Island biogeography of native and alien plant species: Contrasting drivers of diversity across the Lesser Antilles

Understanding the factors driving the diversity of alien and native species on islands is crucial for predicting the spread of alien species and for proposing management practices to protect the unique native biodiversity that often occurs in insular ecosystems. The main objective of this study was to evaluate whether native and alien plant species respond similarly to natural biogeographic and human‐related drivers.

Globalization and the intensification of international trade and travel have been transgressing the natural barriers to dispersal (Levine & D'Antonio, 2003;Hulme, 2009) and are leading to an unprecedented increase in the rates of introduction and naturalization of alien species (Seebens et al., 2017). Ultimately, this may cause the breakdown of the "classical" biogeographic theory and the emergence of new biogeographic arrangements determined primarily by environmental factors and by patterns in human activity (Burns, 2015;Capinha et al., 2015;Dawson et al., 2017;Essl et al., 2019). On islands, the introduction of alien species circumvents natural geographical barriers (isolation), which may lead to increases in the numbers and distribution of naturalized species. These events can potentially amplify negative impacts on native species with important implications for conservation of insular ecosystems (Moser et al., 2018;Russell, Meyer, Holmes, & Pagad, 2017;Sax & Gaines, 2008;Simberloff, 2000). In fact, previous studies have shown that alien diversity on islands is strongly and positively influenced by human-related factors, such as economy and market size, human population density and exchange and transportation rates (Blackburn et al., 2016;Dawson et al., 2017;Denslow, Space, & Thomas, 2009;Kueffer et al., 2010;Rojas-Sandoval, Tremblay, Acevedo-Rodríguez, & Díaz-Soltero, 2017).
Globally, island ecosystems are under severe threat due to habitat loss and climate change and are areas particularly susceptible to biological invasions (Caujapé-Castells et al., 2010;Denslow, 2003;Pyšek & Richardson, 2006;Rejmánek, 1996;Russell et al., 2017;Taylor & Kumar, 2016). Islands are recognized as hotspot of alien species (Dawson et al., 2017;Essl et al., 2019), and in the case of plants, the number of naturalized alien species on islands is almost twice the number of those recorded in similar sized patches on the mainland with approximately the same number of native species (Sax, Gaines, & Brown, 2002).
The Caribbean region is a hotspot of biodiversity with high priority for conservation due to its remarkable biological richness and high levels of endemism (>70% for Caribbean flowering plants; Acevedo-Rodríguez & Strong, 2012;Myers, Mittermeier, Mittermeier, Da Fonseca, & Kent, 2000;Roncal, Nieto-Blázquez, Cardona, & Bacon, 2020;Santiago-Valentin & Olmstead, 2004). Historically, Caribbean islands have played a crucial role as crossroads between continental America, Europe and Africa, resulting in multiple introductions of alien species over centuries as well as changes in land use and biota composition (Lugo, 2004;Rojas-Sandoval et al., 2017). Additionally, the flora of the Lesser Antilles is among the best documented for tropical archipelagos; comprehensive and updated botanical inventories are available providing the information needed for this study. Owing to these unique circumstances, the Lesser Antilles is an excellent place to study the biogeography of alien and native plant species and to evaluate whether alien species follow the same biogeographic patterns as native species, and under which circumstances human-related activities could lead to a failure of the traditional biogeographic rules.
In this study, we address the following questions: (1) What is the relationship between native and alien species richness in the Lesser Antilles, (2) what are the biogeographic and human-related predictor variables driving native and alien plant species richness on these islands and (3) are these variables exerting the same effects on native species as for alien species? To address these questions, we compiled a dataset with the verified status of native and alien plant species occurring on 15 islands across the Lesser Antilles. We hypothesized that if natural biogeographic variables exert the same effects on alien species as they do on native species (Whittaker & Fernandez-Palacios, 2007, Whittaker, Fernández-Palacios, Matthews, Borregaard, & Triantis, 2017, then we would expect biogeographic and biotic variables to be dominant drivers of both alien and native species richness on these islands (e.g. alien species would be more frequent and abundant on large islands as these can support a higher number of species). On the other hand, if the effects of human activities overshadow natural processes, then we would expect the alien species richness to be dominated by human-related variables (e.g. alien species would be more frequent and abundant on islands with larger human populations and higher economic development).

| Study site
We used plant diversity of native and alien species from 15 islands in the Lesser Antilles ranging from Anguilla in the north TA B L E 1 Summary of the biogeographic, socio-economic and historical variables compiled for 15 islands in the Lesser Antilles Percentage of total land area cover by forest. Forest area is land spanning more than 0.5 hectare with trees higher than 5m and a canopy cover of more than 10% to include windbreaks, shelterbelts and corridors of trees greater than 0.5 hectare and at least 20 m wide. to Grenada in the south ( Figure S1). These 15 islands may be divided into two major groups based on their geological histories and elevation. The first group comprises an inner arc of islands  (Bouysse, 1984;Donnelly, 1989;Joseph, 2012Joseph, , 2013Ricklefs & Lovette, 1999

| Data collection
We compiled a dataset with information on the presence/occurrence of plant species (including only spermatophyte species) for each of the 15 islands included in this study (see Appendix S1 in Supporting Information for a list of the references consulted).
Each species was classified as either native or alien for each island.
For this study, only naturalized alien species were considered and species with unreliable records (e.g. listed as both native and alien) were excluded. We built a second dataset comprising 15 biogeographic, biotic and human-related factors (i.e. socio-economic and historical variables) of each island that we suspected to be drivers of plant diversity ( Table 1). Most of these data were obtained in May 2019 from the World Factbook (https://www.cia.gov/libra ry/publi catio ns/the-world -factb ook/) and some local and international websites (see Appendix S2 for a complete list of the sources consulted). Distances to the nearest mainland (hereafter "distance to mainland") were calculated using the Google Maps Distance Measurement Tool as the distance in kilometres from the centre of each island to the coastline of the nearest continental mainland (Weigelt & Kreft, 2013

| Statistical analyses
Generalized linear models (GLMs) were used to assess potential relationships between native and alien species richness and biogeographic, biotic and socio-economic variables. Given the high number of variables (Table 1) and the potential collinearity problems that this might cause on multiple regression models, we first performed pairwise correlations to identify potential drivers of plant diversity on these islands ( Figure S2). The variables evaluated as predictors were as follows: island area, elevation, distance to mainland, human population size, kilometres of paved roads, per capita GDP (hereafter GDP) and percentage of area covered by agriculture and forest (Table 1). Latitude and number of tourists were not included in subsequent analyses owing to their strong pairwise correlations with other less intercorrelated variables that are at least as relevant based on ecological theory and our hypotheses ( Figure S2). Predictor variables island area, human population size and kilometres of paved roads were log-transformed to reduce heteroscedasticity. As species richness is count data, we first tested different GLMs with Poisson and negative binomial families. However, we decided to use GLMs of the Gaussian family because they consistently outperformed GLMs with Poisson and negative binomial in terms of model fit and model diagnostics (Crawley, 2007;Weigelt & Kreft, 2013). We also inspected the homoscedasticity and normality of the residuals with Q-Q plots obtaining similar results (Quinn & Keough, 2002). GLMs were fitted using linear or second-order polynomials, and models including predictor variables with strong pairwise correlation (r > .7; Figure S2) were removed from the list of plausible models to ensure that collinearity did not affect our results (Dormann et al., 2013).
Models were ranked according to the corrected Akaike's information criterion (AICc), and the model with the lowest AICc is considered the best-fit model (Akaike, 1974;Sugiura, 1978). Those models with AICc differences < 2 relative to the best-fit model were also considered as plausible (Arnold, 2010

| RE SULTS
The final dataset comprises a total of 2,438 plant species of which 1,825 are native species and 613 (25%) are alien species. We found that alien and native species richness is strongly and positively correlated (Kendall correlation = 0.92; p < .001; Figure 1), but different variables explained their diversity patterns. Our combined results showed that across the Lesser Antilles, both native and alien species are not distributed evenly among islands. As expected, small islands are sustaining smaller numbers of both native and alien species, while large islands are sustaining larger numbers of native and alien species (Figure 1). For most islands, the number of native and alien species also varied with distance to the mainland (Figure 2a).
The highest species richness for both groups occurred at the middle point of the maximum distance from the continent. A significant reduction in the number of both native and alien species occurs as we move away from the mainland, with a more significant reduction in the number of native species on islands located further away from the continent (Figure 2a). We also detected a positive correlation between elevation and number of native and alien species, but the magnitude of the slope is steeper for native species than for alien species (Figure 2b). A similar pattern emerged for native species of the six most species-rich plant families (Orchidaceae, Poaceae, Fabaceae, Rubiaceae, Cyperaceae and Asteraceae), for which we detected a strong and significant increment in the number of native species as elevation is increased ( Figure S3).
Pairwise correlation analyses showed strong correlations between island area, human population and kilometres of paved roads and between GDP, human population and kilometres of paved roads (r > .7 in all cases; Figure S2). Therefore, models including any pair of these correlated variables were removed from the GLM analyses to avoid collinearity (Dormann et al., 2013). For native species, the best-fit model (AICc < 2) includes three variables: island area, elevation and distance to mainland (Table 2; Appendix S3). The sum of the AIC weight across all plausible models showed that island area (AIC weight = 0.85) is the most important factor explaining native species richness followed by elevation (AIC weight = 0.71) and distance to the mainland (AIC weight = 0.69; Figure 3a). The relative importance of all the remaining variables was very low (AIC weight ≤ 0.2 in all cases), indicating that they have negligible effect on models predicting native species richness (Figure 3a). The best-fit model suggested that the number of native species increased with island area and elevation and mostly decreased as we moved away from the mainland ( Figure 2). Overall, our combined models showed that native species richness is strongly driven by biogeographic variables, whereas biotic and socio-economic variables play a minor role (Figure 3a).
For alien species, the AICc ranking identified three equivalent models (differences in AICc < 2) that are good predictors of alien species richness (

| D ISCUSS I ON
Native and alien species richness of the Lesser Antilles is positively correlated, meaning that islands richer in native species are also hosting more alien species. This positive relationship between native and alien species diversity is counter to Elton's biotic resistance hypothesis (Elton, 1958), which proposes that areas of high species richness should be resistant to biological invasions and assumes that niches are occupied, and native species are better competitors (Elton, 1958).

F I G U R E 2
Relationship between species richness and (a) distance to the mainland, (b) elevation, (c) island area and (d) GDP for 15 islands across the Lesser Antilles. Solid lines indicate the predicted GLM relationships for native plant species (green) and alien plant species (red) using first-or second-order polynomials. The grey areas correspond to the 95% confidence intervals of the model

While this may be true at very small scales and certain vegetation
types, it is difficult to find evidence for it in scales ranging from forest plots to regions Kennedy et al., 2002;Stohlgren, Barnett, & Kartesz, 2003;Stohlgren, Jarnevich, Chong, & Evangelista, 2006). Nevertheless, local areas of high species richness within islands, such as forest reserves, can show a weaker relationship between native and alien species richness suggesting that somewhat permeable biotic resistance may occur . For the Lesser Antilles, this strong positive correlation detected between native and alien species richness implies that native diversity could be used as a good predictor for alien species richness. This correlation has been previously explained in terms of ecological processes including species interactions and comparable responses of both native and aliens to the same drivers, suggesting that ecological drivers that favour high number of native species may also increase niche opportunities for alien species (Fridley et al., 2007;Lonsdale, 1999;Sax et al., 2002;Stohlgren et al., 2003).
We also found that, as expected, larger islands are hosting more native species than smaller islands. This positive correlation observed between native diversity and island area is one of the core relationships proposed by the theory of island biogeography (MacArthur & Wilson, 1967;Whittaker & Fernandez-Palacios, 2007). However, we also detected that this relationship is not restricted to native species, and the number of alien species also increases with island area.
The theory predicts higher species diversity on large islands relative to small islands because populations are likely to be larger on big islands and thus less prone to stochastic extinction than on small islands (MacArthur & Wilson, 1967). Large islands are also more likely to have higher diversity of habitats than small islands, and like area, habitat heterogeneity is often positively correlated with species diversity (Hortal, Triantis, Meiri, Thébault, & Sfenthourakis, 2009; Patiño F I G U R E 3 Sum of the AIC weight for the predictor variables used to explain the richness of (a) native and (b) alien plant species in the Lesser Antilles. The best-fit modes are provided in Table 2, and the full outputs for all the GLMs evaluated are included in Appendix S3 (native species) and S4 (alien species). Asterisks (*) indicate predictor variables with negative effect on species richness et al., 2013; Ricklefs & Lovette, 1999). The habitat heterogeneity hypothesis proposes that an increase in the number of habitats across a landscape leads to an increase in native species diversity because the number of partitionable niche dimensions expands (Crane & Willig, 2005). Consequently, large islands will not only have more available habitats for the evolution of native species but also have more habitats that may be colonized by alien species resulting in a higher number of both native and alien species (Jarnevich, Stohlgren, Barnett, & Kartesz, 2006). In the Lesser Antilles, islands with high topographic relief also have a broader range of temperatures and rainfall, and often support greater range of habitats and communities than islands with low relief (Ricklefs & Lovette, 1999;Ackerman, Trejo-Torres, & Crespo-Chuy, 2007;Joseph 2012;Joseph, 2013;Traxmandlová et al., 2017;Debrot, Madden, Becking, Rojer, & Miller, 2020). Not surprisingly, our data show that large islands with wide elevation ranges such as Guadeloupe and Martinique are hosting more species than small low islands such as Anguilla and St Martin.
The relationship between species richness in the Lesser Antilles and distance to mainland is not always as theory would predict, rather it can be taxon-dependent (Scott, 1972;Staats & Schall, 1996;Rodríguez-Durán & Kunz, 2001;Ricklefs & Bermingham, 2004). The largest islands of the Lesser Antilles are those in the middle of the island chain so that species richness actually increased with distance from continental mainland, and then dramatically dropped off in the northernmost part of the archipelago, where the islands all tend to be small. We therefore cannot clearly determine whether the decline in native species numbers in the northernmost part is due to reduced habitat diversity or distance to mainland; we suspect the former as species richness is related to island area (thus habitat diversity), and Guadeloupe, Martinique and Dominica are the largest islands and have the wider elevation ranges of the 15 islands included in our study. As was expected for alien species richness, distance to the mainland was less important than human-related activities (see below).
Historical colonial identity appears to have a legacy, but not with respect to species richness. Rather, it is associated with the species composition of alien floras in the Lesser Antilles. The clusters formed by the French (Martinique and Guadeloupe) and Dutch islands (Saba, St Eustatius and St Martin) imply that for these islands the composition of alien floras could be explained (at least in part) by their colonial identity and the political, cultural and economic connections that remain today (Fortune, 1984;Hall, 1982). These islands are dependencies of France and the Netherlands, meaning a closer connection, higher commercial and population exchange, and greater interdependence with these European countries than with other countries within the Caribbean region (Girvan, 2012;McElroy & De Albuquerque, 1995). The historical connection with these European countries and preferential trade with islands within the region sharing the same colonial identity appear to be influencing the pathways of introduction and producing unique clusters of alien species.
In agreement with the theory of island biogeography, our GLM analyses showed that biogeographic variables including island area, elevation (surrogate for habitat and niche diversity) and distance to the mainland were the most important drivers of native species richness. In contrast, our models indicated that socio-economic variables played a key role in driving alien species richness. Our findings show that alien species are more frequent in large islands with greater socio-economic development. Island area influences the number of people that inhabit islands and consequently also the number of human-mediated introductions. Because aliens are defined as species introduced as a consequence of human activities, large and more populated islands are more likely to host a larger number of alien species than small, less populated islands. Therefore, larger islands with substantial human populations will not only generate more trade, economic activities and more transport interchange, but they will also have higher anthropogenic disturbance rates and thus more opportunities for accidental and deliberate introduction of alien species (Levine & D'Antonio, 2003;Catford et al., 2012;Hulme, 2009;Lockwood, Cassey, & Blackburn, 2009;Pyšek et al., 2010).  First, there appears to be a strong conflict between human activities and the preservation of native plant diversity on these islands. Islands of the Lesser Antilles are exposed to huge residential and tourism pressures, and these activities are now primarily responsible for the reduction in forest cover and habitat degradation, but they are also the most relevant for the local economy of most islands (Dixon et al., 2001). More suitable and sustainable practices, as well as control in the expansion of urban and tourism activity, should be emphasized (Weaver, 2001). The preservation of areas of natural forests on these islands seems to be crucial not only for the conservation of their native diversity, but also for ecosystem services (e.g.

| Conservation perspective
water supply) most valued by humans (Anadón-Irizarry et al., 2012;Weaver, 2002). On the other hand, the remarkable importance of socio-economic variables driving alien plant richness observed in our results clearly emphasizes the vulnerability of large islands with high economic activity to the introduction and potential invasion of alien species. This link seems to be important to predict future trajectories of plant invasions on these islands, but also an opportunity to mitigate them. Finally, the implementation of more effective policies and management strategies to control the introduction and expansion of alien species should be a priority on these islands.

| Conclusions
As

ACK N OWLED G EM ENTS
We thank Pedro Acevedo-Rodríguez for help and advice on compiling the dataset and Ahme Cruz for help in the organization of hypotheses in the initial phases of this study. Support for this research was provided by the Smithsonian National Museum of Natural History and the U.S. Department of Agriculture.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13139.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used in this study can be found in Table 1.