Invasive plant species in the West Indies: geographical, ecological, and floristic insights

Abstract The level of invasion (number or proportion of invasive species) in a given area depends on features of the invaded community, propagule pressure, and climate. In this study, we assess the invasive flora of nine islands in the West Indies to identify invasion patterns and evaluate whether invasive species diversity is related to geographical, ecological, and socioeconomic factors. We compiled a database of invasive plant species including information on their taxonomy, origin, pathways of introduction, habitats, and life history. This database was used to evaluate the similarity of invasive floras between islands and to identify invasion patterns at regional (West Indies) and local (island) scales. We found a total of 516 alien plant species that are invasive on at least one of the nine islands studied, with between 24 to 306 invasive species per island. The invasive flora on these islands includes a wide range of taxonomic groups, life forms, and habitats. We detected low similarity in invasive species diversity between islands, with most invasive species (>60%) occurring on a single island and 6% occurring on at least five islands. To assess the importance of different models in predicting patterns of invasive species diversity among islands, we used generalized linear models. Our analyses revealed that invasive species diversity was well predicted by a combination of island area and economic development (gross domestic product per capita and kilometers of paved roadways). Our results provide strong evidence for the roles of geographical, ecological, and socioeconomic factors in determining the distribution and spread of invasive species on these islands. Anthropogenic disturbance and economic development seem to be the major drivers facilitating the spread and predominance of invasive species over native species.

Islands provide exceptional systems for testing hypotheses on biological invasions (Denslow, Space, & Thomas, 2009;Kueffer et al., 2010;Simberloff, 2000). Island communities have been regarded as more vulnerable to biological invasions than continental ecosystems due to their relatively higher numbers and densities of alien species (D'Antonio & Dudley, 1995;Denslow, 2003;Pyšek & Richardson, 2006;Sax, Gaines, & Brown, 2002). Nevertheless, this apparent vulnerability of islands to biological invasion, which has been considered as one of the "classic" generalizations in invasion ecology, remains debatable. The high rates of invasibility and naturalization of alien plant species observed on temperate and tropical islands have been traditionally explained by inherent features of islands such as their geographical and historical isolation, high habitat diversity, disharmonic and unsaturated floras, low biotic resistance, and low pest pressures, which are factors that would provide limited barriers to the establishment and naturalization of alien plants (D'Antonio & Dudley, 1995;Denslow, 2003;Mack et al., 2000;Simberloff, 1995). However, these assumptions have been rarely supported by empirical data and other insular features such as small territories, reduced habitats, and small population sizes have been proposed as alternative explanations (Simberloff, 1995(Simberloff, , 2000. Although the susceptibility to biological invasion does not necessarily coincide with the vulnerability to the deleterious effects caused by the spread of alien species, there is a consensus that islands are more susceptible to the effects of invasive species than continental ecosystems (D'Antonio & Dudley, 1995;Pyšek et al., 2012;Reaser et al., 2007). Insular biotas are considered to be fragile ecosystems characterized by high levels of endemism (Acevedo-Rodríguez & Strong, 2008;Whittaker & Fernández-Palacios, 2008). On these biotas, the introduction of alien species could lead to changes in community structure and the loss of native species, ecosystem functions, and services (Reaser et al., 2007;Rojas-Sandoval, Meléndez-Ackerman, & Anglés-Alcázar, 2016;Spotswood, Meyer, & Bartolome, 2012). In many cases, these changes lead to taxonomic and functional simplification, a phenomenon known as biotic homogenization, which has been extensively investigated for oceanic islands (Castro et al., 2010;Castro & Jaksic, 2008;Stohlgren et al., 2011;Vergara, Pizarro, & Castro, 2011).
The West Indies, comprising the Bahamas, the Greater and Lesser Antilles, and the islands off the coast of northern South America (Netherlands Antilles and Trinidad and Tobago), is considered one of the biodiversity hotspots with high global priority for conservation due to its biological richness and high levels of endemism (>70% for seed plants; Myers, Mittermeier, Mittermeier, da Fonseca, & Kent, 2000;Mittermeier, Myers, Thomsen, da Fonseca, & Olivieri, 1998;Acevedo-Rodríguez & Strong, 2012). However, in spite of such diversity, islands in the West Indies have been subjected to extreme levels of anthropogenic disturbance and massive degradation of their natural environments since the beginning of European colonization (Grove, 1996;Lugo, 2004;Maunder et al., 2008). Currently, the expansion of urban areas and tourism complexes coupled with high population densities and accelerated changes in socioeconomic conditions are generating unprecedented pressure over the natural resources on these islands (Dixon, Hamilton, Pagiola, & Segnestam, 2001).
In this study, we used data on invasive plant species collected on nine island groups to identify patterns of invasion across the West Indies. Although several studies and checklists of alien and invasive plants have been produced for some of these islands (Kairo, Ali, Cheesman, Haysom, & Murphy, 2003;Oviedo et al., 2012;Rojas-Sandoval & Acevedo-Rodríguez, 2015;Van der Burg, Freitas, Debrot, & Lotz, 2012), there have been no systematic studies comparing the patterns of invasion at the regional scale. Moreover, assessments of factors influencing invasibility and invasiveness are still limited for the West Indies (but see Ackerman et al., 2014;Ackerman, Tremblay, Rojas-Sandoval, & Hernández-Figueroa, 2017;Rojas-Sandoval & Acevedo-Rodríguez, 2015). Within this context, the main goals of this work were to: (1) describe the invasive flora of nine island groups across the West Indies, (2) evaluate the similarity of the invasive floras between islands, and (3) identify and evaluate geographical, ecological, and socioeconomic factors that best correlate with the invasive species diversity on these islands. The results derived from this study provide a general perspective of the patterns of distribution of invasive species within the Caribbean region and may serve as a base for future comparative studies on plant invasions and the design of management strategies to control and mitigate the impact of invasive plants on these islands.

| Study site
We collected information on the presence of invasive plant species for the following nine island groups and/or political units: (1) Bahamas, (2) Cuba, (3) Dominican Republic, (4) Jamaica, (5)

| Data collection
Our database of invasive species was compiled from published literature including research articles, dissertations, and reports from local governments and conservation organizations (see Appendix S1 in Supporting Information for bibliographic references). We only included alien spermatophyte species classified as "invasive" on the original sources consulted. These sources followed standardized criteria (e.g., Beck et al., 2008;Richardson et al., 2000) to classify species as invasive: (1) species are non-native, (2) they have established in new habitats without direct human assistance after their introduction, and (3) they are spreading over considerable areas causing negative (economic or environmental) impacts. We verified the invasive status and filtered the data by examining additional sources (specialist revisions, scientific publications, and herbarium collections). Species occurring exclusively in human-made habitats such as gardens and parks were excluded from the analyses. Similarly, species with unreliable records or uncertain origin (i.e., listed as both native and alien) were also excluded. The species entries were supplemented with data on their taxonomy, native distribution, life form (aquatic herb, grass, herb, shrub, succulent, tree, vine), life history (annual, biennial, perennial), type of introduction (intentional, unintentional, escaped), and planting purposes/uses (ornamental, agriculture/food crops, forage/ fodder, agroforestry, soil improver, timber tree) obtained from a list of data sources included in Appendix S2. Habitat categories were used to classify species occurrence in: agricultural habitats (traditional agriculture lands), ruderal (human-made habitats, excluding agriculture lands), seminatural (low to medium disturbed areas, including secondary forest), and mature forest (very low disturbed areas) (Lloret et al., 2005;Rojas-Sandoval & Acevedo-Rodríguez, 2015). Note that some species can occur in more than one habitat or can be assigned to several planting purposes/uses categories. Similarly, if the native distribution of a species covers more than one continent, it is assigned to each of them.

| Data analysis
To evaluate patterns of similarity on invasive species richness among islands, we calculated Sørensen coefficients as S ø = (2C/A + B)*100, where A and B are the number of invasive species found on island T A B L E 1 Geographical, ecological, and socioeconomic data for the nine island groups studied in the West Indies  A and B, respectively, and C is the number of invasive species common to both islands (McGarigal, 2000). The S ø coefficient varies between 0 (completely different) and 100 (identical). Additionally, to test if there was a correlation between similarity coefficients and the geographical distance among islands, we used a Mantel test comparing the matrix of S ø values for all island pairs and a matrix of corresponding geographical distances. The statistical significance of the resultant correlation was determined by a Monte Carlo simulation with 10,000 permutations (Legendre & Vaudor, 1991). Generalized linear models (GLM) were used to assess the relationships between invasive species diversity and geographical, socioeconomic, and environmental factors among islands. In these analyses, the factors evaluated as explanatory variables were the following: island area, elevation, forest cover, population density, gross domestic product (GDP) per capita, and kilometers of paved roadways (Table 1), while the number of invasive species per island and the ratio of invasive to native species were used as dependent variables. When the number of invasive species per island was evaluated as the dependent variable, we used GLM with a negative binomial model, but when the ratio of invasive to native species was the depended variable, we used GLM with a beta regression model. The beta regression method is an extension of the GLM and it is especially meant for modeling rates and proportions, as these data are non-normally distributed, while the negative binomial method is more appropriate for count data with a variance that is greater than the mean (McCullagh & Nelder, 1989). We started by creating a "full model" including all variables together, and then new models were created by simplifications of the full model to evaluate explanatory variables. As a result, a total of 64 models including all possible explanatory variable combinations were created for each dependent variable. All models evaluated are included in Appendices S3 and S4. Model selection was performed using the corrected Akaike's information criterion (AICc), a modification of the AIC that include a bias correction term for small sample size (Burnham & Anderson, 2002). In general, the model with the smallest AICc is considered the best model and AICc differences between 0 and 2 suggest weak differences among the models while differences larger than 4 are considered to be large (Arnold, 2010

| RESULTS
The final database comprises a total of 516 invasive plant species representing 348 genera and 96 families and contributing to 4% of the total flora of the West Indies (Table S1). This database comprises a heterogeneous group of plants in terms of their taxonomical and ecological attributes (Table S2). The families with the largest number of invasive species were Fabaceae (88 species), Poaceae (77 species), and Asteraceae (33 species), which accounted for 38% of all invasive species. A total of 34 families were represented by a single species and 78 families were represented by five species or less. The genera with the highest number of invasive species were Ipomoea (10 species), Paspalum (7 species), and Eragrostis (6 species), while 252 genera (72%) were represented by a single invasive species (Table S2).
The frequency distribution of invasive plant species across islands  (Table S3).
At the island scale, invasive species showed considerable variation in diversity ( Figure 2; Table S1). The island with the lowest number  Table S1).
F I G U R E 1 Frequency distribution of occurrences of invasive plant species across nine islands in the West Indies. The dot track represents the proportion relative to the total number of invasive plant species for all islands pooled (total = 516 species)

| Factors influencing invasive species diversity
Our GLM analyses for the number of invasive species as the dependent variable showed that the model including all factors explained 36% of the variation (Table 4; Fig. S1). However, the AICc analyses  AIC weight = 0.04), but it was less well supported than the two best-fit models and the difference in AICc was larger than 4 (Table 4).
Using the ratio of invasive to native species as the dependent variable, the GLM analyses showed that the model including all factors explained 75% of the variation (Table 4; Fig. S1). Nevertheless, from all 64 models generated (Appendix S4), the simplest best-fit (most parsimonious) model (AICc = −32.56; AIC weight = 0.39; Figure 5) included only one variable: the GDP. The next best models included human population density (AICc = −28.80; AIC weight 0.06) and a combination of island area and GDP (AICc = −28.35; AIC weight = 0.05) but both were less well supported than the best-fit model (Table 4). Overall, these results suggest that the number of invasive species increases consistently with island size and that paved roadways (an index of economic development) are facilitating the entrance and movement of invasive species on these islands. On the other hand, the ratio of invasive to native species increases with economic development and human population density, suggesting that human activities are favoring the establishment of invaders over native species.

| DISCUSSION
We have identified a total of 516 alien plant species that are invasive on at least one of the nine islands studied. habitats while our study included natural terrestrial and aquatic habitats, but also ruderal and semi-natural habitats. Indeed, for the nine islands included in our study, the majority of invasive species occur (but not exclusively) on ruderal and disturbed habitats. Our results show that the proportion of alien species that have become invasive, at both regional and local scales, is also higher than the corresponding value reported for other islands: This proportion ranges from 6% to 54% on selected islands across the West Indies (this study), while the values reported for other islands are generally lower and vary from 3% in the Juan Fernandez Archipelago (Castro & Jaksic, 2008) and 5% on Hawaii Islands, to 11% in Fiji (Meyer, 2014) and 12% on Galapagos Islands (Trueman, Atkinson, Guézou, & Wurm, 2010).
The high number of invasive species as well as the high proportion of alien species that have become invasive in the West Indies is likely a combination of (1) the historic role of these islands as a hub for the Europe-America trading routes and (2) the early and widespread degradation of their natural habitats. For more than three centuries (1500s-1890s), European colonizers used the Caribbean islands as ports of call for trading routes between the Americas and the "Old World," facilitating intentional and accidental introductions of numerous alien species coming from Asia, Africa, Europe, and continental America (Oviedo et al., 2012;Rojas-Sandoval & Acevedo-Rodríguez, 2015). Additionally, Caribbean islands have been subjected to high levels of anthropogenic disturbance first due to agriculture expansion and exploitation of their natural resources (e.g., logging), and more recently F I G U R E 3 Donor regions of the invasive plant species in the West Indies. The percentages in the charts represent the contribution of each region to the total number of invasive species on each island and for all islands pooled. The number in parentheses denotes the number of species native to each region. Species with native distribution in more than one continent were assigned to each of the continents 19% 30% 5% 10% 25%

11%
Africa (151) Asia (236) Europe (43) Oceania (77) America (202) West Indies (85 2008;Dixon et al., 2001). Several studies have demonstrated that the level of disturbance (i.e., habitat quality) plays a crucial role in determining the level of invasion in invaded communities and that alien species are more likely to establish in disturbed habitats than in pristine habitats (Catford et al., 2011;Lockwood et al., 2005;Rojas-Sandoval & Acevedo-Rodríguez, 2015). In the case of the West Indies, islands are not only exposed to high levels of anthropogenic disturbances but also recurrently impacted by large-scale natural disturbances such as hurricanes and tropical storms, which may drive changes in the structure, composition, and functioning of plant communities (Bellingham, Tanner, & Healey, 2005;Lugo, 2008;Walker, Lodge, & Waide, 1991).
Consequently, it is plausible that the high disturbance rates prevailing on these islands are facilitating the successful establishment of invasive species populations (Lockwood et al., 2005;Richardson & Pyšek, 2012).
T A B L E 4 Coefficients estimated from the generalized linear models. Coefficients for the number of invasive species as the dependent variable were estimated with negative binomial models. Coefficients for the ratio of invasive to native species as the dependent variable were estimated with beta regression models.   Daehler, 1998;Pyšek, 1998) and they are the two most diverse families in the alien and invasive floras of other insular systems Kueffer et al., 2010;Silva & Smith, 2004). On Caribbean Islands, most Fabaceae and Poaceae invaders are species that have escaped from active and abandoned pastures and agroforestry systems (introduced primarily as nitrogen fixers, soil improvers, forages, and fodders) and ornamentals escaped from cultivation (Oviedo et al., 2012;Rojas-Sandoval & Acevedo-Rodríguez, 2015;Van der Burg et al., 2012). In fact, our results showed that most invasive plant species in the West Indies (75%) have escaped from cultivation. We also found that more than 50% of plant invaders were introduced as ornamentals on these islands. Ornamental trade has been recognized as the main pathway for plant invasions worldwide (Dehnen-Schmutz, Touza, Perrings, & Williamson, 2007;Hulme, 2011) and has been shown to be a crucial factor explaining the establishment, naturalization, and subsequent invasion of alien plant species on Caribbean islands (Rojas-Sandoval & Acevedo-Rodríguez, 2015). The recurrent introduction of certain species associated with ornamental trade, directly linked to economic development (see below), may increase propagule pressure and the establishment probability of invasive species (Lockwood, Cassey, & Blackburn, 2009;Mack & Lonsdale, 2001;Pemberton & Liu, 2009;Reichard & White, 2001).
We found that the total length of paved roadways is indeed a good predictor of invasive species diversity for islands in the West Indies.
Previous studies have also shown that paved roadways strongly favor the spread and establishment of opportunistic invasive plant species (Mortensen et al., 2009;Jolly et al., 2011;Meunier & Lavoie, 2012

| CONCLUSIONS
Invasive species richness among islands in the West Indies is well predicted by a combination of island area and economic development including GDP per capita and kilometers of paved roadways. Low similarities in invasive species diversity between islands were detected, with most invasive species (>60%) occurring on a single island and only 6% occurring on at least five islands. Evidence suggests that species behaving as invasive in one area have higher probability to become invasive in other areas (Richardson & Pyšek, 2012). The dataset of invasive species presented here could thus be used as a blacklist of unwanted species to prevent their spread across the region and to avoid new introductions of likely invasive species. Overall, our data emphasize the need to limit deliberate introduction of alien species (particularly ornamentals) and reduce anthropogenic disturbance on these fragile ecosystems. Among the 36 plant species listed as the "world's worst invasive alien species" (Lowe, Browne, Boudjelas, & De Poorter, 2000), 19 species appeared in our dataset. Considering the extraordinary species richness and extent of endemism of the West Indies floras, the potential negative impact of these invasive species is very high (e.g., Rojas-Sandoval et al., 2016). Future studies should consider complete floristic inventories of alien species, including casual and naturalized species that could eventually become invasive.
These inventories are particularly relevant because they allow comparative regional analyses crucial for understanding invasion patterns.
From a management perspective, there is an urgent need to implement more effective prevention policies at all scales and enforce more stringent national and regional legislations.

ACKNOWLEDGMENTS
We thank J.D. Ackerman for suggestions that improved the manuscript and D. Anglés-Alcázar for inputs and discussions. Logistic support for this research was provided by the Smithsonian Institution (National Museum of Natural History) and the US Department of Agriculture.

AUTHOR CONTRIBUTIONS
JRS conceived and designed the study, compiled the database and prepared the data for analyses; JRS and RLT analyzed the data and created the figures. JRS wrote the manuscript including comments from all authors.