SEARCH

SEARCH BY CITATION

Keywords:

  • Arctic;
  • migration;
  • dispersal;
  • dispersal distance;
  • colonization;
  • island biogeography;
  • scale dependence;
  • plants

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • The vascular plant flora of 66 arctic islands was studied to determine whether the islands have been occupied by random long-distance dispersal (LDD) or in a highly structured northward migration pattern via intervening islands as stepping-stones.
  • A maximum parsimonious migration model minimizing dispersal distances of 1256 vascular plant taxa was calculated in the framework of network analysis.
  • Plant dispersal is not stochastic in the Arctic at the global scale. Inferred mean dispersal distances of the plants occurring on arctic islands are c. 580 km (median 460 km). A LDD across the North Pole could not be inferred in the model and species may be recruited mainly from the nearest mainland or islands. At smaller scales, among adjacent islands, dispersal of vascular plants may be incomplete. Arctic islands do not yet appearto be saturated with species.
  • The results suggest that changes in biodiversity in Arctic islands can be more easily predicted at the global scale than at the local scale. Because islands are not yet saturated with species, new colonizations may not necessarily be linked to climate change.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Comprehension of migration patterns is compulsory for understanding the fate of ecosystems in the past and during future environmental changes. Migration is a salient feature shaping global plant biodiversity (Woodward, 1987; Bell, 2001; Hubbell, 2001). Steady-state biodiversity on islands depends in many cases on immigration and extinction of species as well as on the size of the islands (e.g. Rosenzweig, 1995). Particularly isolated and old islands tend to have high diversification rates of plants (Böhle et al., 1996; Baldwin & Sanderson, 1998), a pattern rarely observed in the Arctic (Hoffmann & Röser, 2009). Migration thus plays the most important role for shaping biodiversity patterns in the Arctic, because arctic islands were repeatedly and seriously affected by Pleistocene glaciations. During the glacial periods almost all plant species were lost from North Western Eurasia and parts of North America that were covered by ice sheets, but were recolonized in the interglacial periods (Brochmann et al., 2003). However, a few species may have survived the glaciations in ice-free habitats at high latitudes (Abbott & Brochmann, 2003; Westergaard et al., 2011). This pattern of survival appears to be rare. Predicted climate change may also have strong effects on the arctic island species composition (ACIA, 2006), owing to climatically available new space for plants and ongoing dispersal. Models of plant migration and inference of migration distances should help to comprehend the dynamics of this ecosystem assembly and its putative responses to predicted environmental changes.

Arctic islands provide a unique possibility to study plant migration patterns and distances because the islands are almost circularly enclosed by mainland belonging to the same ecosystem (Elven et al., 2005; Tkach et al., 2008). Open landscapes, storms, drift wood and ice, migrating animals and birds provide means that facilitate long-distance dispersal (LDD) of plants in the Arctic (Savile, 1972; Johansen & Hytteborn, 2001; Nathan et al., 2002). This might result in frequent LDD events and an almost stochastic distribution of the plants among the arctic islands. Conversely, the arctic island species may be recruited from founder populations of the closest continental area, leading to a highly structured northward migration pattern via islands used as stepping-stones, as has been observed for the Aleutian Islands (Woodward, 1987). However, several glacial cycles have disturbed early plant and animal distributions (Coope, 2004) and what we observe today is a snapshot in the dynamics of this ecosystem going back mostly to the time following the last ice age.

The aims of this study were to decide between a stochastic recolonization of arctic islands in postglacial times and a structured northward migration of plants starting from the nearest continents. For this purpose, a most parsimonious migration (MPM) model that minimizes dispersal distances was constructed and compared with random dispersal models in order to test if the arctic islands are saturated with species and to estimate migration distances of the arctic island species. For this purpose, plant inventories of 66 arctic islands as well as data on species occurrences on the arctic mainland were analysed and quantified. At the global scale, plant dispersal in the Arctic appears structured and nonrandom. Most species have shorter dispersal distances than expected by chance. At small scales, islands appear not yet to be saturated with plant species because of incomplete dispersal.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Sources of the distribution data of plant species on the 66 arctic islands are given in the Supporting Information, Table S1. The most comprehensive taxonomic source of the arctic flora is the Panarctic Flora checklist (PAF, Elven et al., 2005) that was used as taxonomic reference for the taxon list (species and subspecies). The PAF indicates presences in the arctic parts of Iceland, Norway, European Russia, Siberia, Russian Far East, Alaska, Canada and Greenland. Mainland and island distribution is not distinguished in the PAF except for Greenland and Iceland. The arctic mainland distribution in the above-mentioned sectors of PAF was inferred by using published dot maps for Alaska and circumpolar region (Hultén, 1968; Hultén & Fries, 1986), the Russian Arctic (Tolmachev, 1960–1987), Siberia (Krasnoborov et al., 1988–1997), the Russian Far East (Charkevicz, 1985–1996) and Arctic Canada (Porsild & Cody, 1980).

Plant dispersal does fit the theory of network analysis (Hanneman & Riddle, 2005), which consists of two elements: nodes (actors, islands, populations) and ties (relations, connections, that is, distances between islands and populations, respectively). Ideally, knowledge of the exact positions of the populations on the islands and continents should be used to infer plant dispersal by network analysis. These data are, unfortunately, not yet available for many islands, continental areas and plant species. Distances among the midpoints of islands are easy to infer but owing to the large size differences of the islands are not meaningful to apply in this case. For example, distances between the midpoint of Greenland and neighbouring islands of the Canadian Arctic Archipelago would be very large although the coasts are rather close. This would seriously distort a dispersal model among the islands. To circumvent this problem a coast to coast distance matrix among the islands and the mainland areas was constructed using the visual distance calculator of Microsoft Encarta World Atlas 2001. This matrix comprises the shortest distances among the shorelines of islands and mainland areas (Fig. 1). The distance between adjacent mainland areas was arbitrarily set to 1 km, indicating that species might freely migrate between them. A finer division of continental areas is currently not possible because the distribution of plants is still incompletely known in parts of the Arctic mainland. Implicit to this approach is the assumption that plants disperse more easily on land than across ocean barriers. Owing to the irregular shapes of the islands the points to other islands from where the shortest distances are measured are seldom the same and move around the island. This would not be reflected in the tie-relation data model of network analysis. Therefore, it was necessary to introduce a dichotomy in the data analysis between the model calculations and the graphical representation of the results. The modelling of plant dispersal among islands and the mainland areas uses the matrix of shortest distances between the areas (dashed lines in Fig. 1). The visualization of the model results uses island midpoints and an arbitrary point on the continents that is usually placed distantly from the coast for better visibility (bold straight lines in Fig. 1).

image

Figure 1. Graphical explanation of the network analysis for inferring the most parsimonious migration among islands and mainland areas (see the Materials and Methods section for a detailed explanation). The scheme shows the shortest coast to coast distances (dashed lines) among three islands and the mainland area. These distances were used for calculations of the maximal dispersal distance of the species and the colonization model of the arctic islands. For the graphical network presentation of the dispersal model (bold lines) roughly island midpoints and arbitrary points on the continents were used. This increases the visibility of the figure at the expense of nontrue scale presentation of dispersal distances. Judging from this scheme the simplification seems to be unnecessary, but with the real data the network would appear as a smear of lines. In the scheme a complete net-like dispersal among the four areas is shown, that is, each area is occupied by a species and may have been occupied from all areas. This would result from an inferred larger dispersal distance outside this schema (the arrow indicates this remote area). If only the four areas were occupied by a species, the longest coast to coast distances, for example, from mainland to island 2, between islands 1 and 3 as well as between islands 2 and 3, would not appear in the most parsimonious migration model because the shortest dispersals were from mainland to islands 1 and 3, respectively. For island 2 an occupation from island 1 is the most parsimonious to assume.

Download figure to PowerPoint

Calculations and modelling was performed using the software r (R Development Core Team, 2009) with the libraries MCMCpack (Martin et al., 2009) and sna (Butts, 2009). The MPM model of a taxon was obtained by selecting all coast-to-coast distances between the islands and mainland areas occupied by a taxon; ‘n’ is the number of islands and mainland areas occupied by a taxon. In a first step the n – 1 shortest distances were determined and selected from the intersection of the taxon’s occurrence matrix with the coast to coast matrix. The n – 1 shortest distances are theoretically sufficient for connecting the occurrences on islands and mainland areas into a linear graph. The resulting graph or graphs in the sense of network analysis (sna) were then checked for connectedness, that is, if all occurrences are connected or if two or more splitter graphs exist. If the graph is not connected the number of shortest distances is increased stepwise until the graph is connected, that is, each occurrence can be ‘reached’ from all other occurrences. This increasing of the shortest distances introduces reticulations to the graph and it turns from a linear to a network graph. In terms of dispersal biology, islands in the net-like part of the graph might then be occupied from two or more sources or can be sources of dispersal. The graphs of all taxa were then stacked to obtain a general picture of migration of arctic island plants (visualized using netdraw; Borgatti, 2002).

To test if the distribution pattern is obtained by chance alone, a randomization test was performed. The species’ occurrences on the continents were kept as observed and random dispersal was allowed to an equal number of islands as observed for each taxon. The dispersal distances of the MPM model and the random model were compared using a nonparametric Wilcoxon test.

The final distance needed for obtaining a connected graph among all occurrences of a given taxon is the taxon’s fundamental dispersal distance. For each taxon, the inferred individual fundamental dispersal distance was then used to test whether the taxon occupies all neighbouring islands within its fundamental dispersal distance. Starting from the islands known to be occupied by a taxon, all islands and mainland areas within the migration distance were selected. A correction for zonal distribution amplitudes of the species was applied for the following reasons. Northwards decreasing temperatures and thus detrimental environments for plant growth have led to the recognition of vegetation zones in the Arctic that act as environmental filters (Elven et al., 2005; Walker et al., 2005). For each species the zonal amplitude was determined to infer how far North it may grow. In the model each species was then not allowed to migrate to islands situated into colder zones of the Arctic than it is currently observed in. This approach reveals the expected species number on the islands if migration were complete. It also serves as an estimate for future development of the biodiversity if species saturation later becomes more complete.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

The arctic flora comprises c. 2775 taxa, that is, species and subspecies (Elven et al., 2005). Almost half of the species of the Arctic flora (45.3%, 1256 taxa) occur at least on one of the 66 arctic islands (minimum one island, maximum 60 islands: Saxifraga cernua) and were used in the model. An exemplary migration model of a species is given in Fig. 2. The MPM model of the arctic island flora is shown in Fig. 3. This summary picture shows a highly structured migration among the continents and arctic islands and rare LDD events. This pattern is significantly different from random models that keep the taxon occurrences on the continents constant and allow random dispersal to an equal number of islands, as observed for each taxon (Wilcoxon test, 100 replicate runs, all runs P << 0.001). The inferred migration distances are significantly shorter than the distances in the random dispersal models.

image

Figure 2. Most parsimonious migration model of Bistorta vivipara (Polygonaceae), one of the most widespread and abundant species of the Arctic. The dots indicate the islands as well as the six mainland areas as defined in the Panarctic Flora checklist (Elven et al., 2005). The lines are the possible migration routes of the species among islands and mainland, for which the species was reported to occur. The other points indicate islands that are not occupied by B. vivipara. For this species the graph is closed, for other species it may be open and not encircle the whole Arctic. The continental mainland areas, Greenland and Novaja Zemlja are represented by a single dot although they represent large areas. This point is not necessarily the starting point for the migration and placed partly outside the Arctic for better visibility. For example, the dot for Greenland is centred on this island but the plant distribution is confined to coastal areas. Greenland shares five possible migration routes with the Canadian Arctic Archipelago. Iceland and Jan Mayen may be occupied from Greenland. Svalbard may be occupied either from Greenland or from Europe via Bjornoya (Bear Island). It might even be possible that Greenland is occupied from Europe via Svalbard. The exact route is not discernable from the model. However, it is not parsimonious to assume, that, for example, Svalbard has been occupied from European Russia or Siberia by long-distance dispersal. For better visibility the island names are omitted in this figure. They are given in the summary picture in Fig. 3. CAN, Canada; ALA, Alaska; RFE, Russian Far East; SIB, Siberia; RUS, European Russia; EUR, arctic Northern Europe.

Download figure to PowerPoint

image

Figure 3. Most parsimonious migration model of the arctic island flora. The number of migrations among the islands and mainland areas are given as proportional tie sizes. The thinnest lines refer to singular migration events. The locations of the islands and mainland in the graphs are repositioned slightly for better visibility. Note that the mainland areas are large but are represented by dots on the map. The tie length between island and mainland is thus misleading in the figure. Calculations were performed with the shortest distances. Abbreviations of the arctic areas: CAN, Canada; ALA, Alaska; RFE, Russian Far East; SIB, Siberia; RUS, European Russia; EUR, arctic Northern Europe; ICE, Iceland; GRL, Greenland.

Download figure to PowerPoint

Most taxa migrated < 1000 km (mean 577 km, median 460 km, Fig. 4). The peaks at 770 km result from species that occur on the Canadian mainland and Greenland without occurrences in the Canadian Arctic Archipelago. The other peak at 980 km results from species that occur disjunctively in Iceland and the European mainland. These distances are within the ranges that were proposed in the literature (Savile, 1972). Only a few species have very disjunctive ranges that may result from LDD, for example, Blechnum spicant, Luzula arcuata and Carex livida.

image

Figure 4. Inferred migration distances of the 1256 taxa occurring on arctic islands.

Download figure to PowerPoint

If there were a random dispersal pattern in the Arctic, the inferred frequency distribution of the maximal migration distances (Fig. 4) would not be significantly different from the frequency distribution of the coast-to-coast distances (Wilcoxon test, W = 2 450 819, P << 0.001, mean and median of the coast-to-coast distances 2111 km and 2260 km, respectively). The significant differences point also to a nonrandom dispersal in the Arctic.

Using the species’ individual migration distances it may be inferred how completely islands are occupied in the vicinity of reported occurrences. The graph obtained is similar to the MPM model of the arctic island flora. However, species number (α diversity) on the islands should be generally higher in this model than observed (Fig. 5). The logarithm of the difference between expected and observed species number is not correlated with the logarithm of available ice-free area on the islands (Pearson correlation coefficient = 0.084, P = 0.504).

image

Figure 5. Observed species numbers on the 66 arctic islands and expected species number if migration was complete within the inferred migration distance and vegetation zones. The dashed line indicates equality.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Plant migration to and among arctic islands is not random. Very strong floristic similarities exist between the neighbouring arctic mainland areas, as may be inferred from the bold lines in Fig. 3. This is in accord with previous studies revealing a low species turnover along the northern rims of the continents (Tkach et al., 2008). In terms of plant migration this indicates that species migrate rather easily across these continental areas, depending on their ecophysiological capacity (Ackerly, 2003; Hoffmann, 2005). Arctic islands appear to be occupied mainly by dispersal from the nearest continent. For example, Wrangel Island shares most species with the Russian Far East, and Victoria Island with the arctic mainland of Canada (bold lines in Fig. 3). Because of the vicinity to the mainland, it is parsimonious to assume that the species of these islands were recruited from the mainland areas instead of invoking LDD (for a similar pattern of the Aleutian Islands, see Woodward, 1987). More remote islands, for example Svalbard, have various sources, that is, Greenland, Ellesmere, Europe, Kolguev, Novaja Zemlja or even Franz Josef Land. In this case, colonization of Svalbard directly from, for example, the Russian Far East, Alaska or mainland Canada is mostly not parsimonious to assume. Species shared among these areas have occurrences that are closer to Svalbard (e.g. Greenland and Novaja Zemlja), from which the archipelago may have been colonized; thus LDD is not a parsimonious assumption for the recruitment of species to such remote islands. The MPM model indicates that successful plant migration in the Arctic has never involved LDD across the North Pole.

Disjunctive occurrences of species result from LDD or range fragmentation. Range fragmentation may result in similar distribution patterns as LDD because of extinction of intervening populations. Plant LDD is a comparatively rare event over geological time-scales (Simpson, 1952; de Queiroz, 2005). With increasing distance from the sources the diaspore frequencies decline very rapidly (long tails of the dispersal curve, e.g. Clark et al., 1998) and new establishment of species becomes quite unpredictable, although possible (Simpson, 1952). For a limited number of arctic plant species, dispersal routes were inferred from molecular phylogeographic studies (see later). However, possible dispersal routes for the majority of arctic plants remain elusive. As a hypothesis for plant dispersal in the Arctic a maximum parsimonious migration (MPM) model can be constructed. In this model, known occurrences of the species across the Arctic were connected in a way that dispersal distances were minimized. The long tails of the dispersal curves support the idea that shortest distances among occurrences are preferable to explain the dispersal of plants. This parsimony analysis forces the graph not into a tree-like structure and reticulations, that is, ambiguities about dispersal routes, are allowed. Ad hoc explanations for particular distribution patterns are thus not necessary for this model. However, as in phylogenetics, where the maximum parsimonious tree does not necessarily reflect the true species phylogeny but the shortest number of evolutionary changes to obtain a tree, the MPM model does not necessarily reflect the true migration history of a species. If greater dispersal distances can be deduced for a species than can be inferred from the MPM model, these greater dispersal distances will probably introduce more reticulations of possible dispersal corridors in the MPM model. However, such ambiguities are unlikely to significantly affect the general picture of plant migration in the Arctic.

Over the past years, several phylogeographic studies of arctic plant species have been published that may be used to evaluate some of the dispersal corridors of the MPM model. The MPM model reveals a strong link between Greenland and Svalbard. This is in accord with migrations from Greenland to Svalbard, as inferred from phylogeographic studies in, for example, Cassiope tetragona (Alsos et al., 2007; Eidesen et al., 2007), Empetrum nigrum and Vaccinium uliginosum (Alsos et al., 2007). The occupation of Svalbard from Northern Russia by Betula nana, Dryas octopetala, Salix herbacea and V. uliginosum (Alsos et al., 2007) is revealed in phylogeographic studies and in the MPM model. Svalbard may also be directly colonized from Scandinavia (S. herbacea, Alsos et al., 2009). The links between Iceland, Greenland and Europe are reflected, for example, in Draba spp. (Skrede et al., 2009), Arabis alpina (Alsos et al., 2007), S. herbacea (Alsos et al., 2009) and Saxifraga spp. (Abbott et al., 2000; Alsos et al., 2007; Westergaard et al., 2010). The migration patterns of mainly Amphiatlantic plants, as revealed in the MPM model, were also in accord with phylogeographic studies of Arenaria humifusa and Sagina caespitosa (Westergaard et al., 2011). Distributions of plants that embrace the Arctic mostly in the boreal zone and from where arctic islands were colonized are rather frequent in the MPM model. Phylogeographic studies on some of these widespread species are in accord with the model, for example, V. uliginosum (Alsos et al., 2007). In general, many dispersal routes of the MPM model were matched with routes reconstructed from phylogeographic studies. Other dispersal corridors inferred from the model, for example among the Northern Siberian islands and the Siberian mainland, could not be evaluated, because molecular phylogeographic data are not yet available.

The islands of the Canadian Arctic Archipelago, of Svalbard or the islands close to the Taymyr Peninsula in Northern Siberia are highly connected, that is, the islands may be sources as well as sinks of plant migration. It might be argued that migration among these often small islands within the archipelagos is not very probable, because the small islands sustain only small populations in perhaps suboptimal conditions. In these marginal populations the species may produce only a small amount of diaspores and the sheer mass of diaspores produced by mainland populations would simply conceal the local migration among islands by LDD. However, this is not parsimonious to assume because LDD needs then to be assumed, leading to a narrative and almost untestable explanation of taxon recruitment of arctic islands. Thus, the nonstochastic MPM model of the arctic island plants suggests that there is a highly structured plant migration among the arctic islands at the global scale.

Having inferred the species’ individual migration distances it was possible to estimate how completely islands were occupied. The general migration pattern is essentially the same as for the observed data, but species number (α diversity) on the islands should be generally higher in this model than observed (Fig. 5). The implication of this observation is that although the arctic-wide migration is not random and highly structured, migration among adjacent and not too remote islands may be incomplete. At this smaller, regional scale, plant migration appears to be dispersal-limited. Not all islands within the migration distance of the species were actually occupied, an observation that is not dependent on island size. This observed dispersal limitation might be explained by two factors that have the same result for the species number. First, as implicit in the model, diaspores have not reached all islands. Second, the species were already dispersed to the islands but could not establish on these islands owing to, for example, absence of shelter, suitable climate, certain habitats, bedrocks or relief energy (Sauer, 1991). These factors can only difficultly be incorporated in the model to explain the difference between expected and observed species number. Considering the theory of island biogeography the incorporation of such factors is even not necessary because habitat diversity usually increases with area (Rosenzweig, 1995, and references therein). Moreover, according to the results of the present modelling approach, the number of missing species is not correlated with the available ice-free area on the islands. Small islands with, for example, a small number of habitats and large islands with greater habitat diversity are equally not filled with species. Thus, plant migration in the Arctic appears indeed dispersal-limited. Other reasons for this observation of incomplete migration are insufficient time to complete the migration process or species saturation. The latter appears incomprehensible considering the virtual ample space available in open arctic landscapes. Another explanation is possible in some species, if only southern islands are occupied but not northern islands within the migration distance: these species may be at their northern limit and cannot spread further north. Examples would be species occurring only on islands closest to the continents and not differing in their observed and modelled distribution (261 out of 1256 taxa, e.g. Coeloglossum viride, Salix fuscescens and Rubus arcticus). However, this possible explanation has been eliminated because the model was controlled for the zonal distribution of the species: thus, spurious migration into areas apparently unfavourable for such plants, as inferred from their present distribution, was omitted. Thus, the species show largely an incomplete dispersal among islands that are close.

Changing climate tends to result in plant migration (e.g. Coope, 2004). The MPM model suggests how arctic island biodiversity may have developed. Arctic islands appear to have been recolonized rapidly, since the last Ice Age, from the closest mainland areas or the larger islands (Greenland and Iceland) in a rather predictable way. However, judging from the higher than expected number of species on the islands, exchange of species among adjacent islands appears to be dispersal-limited and not yet complete. Present species numbers on the Arctic islands may thus increase if migration becomes more complete and may not necessarily be linked to climate change.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

I thank Robert Ricklefs and three reviewers for helpful discussions and N. Tkach and M. Röser for their support. Funding of the project was provided by the German Science Foundation.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • Abbott RJ, Brochmann C. 2003. History and evolution of the arctic flora: in the footsteps of Eric Hultén. Molecular Ecology 12: 299313.
  • Abbott RJ, Smith LC, Milne RI, Crawford RMM, Wolff K, Balfour J. 2000. Molecular analysis of plant migration and refugia in the Arctic. Science 289: 13431346.
  • ACIA. 2006. Arctic Climate Impact Assessment: Scientific Report. Cambridge, UK: Cambridge University Press.
  • Ackerly DD. 2003. Community assembly, niche conservatism, and adaptive evolution in changing environments. International Journal of Plant Sciences 164: S165S184.
  • Alsos IG, Alm T, Normand S, Brochmann C. 2009. Past and future range shifts and loss of diversity in dwarf willow (Salix herbacea L.) inferred from genetics, fossils and modelling. Global Ecology and Biogeography 18: 223239.
  • Alsos IG, Eidesen PB, Ehrich D, Skrede I, Westergaard K, Jacobsen GH, Landvik JY, Taberlet P, Brochmann C. 2007. Frequent long-distance plant colonization in the changing Arctic. Science 316: 16061609.
  • Baldwin BG, Sanderson MJ. 1998. Age and rate of diversification of the Hawaiian silversword alliance (Compositae). Proceedings of the National Academy of Sciences, USA 95: 94029406.
  • Bell G. 2001. Neutral macroecology. Science 293: 24132418.
  • Böhle UR, Hilger HH, Martin WF. 1996. Island colonization and evolution of the insular woody habit in Echium L. (Boraginaceae). Proceedings of the National Academy of Sciences, USA 93: 1174011745.
  • Borgatti SP. 2002. NetDraw: graph visualization software. Harvard, MA, USA: Analytic Technologies.
  • Brochmann C, Gabrielsen TM, Nordal I, Landvik JY, Elven R. 2003. Glacial survival or tabula rasa? The history of North Atlantic biota revisited. Taxon 52: 417450.
  • Butts CT. 2009. sna: Tools for social network analysis. R package version 2.0-1.
  • Charkevicz SS. 19851996. Sosudistye rastenija sovetskogo dalnego vostoka – Flora of the Soviet Far East vol. 1–8. Leningrad, Russia: Nauka.
  • Clark JS, Fastie C, Hurtt G, Jackson ST, Johnson C, King GA, Lewis M, Lynch J, Pacala S, Prentice C et al. 1998. Reid’s paradox of rapid plant migration. BioScience 48: 1324.
  • Coope GR. 2004. Several million years of stability among insect species because of, or in spite of, Ice Age climatic instability? Philosophical Transactions of the Royal Society London Series B 359: 209214.
  • Eidesen PB, Carlsen T, Molau U, Brochmann C. 2007. Repeatedly out of Beringia: Cassiope tetragona embraces the Arctic. Journal of Biogeography 34: 15591574.
  • Elven R, Murray DF, Razzhivin V, Yurtsev BA. 2005. Checklist of the Panarctic Flora (PAF): vascular plants. Oslo, Norway: University of Oslo.
  • Hanneman RA, Riddle M. 2005. Introduction to social network methods. Riverside, CA, USA: University of California, Riverside (published in digital form at http://faculty.ucr.edu/~hanneman/).
  • Hoffmann MH. 2005. Evolution of the realized climatic niche in the genus Arabidopsis (Brassicaceae). Evolution 59: 14251436.
  • Hoffmann MH, Röser M. 2009. Taxon recruitment of the arctic flora: an analysis of phylogenies. New Phytologist 182: 774780.
  • Hubbell SP. 2001. The unified neutral theory of biodiversity and biogeography. Princeton, NJ, USA: Princeton University Press.
  • Hultén E. 1968. Flora of Alaska and neighboring territories: a manual of the vascular plants. Stanford, CA, USA: Stanford University Press.
  • Hultén E, Fries M. 1986. Atlas of North European vascular plants north of the tropic of cancer vol. 1–3. Königstein, Germany: Koeltz.
  • Johansen S, Hytteborn H. 2001. A contribution to the discussion of biota dispersal with drift ice and driftwood in the North Atlantic. Journal of Biogeography 28: 105115.
  • Krasnoborov IM, Malyschev LI, Peschkova GA, Polozhij AV.(eds.). 1988-1997. Flora Sibiri vol. 1–13. Novosibirsk, Russia: Nauka.
  • Martin AD, Quinn KM, Park JH. 2009. MCMCpack: Markov chain Monte Carlo (MCMC) Package. R package version 1.0-4.
  • Nathan R, Katul GG, Horn HS, Thomas SM, Oren R, Avissar R, Pacala SW, Levin SA. 2002. Mechanisms of long-distance dispersal of seeds by wind. Nature 418: 409413.
  • Porsild AE, Cody WJ. 1980. Vascular plants of the continental Northwest Territories, Canada. Ottawa, Ontario, Canada: National Museum of Canada.
  • de Queiroz A. 2005. The resurrection of oceanic dispersal in historical biogeography. Trends in Ecology and Evolution 20: 6873.
  • R Development Core Team. 2009. R: A language and environment for statistical computing. Vienna, Austria: Foundation for Statistical Computing.
  • Rosenzweig ML. 1995. Species diversity in space and time. Cambridge, UK: Cambridge University Press.
  • Sauer JD. 1991. Plant migration. Berkeley, CA, USA: University of California Press.
  • Savile DBO. 1972. Arctic adaptations in plants. Research Branch Canada Department of Agriculture. Monograph 6: 1181.
  • Simpson GG. 1952. Probabilities of dispersal in geological times. All-or-none versus degrees of probability. Bulletin of the American Museum of Natural History 99: 163176.
  • Skrede I, Borgen L, Brochmann C. 2009. Genetic structuring in three closely related circumpolar plant species: AFLP versus microsatellite markers and high-arctic versus arctic–alpine distributions. Heredity 102: 293302.
  • Tkach NV, Röser M, Hoffmann MH. 2008. Ranges and range size variation in the vascular plant flora of the Eurasian Arctic. Organisms Diversity and Evolution 8: 251266.
  • Tolmachev AI (ed.). 19601987. Arkticheskaya Flora SSSR. Moscow, Leningrad, Russia: Nauka.
  • Walker DA, Raynolds MK, Daniels FJA, Einarsson E, Elvebakk A, Gould WA, Katenin AE, Kholod SS, Markon CJ, Melnikov ES et al. 2005. The circumpolar arctic vegetation map. Journal of Vegetation Science 16: 267282.
  • Westergaard KB, Alsos IG, Popp M, Engelskjoen T, Atberg KI, Brochmann C. 2011. Glacial survival may matter after all: nunatak signatures in the rare European populations of two west-arctic species. Molecular Ecology 20: 376393.
  • Westergaard KB, Jorgenson MH, Gabrielsen TM, Alsos IG, Brochmann C. 2010. The extreme Beringian/Atlantic disjunction in Saxifraga rivularis (Saxifragaceae) has formed at least twice. Journal of Biogeography 37: 12621276.
  • Woodward FI. 1987. Climate and plant distribution. Cambridge, UK: Cambridge University Press.

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Table S1 Arctic islands, geographic coordinates and references for plants recorded on these islands

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

FilenameFormatSizeDescription
NPH_3924_sm_TableS1.doc105KSupporting info item