenging Theophrastus: A common core list of plant traits for functional ecology. J Veg Sci 10: 609–620.
- 70. Kelly CK, Woodward FI (1996) Ecological correlates of plant range size: Taxonomies and phylogenies in the study of plant commonness and rarity in Great Britain. Philos Trans R Soc Lond B Biol Sci 351: 1261–1269.
- 71. Wald EJ, Kronberg SL, Larson GE, Johnson WC (2005) Dispersal of leafy spurge (Euphorbia esula L.) seeds in the feces of wildlife. Am Midl Nat 154: 342–357.
- 72. Carlquist S (1967) The biota of long-distance dispersal. V. Plant dispersal to Pacific Islands. Bull Torrey Bot Club: 129–162.
- 73. Goodwin NB, Dulvy NK, Reynolds JD (2005) Macroecology of live-bearing in fishes: latitudinal and depth range comparisons with egg-laying relatives. Oikos 110: 209–218.
- 74. Lester SE, Ruttenberg BI, Gaines SD, Kinlan BP (2007) The relationship between dispersal ability and geographic range size. Ecol Lett 10: 745–758.
- 75. Knappová J, Hemrová L, Münzbergová Z (2012) Colonization of central European abandoned fields by dry grassland species depends on the species richness of the source habitats: a new approach for measuring habitat isolation. Landscape Ecol 27: 97–108.
- 76. Kelly CK (1996) Identifying plant functional types using floristic data bases: Ecological correlates of plant range size. J Veg Sci 7: 417–424.
- 77. Gabrielová J, Münzbergová Z, Tackenberg O, Chrtek J (2013) Can we distinguish plant species that are rare and endangered from other plants using their biological traits? Folia Geobot 48: 449–466.
- 78. Klotz S, Kühn I, Durka W (2002) BIOLFLOR - Eine Datenbank mit biologisch-ökologischen Merkmalen zur Flora von Deutschland. Schriftenreihe für Vegetationskunde
- 79. Van der Pijl L (1982) Principles of dispersal. Berlin: Springer-Verlag.
- 80. Vargas P, Nogales M, Jaramillo P, Olesen JM, Traveset A, et al. (2014) Plant colonization across the Galápagos Islands: success of the sea dispersal syndrome. Bot J Linn Soc 174: 349–358.
- 81. Couvreur M, Vandenberghe B, Verheyen K, Hermy M (2004) An experimental assessment of seed adhesivity on animal furs. Seed Sci Res 14: 147–162.
- 82. Bonet A, Pausas JG (2004) Species richness and cover along a 60-year chronosequence in old-fields of southeastern Spain. Plant Ecol 174: 257–270.
- 83. Malo J, Jiménez B, Suarez F (2000) Herbivore dunging and endozoochorous seed deposition in a Mediterranean dehesa. J Range Manage: 322–328.
- 84. Médail F, Quézel P (1999) The phytogeographical significance of SW Morocco compared to the Canary Islands. Plant Ecol 140: 221–244.
- 85. Bramwell D (1985) Contribución a la biogeografía de las Islas Canarias. Botan Macaron 14: 3–34.
- 86. Melendo M, Giménez E, Cano E, Mercado FG, Valle F (2003) The endemic flora in the south of the Iberian Peninsula: taxonomic composition, biological spectrum, pollination, reproductive mode and dispersal. Flora 198: 260–276.
- 87. Heinken T, Raudnitschka D (2002) Do wild ungulates contribute to the dispersal of vascular plants in central European forests by epizoochory? A case study in NE Germany. Forstwiss Centralbl 121: 179–194.
- 88. Sadeh A, Guterman H, Gersani M, Ovadia O (2009) Plastic bet-hedging in an amphicarpic annual: an integrated strategy under variable conditions. Evol Ecol 23: 373–388.
- 89. Evenari M, Kadouri A, Gutterman Y (1977) Eco-physiological investigations on amphicarpy of Emex spinosa (L.) Campd. Flora 166: 223–238.
- 90. Talavera M, Balao F, Casimiro-Soriguer R, Ortiz MA, Terrab A, et al. (2011) Molecular phylogeny and systematics of the highly polymorphic Rumex bucephalophorus complex (Polygonaceae). Mol Phylogen Evol 61: 659–670.
- 91. Koch MA, Kiefer C, Ehrich D, Vogel J, Brochmann C, et al. (2006) Three times out of Asia Minor: the phylogeography of Arabis alpina L. (Brassicaceae). Mol Ecol 15: 825–839.