SEARCH

SEARCH BY CITATION

Keywords:

  • Autocorrelation;
  • Diversity;
  • Introduced species;
  • Island biogeography;
  • Pine;
  • Richness;
  • Spatial Autoregression;
  • Species space

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Questions

Are there differences in species composition and richness between islands that were reforested more than 70 yr ago with the introduced Pinus mugo compared with islands supporting the native Pinus sylvestris? Do the results depend on autocorrelation in geographical space and species ordination space? Species richness is expected to increase as a function of the size of an island; are the responses to island size different between P. mugo and P. sylvestris islands. Does the land-use history have an impact on the current species composition and richness pattern?

Location

The archipelago is in the oceanic section of the Atlantic bioclimatic zone, west Norway. This archipelago was part of the ancient and widespread treeless heathland found along the European west coast.

Methods

Data on vascular plants were compiled from the forested islands, and their differences in species composition were analysed by ordination. The hypotheses were tested by means of t-tests and generalized linear models, the spatial component was accounted for by means of Moran's I and spatial autoregression with the moving average approach. This was done both in geographical space and species ordination space.

Results

There are more vascular plants on the islands with introduced P. mugo than on the islands with native P. sylvestris. The latter have rather homogenous undergrowth dominated by bryophytes. This may explain lower richness on islands with native forest and why island size is not correlated with species richness on these islands. In contrast, P. mugo is easily wind-felled in autumn storms, which keeps rocky microhabitats exposed to air and new forest habitats are created. Species that are associated with the previous land-use system (grazing) prevail on islands with introduced pine, and thus contribute to higher plant richness.

Conclusions

The difference in species richness and island species–area relationship (ISAR) between P. mugo and P. sylvestris islands may relate to the same underpinning causes. Species from the old land-use system have survived on P. mugo islands, but not in the late-successional forest with a more closed canopy that has developed on P. sylvestris islands. Thus habitat and species richness is higher and increases with area on P. mugo islands but not on P. sylvestris islands.


Nomenclature
Lid & Lid

2005

Abbreviations
DCA

detrended correspondence analysis

GLM

generalized linear models

ISAR

Island species–area relationship

MA

moving average

Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

The Millennium Ecosystem Assessment (MEA 2005) identifies the transport of species across biogeographic boundaries as a major threat to biodiversity worldwide. This has been widely accepted by conservationists (cf. Lodge & Shrader-Frechette 2003; Perrings et al. 2005). However, while some studies find that invasive species have strong negative impacts on native biodiversity, comparable to the effects of habitat destruction (Wilcove & Chen 1998; Pimentel et al. 2001; Pauchard & Shea 2006), others fail to document any negative impacts and conclude that plant invasions do not cause species extinctions, at least not on regional or broader scales (Sax et al. 2002; Gurevitch & Padilla 2004; Maskell et al. 2006; Stohlgren et al. 2006). It has thus been argued that direct effects of alien species on the species richness component of biodiversity is not very well documented (Gurevitch & Padilla 2004; but see Powell et al. 2011), and the overall global impact is still rather enigmatic (Rosenzweig 2001; Slobodkin 2001; Barney & Whitlow 2008).

There are several reasons for these ambiguous reports (Davis 2003). First, there is a lack of coherent scaling across studies, which makes direct comparisons of ecological impacts questionable (Shea & Chesson 2002; Gooden et al. 2009). Second, it is often difficult to make clear causal links between the spread of alien plants and the demonstrated negative impact. Third, the concepts and terminology in the emerging field of invasive biology are often ambiguous (Colautti & MacIsaac 2004), for instance, it is important to distinguish between infrequent introduced species that have minimal ecological impacts and species that are invasive due to high rates of population growth, rapid spread and strong negative impacts on native species richness and ecosystem function (Young & Larson 2011).

The potential impact of introduced species will also depend on biogeographic conditions. Alien species may represent a threat to young terrestrial ecosystems, such as forest ecosystems on small islands along biogeographically isolated coasts (Pretto et al. 2010). A number of small islands along the North Atlantic coast of Norway have emerged since 7000 BP due to sea level fluctuations in postglacial times (Kaland 1984; Bondevik et al. 1998). This relatively young archipelago has a very peripheral location at the north-western fringe of the Palaearctic flora region, and it is not certain if these areas are saturated with species from the Eurasian species pool (cf. Svenning & Skov 2007). The flora of Norway, which covers this study area, consists of a high proportion of introduced species, partly because of an almost complete glacial cover during the last glacial period. Between 22% and 50%, depending on the temporal criteria applied (Fremstad & Elven 1997), of the vascular plants in Norway have been introduced. Only a subset (ca. 4%) of these has spread, and only a small proportion of these species has had a potentially significant negative ecological impact (Gederaas et al. 2008). The biogeographical conditions along the North Atlantic coast indicate that this area might be vulnerable to the potential effects of introduced species.

Scots pine (Pinus sylevstris L.) is a potentially dominant canopy tree in large fractions of the boreo-nemoral and North Atlantic parts of the Palaearctic, both on the mainland and the islands. The natural pine forests along the North Atlantic coast of Europe were, however, transformed by humans 1–5 kyr BP. Fire was used to create open heathlands suitable for year-round grazing by domestic animals (Kaland 1986; Loidi et al. 2010). During the last 130 yr, a decreasing farming population, changes in farming practices and public reforestation schemes have resulted in pine forest on many of the islands. Some islands have native Scots pine (P. sylvestris) and others have a mix of mountain dwarf pines (Pinus mugo Torra complex, mainly subsp. mugo and subsp. uncinata). Alongside this reforestation, the traditional land-use regime and open heathlands have persisted. The rugged coast and many islands of western Norway have allowed different forest types to persist alongside each other within the same landscapes. This study is carried out within an archipelago of coastal islands, and compares islands that have been reforested for at least 70 yr with native Scots pine (P. sylvestris) with islands that were planted at least 70 yr ago with introduced mountain dwarf pines from the sub-alpine zone in the Pyrenees, i.e. P. mugo subsp. mugo and P. mugo subsp. uncinata (hereafter P. mugo coll.).

The first step in the study is to analyse the relationship between the size of the islands and their number of species, i.e. the well-documented island species–area relationship (ISAR; Rosenzweig 1993; Whittaker & Fernandez-Palacios 2007; Tjorve & Tjorve 2011). Larger islands have a higher probability of harbouring more species than smaller islands, which is often explained by an increased number of habitats as a function of island size (Whittaker & Fernandez-Palacios 2007). However, some studies have tried to disentangle the ‘size effect’ from the ‘habitat effect’ (Westman 1983; Kallimanis et al. 2008). The number of habitats or ‘entities’ within a landscape is not easy to delimit, and therefore the relative importance of area vs habitat effects is elusive. In this study, we explore these phenomena by contrasting the ISAR of islands reforested with native Scots pine vs islands of the same size range that are planted with introduced pine. We assume that the size of the island correlates with the number of topographically and edaphically defined habitats, but the two tree species have different effects on habitat heterogeneity, as the old-growth Scots pine forests can tolerate the severe wind conditions at the coast and are therefore relatively undisturbed. Within these forests, light conditions appear more uniform due to a closed tree canopy, and on the ground there are extensive moss carpets that cover the microtopographic variation. Thus, we hypothesize that Scots pine forests, which are in a moss-dominated late-successional phase, may counteract the positive effect of island area. Due to proportionally fewer plant habitats on islands with old-growth Scots pine forest, we predict a relatively shallow regression slope of ISAR (i.e. lower z-value) on islands reforested with native Scots pine.

The direct effect of P. mugo coll. on species richness is the overall research question. It is either a passenger species that does not influence the local environment (Chabrerie et al. 2008), or it is more of an ecosystem engineer that directly influences species composition and richness (Jones et al. 1994). The potential biological life span of the introduced pine is much shorter than that of P. sylvestris, and it is easily wind-felled during storms, resulting in higher disturbance and more variable light and temperature conditions on islands reforested with the introduced pine. The higher disturbance frequency may also result in higher colonization success on islands with P. mugo (hereafter P. mugo islands) than islands with P. sylvestris forest (hereafter P. sylvestris islands). These ecological differences predict higher species richness per unit area on P. mugo islands compared with P. sylvestris islands. At the same time, however, many coniferous species are invasive and particularly within the genus Pinus (Richardson & Rejmanek 2004; Richardson 2006), which predicts negative impacts on native diversity. Hence, we test the null hypothesis of no difference in species richness between P. mugo and P. sylvestris islands, and interpret any deviations from this null hypothesis (positive or negative) in light of the invasive species effect and habitat heterogeneity hypotheses outlined above. The clear spatial nature of studies on oceanic islands requires that we take autocorrelation into account (Selmi & Boulinier 2001), and we have explicitly considered both the geographical space and species ordination space in the analyses. The primary aims are to test the following hypotheses

  1. The increase in number of species as a function of the size of an island is steeper on P. mugo than P. sylvestris islands
  2. There is no difference in species richness between islands reforested with P. sylvestris and those that are reforested with P. mugo coll.
  3. The significance of the differences in species richness is independent of the positions of islands in geographical space and species ordination space.

If the two last hypotheses are rejected, we aim to use differences in species composition and ecological attributes among the species unique to one type of island to aid the causal interpretation of the differences. The land-use argument will be reinforced if shade-tolerant forest core species prefer islands with old-growth Scots pine forest, whilst anthropogenic grazing-tolerant species are more frequent on islands that have been reforested by P. mugo coll.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Area

The study took place on an archipelago in the south-western part of the Scandinavian Peninsula, bordering the North Atlantic Ocean (59°42′–60°44′ N, 05°02′–05°35′ E; Fig. 1). The islands range from 0.1 to 4.6 ha (mean = 1.0 ± 0.12 ha). The bedrock consists mainly of gneiss and granite, mostly with an acidic and thin soil layer, although there are large areas that consist of humus directly on bedrock. Typical podsol profiles develop in favourable microtopographic locations. The islands are uninhabited and are not actively used for agricultural or forestry purposes.

image

Figure 1. Map showing the study area in Norway and a crude indication of the area where the 70 islands were sampled.

Download figure to PowerPoint

The climate is oceanic with mild winters with high precipitation and cool and moist summers. The mean annual temperature is 6.7 °C, mean summer temperature (June–July) is 13.3 °C, and mean winter temperature (February) is 0.7 °C. Mean annual precipitation is 1815 mm, and winds are strong, with an October average of 5.5 m·s−1, but events with wind speeds above 20 m·s−1 are common every year (Flesland airport climate station). The archipelago is located in the euoceanic section of the Atlantic bioclimatic zone (Moen 1998). Pinus sylvestris is the only coniferous forest tree native to this region, and forest cover is typically not dense, with several deciduous trees contributing to the lower canopy, including Betula pubescens, Sorbus aucuparia, Prunus padus and Alnus glutinosa, with Juniperus communis common in the shrub layer. Dwarf shrubs (chamaephytes) such as Empetrum hermaphroditum and Calluna vulgaris dominate, together with species belonging to the genera Vaccinium and Erica. Herb and graminoid richness and cover are relatively low, but a few ferns are very common, e.g. Polypodium vulgare and Dryopteris dilatata. The ground layer is well developed and dominated by bryophytes, such as Hylocomium splendens (H edw.) B.S.G., Pleurozium schreberi (Brid.) Mitt. and Sphagnum spp. Nomenclature for all vascular plants follows Lid & Lid (2005), except Pinus mugo Torra complex (Christensen 1987) and Picea sitchensis (Bong) Carr.

Land use

The original deforestation of the pine forest on these islands was not synchronized, which suggests that it was driven by land use (Kaland 1986; Hjelle et al. 2010). This is marginal agricultural land, but rich fishing resources contributed to livelihoods and supported a relatively high coastal population. In the late 1800s these islands were very heavily utilized due to steep human population growth, combined with a largely agricultural economy with more than 85% of the population in western Norway engaged in smallholder farming and animal husbandry (Oyen et al. 2006). The reforestation started in the late 19th century as a result of emigration, urbanization and government schemes to promote P. mugo subsp. mugo, aiming to provide the poor farmers with fuelwood. Later, in the 20th century, P. mugo subsp. uncinata was also planted. The islands with native P. sylvestris were also deforested during the most intensive farming period due to high demand for timber and fuel (Oyen et al. 2006). As judged from the age of the oldest trees on the studied islands, the forests are at least 70 yr old.

Field methods

Small islands were selected based on their forest cover of either native pine (= 36) or introduced pine (= 34) forest of at least 70 yr old. Other native trees are present in the forests (see above), and some individuals of Picea abies and P. sitchensis (Bong.) Carr had also been planted on the Pinus mugo islands. The size class distributions of the two island types are similar, with a mean size of 9811 ± 1789 m2 (native) and 9679 ± 1507 m2 (introduced). The island size and area sampled were restricted to the vegetated (forested) areas, excluding the littoral zone. Consequently, species with habitats restricted to the littoral zone were not included in the species list for each island, but species that occurred on outcrops within the forested area (e.g. Arctostaphylos uva-ursi, Sedum anglicum) were included. We only included native species, which means species such as Cotoneaster spp. and Acer pseudoplatanus were excluded. There were never more than one or two species of this type on each island and they normally amounted to a few individuals. Species were recorded during a clockwise coil-shaped walk from sea level towards the top. We then made a similar walk (counter-clockwise) from the top towards the littoral zone. To improve the species search we did an intensive investigation of two 100-m2 plots, one of which was located in the middle of the island (normally the top), and the other in the forest near sea level in a somewhat flat area. Approximately 90–180 min were spent sampling each island, depending partly on the size of the island, but variable topography and accessibility made it impossible to set a fixed time proportional to the size of the islands for sampling. The islands have relatively few vascular plants (mean 20.8 ± 0.7); thus it is not unreasonable to assume that the number of recorded species approximate the total number of vascular plants on each island.

Numerical analyses

The analytical path

The principal response variable is the number of native species excluding coniferous trees and species restricted to the littoral zone. First, we compared the species–area relationship between islands by means of linear regression, i.e. log species against log area, a model derived from the classical Arrhenius (1921) formula (Tjorve & Turner 2009). There is no significant correlation between species richness and the distance from the mainland (isolation effect). Richness is negatively correlated with the estimated terrestrial area within a 500-m radius around the island centre. The species–area relationship was used to correct for island size before the difference in species number between P. mugo and P. sylvestris islands was analysed using univariate statistics. If the difference was significantly different from zero, we continued the significance test with multivariate regressions, including spatial models in combination with ordination. This enabled a test that takes account of both the location of islands in geographical space and the location of islands in species space (two-dimensional species ordination space; Diniz-Filho et al. 2003). The rationale is that, on average, species richness on nearby islands has a higher probability of being similar compared with distant islands, i.e. standard spatial autocorrelation or distance decay. Spatial autocorrelation violates the standard statistical assumption that observations are independent of one another, and is problematic because it will inflate the degrees of freedom and hence the chance of making a type-I error, i.e. rejecting a correct null hypothesis (Dormann et al. 2007). There are three main approaches to this challenge: (1) the partial approach is to partial out the deviance that correlates with geographical coordinates and then analyse with respect to the target explanatory variable (Borcard et al. 1992); (2) the residual approach aims to test the residuals for remaining spatial structure posterior to testing of the predictor (Hawkins et al. 2007); and (3) the simultaneous approach includes the spatial structure in a spatial regression (Bini et al. 2009). In species richness analyses, however, the similarity in species composition should also be taken into account, i.e. location in species ordination space (Diniz-Filho et al. 2003). To obtain this, we used detrended correspondence analysis (DCA) on a binary data set based on presence of all species in the forest in each of the 70 islands, excluding species only found on two islands (69 species in total). The rationale is that if two neighbouring islands have the same number of species, but a totally different flora, the number of species is actually statistically independent (no effect of the proximity in space). Although this extreme example is unlikely in nature, it illustrates how the degree of shared species will also determine the degree of statistical independence between the islands. Many shared species yield high statistical dependence and vice versa, i.e. species ordination space autocorrelation. The location of sampling points in species ordination space will be almost as important as the geographical distance because this determines the statistical independence of species richness observations. We therefore used both geographical coordinates and ordination scores on first and second ordination axes (see below) and calculated Moran's I as a measure of autocorrelation. This was done both for the response variable (species richness) and for the residuals after multiple regressions, where the explanatory variables are area and native vs introduced trees. If there is significant autocorrelation in several distance classes in the residuals it may indicate that an important variable is missing (Hawkins et al. 2007). In our case, we added an explanatory variable that indicates how well the grazing indicators have survived after the gazing practice ceased, i.e. number of grazing indicator species based on Norderhaug et al. (1999) and Fremstad (1997). This is merely a variable that may aid interpretation of the results. Finally, we used moving average (MA) as a spatial regression approach to include the spatial structure of geographical space and species space in the regressions, where area was entered before the target predictor of native vs non-native forest, and the grazing indicator.

Numerical tests and software

We used a simple t-test to check if the there are significant differences in species richness between P. mugo and P. sylvestris islands. We used generalized linear models (GLM) with a log-link function (assuming a Poisson distribution of errors) to test if species richness is explained by the forest type on the islands, and included island size as a covariable. The residuals from this regression were then checked with Moran's I for significant positive spatial autocorrelation, both in geographical space as well as in species ordination space (applying the DCA axes score for ordination above). We used Moran's I to test for positive spatial autocorrelation of the response variable before and after the GLM regressions. As a third alternative, we used moving average (MA) as a simultaneous spatial regression approach because it has the ability to minimize spatial autocorrelation in residuals, although we are not able to quantify all biological and ecological processes generating the spatial structure in the data (Bini et al. 2009). We used default DCA in CANOCO (v. 4.5 for Windows, Microcomputer power, Ithaca, NY, USA) to display graphically the islands on the basis of species present in the forest of the islands, and relate the axes to spatial coordinates, size of islands, type of forest and browsing/grazing indicator species. R software was used for the t-test and GLM regressions (R Foundation for Statistical Computing, Vienna, AT). Spatial analysis in macroecology (SAM, v. 4.1; Rangel et al. 2010) was used for Moran's I and spatial autoregression with the MA approach.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

Island species–area relationships

There is a statistically significant relationship between island size and total number of species found on the islands (Fig. 2a). The z-value (regression slope) is 0.1 for the total data, but the slopes vary between island types: whereas the increase in species number as a function of island size is steep and highly significant on P. mugo islands (z-values = 0.23; Fig. 2b), no statistically significant ISAR relationship is found on P. sylvestris islands (Fig. 2c).

image

Figure 2. Island species–area relationship (ISAR): (a) for all islands together: log(spp) = 0.93 + 0.1 × log(area) (< 0.01), (b) Pinus mugo islands: log(spp) = 0.48 + 0.23 × (log)area (< 0.001). (c) Pinus sylvestris islands (not significant).

Download figure to PowerPoint

Species richness

A total of 77 native vascular plant species were recorded on P. sylvestris islands, whereas 75 species were recorded on P. mugo islands. The mean number of species is 18.5 ± 0.69 on P. sylvestris islands and 23.2 ± 1.14 on P. mugo islands, which is a statistically significant difference (< 0.001, = 3.5). When the species numbers are corrected for island size, mean number of species on P. mugo islands decreases somewhat, but is still significantly different (< 0.001, = 3.1; Fig. 3) from the mean species richness on P. sylvestris islands. The null hypothesis of no difference in species number between P. mugo islands and P. sylvestris islands is therefore rejected, but see below for how the outcome varies depending on the approach used in multivariate analyses to tackle the autocorrelation challenge.

image

Figure 3. Box plots showing significant differences in mean species number on Pinus mugo islands (1) and Pinus sylvestris islands (2). The panel to the left depicts uncorrected values (1 = 23.2 ± 1.4; 2 = 18.5 ± 0.7), and the panel to the right depicts corrected values for P. mugo islands (1 = 20.7 ± 0.7). The overall uncorrected mean is 20.8 ± 0.7.

Download figure to PowerPoint

Islands in species space and species turnover

The difference in species composition between P. mugo and P. sylvestris islands is significant, although not very strong (Fig. 4). The variation along the first ordination axis in Fig. 4 correlates with geographical position (east–west): P. sylvestris islands are more common in the eastern part of the study area, towards the mainland, whereas P. mugo islands are more commonly found in the west. The latter islands are characterized by Erica cinerea, Phegopteris connectilis, Carex rostrata and species that indicate light and/or disturbance, including Digitalis purpurea, Juncus conglomeratus and Rubus idaeus, which are almost exclusively on P. mugo islands. Forest species such as Myrica gale, Linnaea borealis and Melampyrum pratense are almost exclusively in the native pine forest (Fig. 4, Appendix S1). Species turnover (SD units) within each island type differs only slightly between the two types of island (P. mugo: 1.65 SD; P. sylvestris: 1.73 SD). The covariation between species composition and geographic location represents an analytical challenge (see below).

image

Figure 4. DCA ordination diagrams of binary data showing the difference between Pinus sylvestris islands and Pmugo islands. Geographical and biological variables are correlated with DCA axes and superimposed on the diagram. natspp = number of native vascular plants on each island; grazing = number of indicator species that are associated with grazing activity (see Appendix S1); east = eastern coordinates; north = northern coordinates, log-area = log size of the islands; alien = number of alien tree species found on the islands.

Download figure to PowerPoint

Autocorrelation

Species richness on the islands is affected by the location of the island, both in geographical and species ordination space (Fig. 4, Table 1). This is captured by Moran's I where species richness (uncorrected) is significantly autocorrelated in geographical space, particularly over short distances (<0.1–0.5 km; Fig. 5a). The three alternative approaches used to deal with this autocorrelation yield different results. The first approach was to partial out the spatial or compositional structure in the data by including either of these structural variables as covariates in the GLM regression models. Using this approach, the difference in richness between P. mugo and P. sylvestris islands is statistically significant if geographical space is a covariable (Table 2B), but not if species ordination space is used (Table 2C).

Table 1. Summary statistics of detrended correspondence analyses (DCA) on a binary species by island matrix (69 × 70; cf. Fig. 4, Appendix S1). The first and second axes are used in the autocorrelation and regression analyses. Total inertia = 1.548. Significant correlation coefficients are shown for the covariables (area, east, north) as well as explanatory variable (grazing indicators), and the direction of maximum richness (natspp) and number of alien tree species on the islands (alien)
DCA axes12
Eigenvalue0.1370.10
Gradient length1.861.45
Species–env. correlation0.850.73
Northns0.47
East0.23ns
Natspp−0.70ns
Log(area)ns0.58
Alien−0.45ns
Grazing−0.750.22
Table 2. Generalized linear model (GLM) where native vs non-native pine forest is the explanatory variable and null deviance is 114.1 with 69 df. (A) Residual analyses approach where (log)area is the only covariable, and explanatory variables are first native vs alien pine (native) and then grazing indicator (grazing). The residuals are checked for autocorrelation by means of Moran's I with respect to geographical space (cf. Fig. 5b) and species ordination space (cf. Fig. 5c), and here we also added grazing as an explanatory variable (cf. Fig 5d). (B) Partial regression approach: covariables are island size log(area) and geographical position (north and east), and (C) log(area) and locations in species ordination space (DCA axes 1 and 2). Native is not significant when ordination space is added first (C), cf. analyses of residuals (Fig 5c)
 VariablesdfDevianceResid. dfResid. DevPr(Chi)
  1. df, degrees of freedom; Resid., residuals; Dev., deviance; Pr(Chi), P-value of Chi-test.

ALog(area)113.9068100.18<0.001
Native117.046783.14<0.001
Grazing141.286641.86<0.001
BLog(area)113.9068100.18<0.001
North12.046780.93>0.15
East17.026691.55<0.01
Native17.086584.04<0.01
CLog(area)113.9068100.18<0.001
DCA-1 score146.706753.48<0.001
DCA-2 score13.334650.15>0.06
Native10.116550.05>0.74
image

Figure 5. (a) Moran's I of uncorrected species richness values as a function of geographical distance indicates significant spatial autocorrelation in species richness. The shortest distance class of the residuals is significantly autocorrelated (= 0.03), i.e. residuals from GLM regression (cf. Table 2A; island size = covariable; native/alien forest = predictor). (b) Moran's I of residuals after a GLM (model: species richness ~ log(Area)) plotted against geographical distance. (c) The same residuals as in (b) plotted against species ordination space distance. (d) When grazing indicator variable is added to the above GLM model; only marginal positive autocorrelation in the shortest distance class of the residuals remains (cf. Table 2A). Negative autocorrelations, i.e.more different than random expectation are not interpreted. The distance units in (a) and (b) are geographical distances based on UTM coordinates, (c) and (d) are arbitrary based on species ordination space (DCA axes 1 and 2). Conf. Interv., confidence interval.

Download figure to PowerPoint

The second approach was to first regress and then examine the residuals for autocorrelation. The two forest types differ significantly in the GLM regression model with species richness as the response variable and (log)area as a covariable (Table 2A). The residuals from this GLM model show only marginally significant autocorrelation over the shortest distance (Fig. 5b), but the residuals in species composition space show very clear and significant autocorrelation (Fig. 5c). When we added the grazing indicator variable (Table 2A), the residuals from this GLM model show no significant autocorrelation (Fig. 5d).

The third approach was to apply simultaneous spatial regression with moving average, with geographical space implicitly taken into account and island size included as a covariable and the grazing indicator variable added. These yield significant differences between P. mugo and P. sylvestris islands, but the significance is strongest in geographical space and rather weak in species ordination space (< 0.03; Table 3B).

Table 3. Spatial regression with moving average (MA) approach. Log-area, native vs non-native pine forest, and number of grazing indicator species (grazing) are explanatory variables. The spatial structure is entered as (A) geographic coordinates with F = 20.97; R2 = 0.65; spatial autoregressive parameter rho = 0.916; and (B) species ordination space with F = 40.97; R2 = 0.81 and rho = 0.982 (alpha = 1.0 in both cases)
VariablesOLS coeff.MA coeff.Std. coeff.SEt-valueP-value
  1. OLS, ordinary least square regression; MA, moving average; Coeff., coefficient; Std, standard.

(A) Constant−0.311−10.71905.4811.984>0.051
Log-area4.7937.0700.4831.3895.088<0.001
Native1.928−1.4890.2110.9432.638<0.001
Grazing2.0491.9560.6320.2996.535<0.001
(B) Constant−0.3111.71904.2440.405>0.687
Log-area4.7935.2240.3571.0904.792<0.001
Native1.928−1.8600.1570.8842.105<0.039
Grazing2.0491.8560.6000.2577.233<0.001

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

The number of species found in the entire forest on each island is higher on P. mugo islands than P. sylvestris islands, except when using the partial regression approach with DCA axes as covariables. There are also clear differences in the species–area relationships between P. sylvestris islands (z-value ~0) and P. mugo islands (z-value = 0.23). We will discuss how the differences in ISAR and species richness composition may be related to similar underpinning causal factors, and how the demonstrated autocorrelation, particularly in species ordination space, may aid the interpretation and the plausible causal links (cf. Diniz-Filho et al. 2003).

Island species–area relationships

We confirmed the hypothesis that the increase in species as a function of island size is steeper on islands with non-native rather than native pine. The z-value on P. mugo islands is in the expected range (Lomolino & Weiser 2001 and references therein) and very close to the theoretical suggested value of 0.26 (MacArthur & Wilson 1963), whereas for P. sylvestris islands the relationships is not significant. The latter result may relate to the ‘small island effect’ (SIE; Niering 1963): the studied islands are within the size range for which several authors have claimed that there is no systematic trend in species richness (Lomolino & Weiser 2001; Triantis et al. 2006). However, it is unlikely that SIE is only apparent on islands with native pine forest that are very similar environmentally. We therefore hypothesize that the independence between species number and the size of the P. sylvestris islands is caused by the ecological conditions in the forest on these islands. Although we did not quantify the number of habitats in the forests on these islands, we did observe reduced microtopographic relief on P. sylvestris islands. This is due to extensive carpets of Sphagnum mosses that have developed on these islands, which cover most small depressions and form minor blanket bogs on steeper slopes. Dominant forest floor bryophytes (e.g. Sphagnum, Polytrichum, feather mosses) are able to suppress establishment and growth of vascular plants by reducing access to light, water, nutrients and space (Grime et al. 1990; Okland et al. 2004), and this effect may be more pronounced in oceanic climates with long, moist growing seasons (Okland et al. 2004). This is in contrast to P. mugo islands, where disturbance from wind-felled trees creates a variety of microhabitats, including openings in the forest, exposed rock outcrops and small open ponds not overgrown by Sphagnum mosses (these ponds must have been vital waterholes for the animals that used to browse on these islands). Extensive Sphagnum mats are more common on the largest islands that are well protected against the heavy sea storms, whereas on the small islands the wind pressure is not so effectively reduced, and Sphagnum mats cannot expand without interruption. This lack of disturbance causing differences in richness is in line with other studies that have shown that islands not affected by disturbance (e.g. fire, grazing, wind tree-felling) for a long time period may undergo ‘ecosystem retrogression’ (Peltzer et al. 2010), which means a reduction in decomposition rates, microbial biomass, light interception and increase in humus depth. This will not facilitate species richness of vascular plants, but may enhance the bryophyte diversity. Data from pine forests on lake islands in Scotland show that two-thirds of the total species number are bryophytes, whereas the number of vascular plants is in the same range as our plot data (not shown; Kerslake 1982). This indicates that the above rationale is valid for vascular plants, but not for bryophytes.

In conclusion, the number of potential habitats may be significantly reduced on the larger P. sylvestris islands relative to the P. mugo islands, and hence the number of vascular plant species did not increase as a function of island size. This reasoning assumes that the main cause of enhanced number of species on larger islands is mainly related to a positive correlation between habitat diversity and island area (Westman 1983; Whittaker & Fernandez-Palacios 2007; Kallimanis et al. 2008; Hortal et al. 2009).

Species composition, richness and autocorrelation

We found a geographical east–west gradient in species composition that co-varies with the variation in species richness (Fig. 4). This is a typical autocorrelation problem that inflates the degrees of freedom (Legendre 1993; Dormann et al. 2007). If we crudely partial out all deviance correlated to geographical space or species ordination space, the number of species between the two sets of islands is significantly different in the former approach but not in the latter. However, the GLM residual approach indicates that species richness is significantly higher on islands with introduced pine than native pine. The autocorrelation over short distances depicted by Moran's I for species richness is significant (Fig. 5a), but for the residuals after regression this disappears, except for the shortest distance class (Fig. 5b). The autocorrelation remains over several distances in species ordination space (Fig. 5c), but if we add the grazing indicator variable to the model the residuals show no significant autocorrelation (Fig. 5d). This result is followed up by the simultaneous approach where MA in spatial regression indicates that native vs non-native pine forest is a significant predictor together with the grazing indicator variable when geographical space or species ordination space are used in the MA model. The latter is a rather novel approach for autocorrelated species richness analyses (cf. Diniz-Filho et al. 2003). The rationale is outlined in Diniz-Filho et al. (2003), who emphasized the fact that neighbouring sampling locations (islands) may lack statistical independence if they have similar species composition. However, if they, in theory, had totally different species compositions they will actually appear as statistically independent. Thus it is equally important to take account of the location in species space as well as in geographical space, but one should be aware of the fact that the theoretical peak in richness can be anywhere in the geographical space, whereas in species ordination space it is only possible at one single spot were all species are present. This is because species composition and richness are intrinsically dependent on each other because both are an emergent property of species distributions. The location of this richness peak is most probably (but not necessarily) in the centre of a given species ordination space. However, in our case the maximum richness vector is pointing to the border of the species space and correlates with the first ordination axis. This illustrates that the grazing indicator species that have survived on the P. mugo islands are additional to the very common generalist type of species that cover many of the islands. The crucial point here is whether species composition and richness are driven by the same factors (Wagner & Fortin 2005). All the autocorrelation-related analyses clearly indicate that this is the case here, providing both an analytical challenge and an aid for interpretation.

Plausible causes of differences and management implications

In our case, species richness is enhanced on P. mugo islands, but this introduced species is more a passenger than an ecosystem engineer (Didham et al. 2005; MacDougall & Turkington 2005). The islands with higher species number have a set of grazing-tolerant species present that are rare on P. sylvestris islands. It is mainly these species that make the significant differences in species composition and richness. These species are associated with the past land-use regime, i.e. small stock grazing and fire, and have survived after grazing declined and P. mugo coll. was introduced (Fig. 4, Appendix S1). Although both types of islands have been irregularly burned and used for grazing in the first decades of the 20th century, the number of grazing-adapted species is higher on P. mugo islands. We hypothesize that this is related to the inherent properties of the two types of pine tree. The native P. sylvestris has been able to develop a closed canopy and represents an older successional stage compared with P. mugo forest. These conditions have facilitated extensive moss carpets, whereas the light-demanding herbs and grasses have disappeared. Pinus mugo on the other hand, is a sub-alpine species in its natural habitat, and in these extreme oceanic environments the trees are easily wind-felled in the autumn storms. Tree-fall caused by autumn storms is much more common on P. mugo islands than on P. sylvestris islands. This creates open habitats for herbs, grasses and ferns that were more common during the old land-use regime. This is, in part, in line with Lindborg & Eriksson (2004), who found time lags of 50–100 yr in the response of plant species diversity to a change in habitat in the landscape. In conclusion, one may argue that after the anthropogenic pressure was released and forest succession started, the successional phase on the P. mugo islands has been slower because of wind disturbance. It is not uncommon to find more vascular plant species in mid-successional phases than in old-growth forest (Vetaas 1997 and references therein), which explains the difference in species richness.

The initial rationale underpinning the introduction of P. mugo coll. was to conserve the soil, prevent erosion and provide fuelwood (Oyen et al. 2006). It has definitely contributed to the build-up of carbon stocks on these islands and has facilitated a succession towards forest. In this respect one has to re-evaluate P. mugo coll. from a potential invasive species threatening biodiversity to a species that contributes to restoration of the forest cover (cf. Schlaepfer et al. 2011). Although pine, at a global scale, is a genus that consists of many invasive species, particularly in the southern hemisphere (Richardson & Rejmanek 2004), we will argue, along with Schlaepfer et al. (2011), that in some exceptional cases the introduced species may actually contribute to ecological restoration and do not represent a threat to the local flora. Most of the islands do have a few native Scots pine individuals, and there were very few seedlings and saplings of P. mugo on these islands, therefore it is not unlikely that these islands will be forested with Scots pine in the future.

Conclusions

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

The difference in species richness and ISAR between P. mugo islands and P. sylvestris islands may relate to the same underpinning causes. The natural disturbance regimes are the same, but the native P. sylvestris is adapted to autumn storms, whereas the introduced P. mugo forests are in a state of perpetual gap dynamics. This may explain why P. mugo islands have a mixture of forest species and light-demanding species that are associated with grazing. Lower species richness in the mature old-growth P. sylvestris forest is due to an extensive moss carpet that reduces habitat diversity, and all the grazing indicator species have been shaded out. This may also explain why large size does not correspond to more species or a positive species–area relationship on the P. sylvestris islands.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information

We are grateful to Forest and Landscape (Bergen) and EECRG at the University of Bergen for useful comments on earlier presentations of this research. We thank Randi Danielsen, Atle Danielsen, Frida L. Vetaas, Leo Vetaas and Marit Vetaas Karlsen for assistance in the field. We thank Brooke Wilkerson and Cathy Jenks for invaluable editorial help, and two anonymous referees for instructive comments. The research was supported by the Norwegian Research Council (project no. 184099) and the Grolle Olsen legacy.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
  • Arrhenius, O. 1921. Species and area. Journal of Ecology 9: 9599.
  • Barney, J.N. & Whitlow, T.H. 2008. A unifying framework for biological invasions: the state factor model. Biological Invasions 10: 259272.
  • Bini, L.M., Diniz-Filho, J.A.F., Rangel, T., Akre, T.S.B., Albaladejo, R.G., Albuquerque, F.S., Aparicio, A., Araujo, M.B., Baselga, A., Beck, J., Bellocq, M.I., Bohning-Gaese, K., Borges, P.A.V., Castro-Parga, I., Chey, V.K., Chown, S.L., de Marco, P., Dobkin, D.S., Ferrer-Castan, D., Field, R., Filloy, J., Fleishman, E., Gomez, J.F., Hortal, J., Iverson, J.B., Kerr, J.T., Kissling, W.D., Kitching, I.J., Leon-Cortes, J.L., Lobo, J.M., Montoya, D., Morales-Castilla, I., Moreno, J.C., Oberdorff, T., Olalla-Tarraga, M.A., Pausas, J.G., Qian, H., Rahbek, C., Rodriguez, M.A., Rueda, M., Ruggiero, A., Sackmann, P., Sanders, N.J., Terribile, L.C., Vetaas, O.R. & Hawkins, B.A. 2009. Coefficient shifts in geographical ecology: an empirical evaluation of spatial and non-spatial regression. Ecography 32: 193204.
  • Bondevik, S., Svendsen, J.I. & Mangerud, J. 1998. Distinction between the Storegga tsunami and the Holocene marine transgression in coastal basin deposits of western Norway. Journal of Quaternary Science 13: 529537.
  • Borcard, D., Legendre, P. & Drapeau, P. 1992. Partialling out the spatial component of ecological variation. Ecology 73: 10451055.
  • Chabrerie, O., Verheyen, K., Saguez, R. & Decocq, G. 2008. Disentangling relationships between habitat conditions, disturbance history, plant diversity, and American black cherry (Prunus serotina Ehrh.) invasion in a European temperate forest. Diversity and Distributions 14: 204212.
  • Christensen, I.B. 1987. Taxonomic revision of the Pinus mugo complex and P. × rhaetica (P. × rhaetic-mugo × sylvestris) (Pinaceae). Nordic Journal of Botany 7: 348408.
  • Colautti, R.I. & MacIsaac, H.J. 2004. A neutral terminology to define ‘invasive’ species. Diversity and Distributions 10: 135141.
  • Davis, M.A. 2003. Biotic globalization: does competition from introduced species threaten biodiversity? BioScience 53: 481489.
  • Didham, R.K., Tylianakis, J.M., Hutchison, M.A., Ewers, R.M. & Gemmell, N.J. 2005. Are invasive species the drivers of ecological change? Trends in Ecology & Evolution 20: 470474.
  • Diniz-Filho, J.A.F., Bini, L.M. & Hawkins, B.A. 2003. Spatial autocorrelation and red herrings in geographical ecology. Global Ecology and Biogeography 12: 5364.
  • Dormann, C.F., McPherson, J.M., Araujo, M.B., Bivand, R., Bolliger, J., Carl, G., Davies, R.G., Hirzel, A., Jetz, W., Kissling, W.D., Kuhn, I., Ohlemuller, R., Peres-Neto, P.R., Reineking, B., Schroder, B., Schurr, F.M. & Wilson, R. 2007. Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography 30: 609628.
  • Fremstad, E. 1997. Vegetasjonstyper i Norge. Temahefte 12, Norsk Institutt for Naturforskning, Trondheim, NO.
  • Fremstad, E. & Elven, R. 1997. Alien plants in Norway. The large Fallopia species. Blyttia 55: 314.
  • Gederaas, L., Salvesen, I. & Viken, A. 2008. Blacklist for Norway 2007 and further afield. Insekt-Nytt 33: 47.
  • Gooden, B., French, K., Turner, P.J. & Downey, P.O. 2009. Impact threshold for an alien plant invader, Lantana camara L., on native plant communities. Biological Conservation 142: 26312641.
  • Grime, J.P., Rincon, E.R. & Wickerson, B.E. 1990. Bryophytes and plant strategy theory. Botanical Journal of the Linnean Society 104: 175186.
  • Gurevitch, J. & Padilla, D.K. 2004. Response to Ricciardi. Assessing species invasions as a cause of extinction. Trends in Ecology & Evolution 19: 470474.
  • Hawkins, B.A., Diniz, J.A.F., Bini, L.M., De Marco, P. & Blackburn, T.M. 2007. Red herrings revisited: spatial autocorrelation and parameter estimation in geographical ecology. Ecography 30: 375384.
  • Hjelle, K.L., Halvorsen, L.S. & Overland, A. 2010. Heathland development and relationship between humans and environment along the coast of western Norway through time. Quaternary International 220: 133146.
  • Hortal, J., Triantis, K.A., Meiri, S., Thebault, E. & Sfenthourakis, S. 2009. Island species richness increases with habitat diversity. The American Naturalist 174: E205E186.
  • Jones, C.G., Lawton, J.H. & Shachak, M. 1994. Organisms as ecosystem engineers. Oikos 69: 373386.
  • Kaland, P.E. 1984. Holocene shore displacement and shorelines in Hordaland, western Norway. Boreas 13: 203242.
  • Kaland, P.E. 1986. The origin and management of Norwegian coastal heaths as reflected by pollen analysis. In: Behre, K.E. (ed.) Anthropogenic indicators in pollen diagrams, pp. 3752. Balkema, Rotterdam, NL.
  • Kallimanis, A.S., Mazaris, A.D., Tzanopoulos, J., Halley, J.M., Pantis, J.D. & Sgardelis, S.P. 2008. How does habitat diversity affect the species–area relationship? Global Ecology and Biogeography 17: 532538.
  • Kerslake, P.D. 1982. Vegetational history of wooded islands in Scottish lochs. PhD thesis, University of Cambridge, Cambridge, UK.
  • Legendre, P. 1993. Spatial autocorrelation – trouble or new paradigm. Ecology 74: 16591673.
  • Lid, J. & Lid, D.T. 2005. Norsk Flora. Elven, R. (ed) 7th ed. Det Norske Samlaget, Oslo, NO.
  • Lindborg, R. & Eriksson, O. 2004. Historical landscape connectivity affects present plant species diversity. Ecology 85: 18401845.
  • Lodge, D.M. & Shrader-Frechette, K. 2003. Nonindigenous species: ecological explanation, environmental ethics, and public policy. Conservation Biology 17: 3137.
  • Loidi, J., Biurrun, I., Campos, J.A., García-Mijangos, I. & Herrera, M. 2010. A biogeographical analysis of the European Atlantic lowland heathlands. Journal of Vegetation Science 21: 832842.
  • Lomolino, M.V. & Weiser, M.D. 2001. Towards a more general species–area relationship: diversity on all islands, great and small. Journal of Biogeography 28: 431445.
  • MacArthur, R.H. & Wilson, E.O. 1963. Equilibrium theory of insular zoogeography. Evolution 17: 373387.
  • MacDougall, A.S. & Turkington, R. 2005. Are invasive species the drivers or passengers of change in degraded ecosystems? Ecology 86: 4255.
  • Maskell, L.C., Bullock, J.M., Smart, S.M., Thompson, K. & Hulme, P.E. 2006. The distribution and habitat associations of non-native plant species in urban riparian habitats. Journal of Vegetation Science 17: 499508.
  • Millennium Ecosystem Assessment 2005. Ecosystems and human well-being: biodiversity synthesis. World Resources Institute, Washington, DC, US.
  • Moen, A. 1998. Nasjonalatlas for Norge: Vegetasjon. Statens Kartverk, Hønefoss, NO.
  • Niering, W.A. 1963. Terrestrial ecology of Kapinggamarangi atoll, Caroline Islands. Ecological Monographs 33: 131138.
  • Norderhaug, A., Austad, I., Hauge, L. & Kvamme, M. 1999. Skjøtselsboka for kulturlandskap og gamle norske kulturmarker. Landbruksforlaget, Oslo, NO.
  • Okland, T., Bakkestuen, V., Okland, R.H. & Eilertsen, O. 2004. Changes in forest understorey vegetation in Norway related to long-term soil acidification and climatic change. Journal of Vegetation Science 15: 437448.
  • Oyen, B.H., Blom, H.H., Gjerde, I., Myking, T., Saetersdal, M. & Thunes, K.H. 2006. Ecology, history and silviculture of Scots pine (Pinus sylvestris L.) in western Norway – a literature review. Forestry 79: 319329.
  • Pauchard, A. & Shea, K. 2006. Integrating the study of non-native plant invasions across spatial scales. Biological Invasions 8: 399413.
  • Peltzer, D.A., Wardle, D.A., Allison, V.J., Baisden, W.T., Bardgett, R.D., Chadwick, O.A., Condron, L.M., Parfitt, R.L., Porder, S., Richardson, S.J., Turner, B.L., Vitousek, P.M., Walker, J. & Walker, L.R. 2010. Understanding ecosystem retrogression. Ecological Monographs 80: 509529.
  • Perrings, C., Dehnen-Schmutz, K., Touza, J. & Williamson, M. 2005. How to manage biological invasions under globalization. Trends in Ecology & Evolution 20: 212215.
  • Pimentel, D., Zuniga, R. & Morrison, D. 2001. Update on the environmental and economic costs associated with alien-invasive species in the United States. Ecological Economics 52: 273288.
  • Powell, K.I., Chase, J.M. & Knight, T.M. 2011. A synthesis of plant invasion effects on biodiversity across spatial scales. American Journal of Botany 98: 539548.
  • Pretto, F., Celesti-Grapow, L., Carli, E. & Blasi, C. 2010. Influence of past land use and current human disturbance on non-native plant species on small Italian islands. Plant Ecology 210: 225239.
  • Rangel, T.F., Diniz, J.A.F. & Bini, L.M. 2010. SAM: a comprehensive application for spatial analysis in macroecology. Ecography 33: 4650.
  • Richardson, D.M. 2006. Pinus: a model group for unlocking the secrets of alien plant invasions? Preslia 78: 375388.
  • Richardson, D.M. & Rejmanek, M. 2004. Conifers as invasive aliens: a global survey and predictive framework. Diversity and Distributions 10: 321331.
  • Rosenzweig, M. 1993. Species diversity in space and time. Cambridge University Press, Cambridge, UK.
  • Rosenzweig, M.L. 2001. The four questions: what does the introduction of exotic species do to diversity? Evolutionary Ecology Research 3: 361367.
  • Sax, D.F., Gaines, S.D. & Brown, J.H. 2002. Species invasions exceed extinctions on islands worldwide: a comparative study of plants and birds. The American Naturalist 160: 766783.
  • Schlaepfer, M.A., Sax, D.F. & Olden, J.D. 2011. The potential conservation value of non-native species. Conservation Biology 25: 428437.
  • Selmi, S. & Boulinier, T. 2001. Ecological biogeography of Southern Ocean islands: the importance of considering spatial issues. The American Naturalist 158: 426437.
  • Shea, K. & Chesson, P. 2002. Community ecology theory as a framework for biological invasions. Trends in Ecology & Evolution 17: 170176.
  • Slobodkin, L.B. 2001. The good, the bad and the reified. Evolutionary Ecology Research 3: 113.
  • Stohlgren, T.J., Jarnevich, C., Chong, G.W. & Evangelista, P.H. 2006. Scale and plant invasions: a theory of biotic acceptance. Preslia 78: 405426.
  • Svenning, J.C. & Skov, F. 2007. Could the tree diversity pattern in Europe be generated by postglacial dispersal limitation? Ecology Letters 10: 453460.
  • Tjorve, E. & Tjorve, K.M.C. 2011. Subjecting the theory of the small island effect to Ockham's razor. Journal of Biogeography 38: 18361839.
  • Tjorve, E. & Turner, W.R. 2009. The importance of samples and isolates for species–area relationships. Ecography 32: 391400.
  • Triantis, K.A., Vardinoyannis, K., Tsolaki, E.P., Botsaris, I., Lika, K. & Mylonas, M. 2006. Re-approaching the small island effect. Journal of Biogeography 33: 914923.
  • Vetaas, O.R. 1997. The effect of canopy disturbance on species richness in a central Himalayan oak forest. Plant Ecology 132: 2938.
  • Wagner, H.H. & Fortin, M.J. 2005. Spatial analysis of landscapes: concepts and statistics. Ecology 86: 19751987.
  • Westman, W.E. 1983. Xeric Mediterranean-type shrubland associations of Alta and BajaCalifornia and the community continuum debate. Vegetatio 52: 319.
  • Whittaker, R.J. & Fernandez-Palacios, J.M. 2007. Island biogeography: ecology, evolution and conservation, 2nd ed. Oxford University Press, Oxford, UK.
  • Wilcove, D.S. & Chen, L.Y. 1998. Management costs for endangered species. Conservation Biology 12: 14051407.
  • Young, A.M. & Larson, B.M.H. 2011. Clarifying debates in invasion biology: a survey of invasion biologists. Environmental Research 111: 893898.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusions
  8. Acknowledgements
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
jvs12045-sup-0001-AppendixS1.txtplain text document169KAppendix S1. List of species that occur on more than two islands.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.