Human activity drives establishment, but not invasion, of non-native plants on islands
Abstract
Island ecosystems are particularly susceptible to the impacts of invasive species. Many rare and endangered species that are endemic to islands are negatively affected by invasions. Past studies have shown that the establishment of non-native species on islands is related to native plant richness, habitat heterogeneity, island age, human activity, and climate. However, it is unclear whether the factors promoting establishment (i.e. the formation of self-sustaining populations) also promote subsequent invasion (i.e. spread and negative impacts). Using data from 4308 non-native plant species across 46 islands and archipelagos globally, we examined which biogeographic characteristics influence established and invasive plant richness using generalized linear models nested within piecewise structural equation models. Our results indicate that anthropogenic land use (i.e. human modification) is strongly associated with establishment but not invasion, that climate (maximum monthly temperature) is strongly associated with invasion but not establishment, and that habitat heterogeneity (represented by maximum elevation and island area) is strongly associated with both establishment and invasion. Island isolation explains native plant richness well, but is not associated with established and invasive plant richness, likely due to anthropogenic introductions. We conclude that anthropogenic land use on islands is likely to be a proxy for the number of introductions (i.e. propagule pressure), which is more important for establishment than invasion. Conversely, islands with more diverse habitats and favorable (warm) climate conditions are likely to contain more available niche space (i.e. ‘vacant niches') which create opportunities for both establishment and invasion. By evaluating multiple stages of the invasion process, we differentiate between the biogeographic characteristics that influence plant establishment (which does not necessarily lead to ecological impacts) versus those that influence subsequent plant invasion (which does lead to negative impacts).
Introduction
A fundamental goal of invasion ecology is identifying the environmental characteristics that facilitate the success of species in non-native environments (Catford et al. 2011, Bellard et al. 2017). The characteristics that allow introduced species to establish (i.e. form self-sustaining populations in their non-native ranges) may differ from those that allow established species to invade (i.e. cause negative impacts; see Box 1 for a glossary of invasion terms). Hypotheses of invasibility have attempted to distinguish the habitat characteristics that promote success at each stage of the invasion process (Daly et al. 2023). These distinctions are crucial because most non-native species will not become invasive, meaning they require little to no formal management. Thus, to properly manage invasive species without expending limited resources on introducedandestablished species that will not become invasive, it is important to understand which habitat characteristics promote establishment only, and which habitat characteristics also promote invasion.
| Term | Definition |
|---|---|
| Non-native speciesa | Species that have been anthropogenically relocated to areas beyond their historical natural range. This definition inherently includes all species defined as ‘established' or ‘invasive,' as well as species that have been anthropogenically relocated to new areas but do not have self-sustaining populations or negative impacts. |
| Established species | Non-native species that have self-sustaining populations in their non-native ranges. This definition includes all species defined as ‘invasive' but does not necessarily include all species defined as ‘non-native.' This term is synonymous with ‘naturalized species.' |
| Invasive species | Non-native species that have self-sustaining populations in their non-native ranges and are also spreading or causing negative impacts in those locationsb. This definition of ‘invasive species' presented here is consistently applied by all sources of invasive species data used in the present analysis (Pagad et al. 2018, CABI 2022, Laginhas and Bradley 2022). |
| Introduction | The stage of the invasion process where individuals of a species arrive (following anthropogenic relocation) in areas beyond their historical natural range. All non-native species, by definition, have progressed to this stage. ‘Introduction' differs from ‘natural colonization' in that introduction is a human-centered process whereas natural colonization occurs through non-anthropogenic dispersal. |
| Establishment | The stage of the invasion process following ‘introduction,' where non-native species form self-sustaining populations in their non-native ranges. Habitat characteristics that influence the introduction stage inherently also influence the establishment stage since all established species must first be introduced. |
| Invasion | The stage of the invasion process following ‘establishment,' where established species cause negative impacts in their non-native rangesc. Habitat characteristics that influence introduction and establishment inherently also influence invasion since all invasive species must first be introduced and established. |
| Island susceptibility hypothesis (ISH) | The expectation that island habitats should be more vulnerable to invasion than mainland habitats. Since the invasion process requires a non-native species to survive introduction and establishment before becoming invasive, island susceptibility to invasion may be influenced by habitat characteristics that act on any of the three stages in the invasion process. |
The ‘island susceptibility hypothesis' (ISH) posits that island ecosystems are more vulnerable to invasion than mainlands with similar habitats. The ISH is frequently linked to biotic resistance; since islands have proportionally fewer native species than mainland habitats, they are expected to have weaker biotic resistance to non-native species (Elton 1958, Jeschke et al. 2018). However, other variables have also been hypothesized to underlie the ISH (e.g. island size, age and climate) and the exact mechanisms are not fully understood (Carlquist 1965, Ricklefs 1977, Davis et al. 2000). Disentangling the drivers of island susceptibility is particularly complex because different habitat characteristics may influence different stages of the invasion process (introduction, establishment, and invasion; see Box 1 for definitions) in different ways. For example, in the Anthropocene, island size may influence introduction since larger islands usually have more people, who are likely to introduce many non-native species (Chown et al. 1998). However, island size may also influence establishment, since larger islands tend to have greater abiotic heterogeneity than smaller islands, and therefore may contain a habitat suitable for establishment of each introduced species. Thus, although the ISH is relatively well-known by invasion ecologists (Enders et al. 2018), the underpinnings of this pattern are complicated and poorly understood.
Although the mechanisms for the ISH are still being determined, the conservation implications of island vulnerability are clear. Islands frequently harbor populations of rare and endemic species (Kier et al. 2009, Tershy et al. 2015, Steinbauer et al. 2016) that are often threatened by invasive species. Many islands now have more non-native plant species than native plant species (Essl et al. 2019), and the rate of introduction of new non-native species to islands is increasing (Seebens et al. 2017). Thus, improving invasive species management on islands is crucial for ecological conservation (Kueffer and Kinney 2017, Fernández-Palacios et al. 2021).
To date, most studies of island susceptibility to non-native plants have focused on established species rather than invasive species (Chown et al. 1998, Sax et al. 2002, Denslow et al. 2009, Long et al. 2009, Guo 2014, Blackburn et al. 2016, Moser et al. 2018, Irl et al. 2021). These previous studies have consistently found positive correlations between non-native plant richness and island area (Chown et al. 1998, Denslow et al. 2009, Moser et al. 2018), isolation (Moser et al. 2018, Essl et al. 2019), temperature (Chown et al. 1998, Blackburn et al. 2016), elevation (Denslow et al. 2009, Moser et al. 2018, Essl et al. 2019), human population size (Denslow et al. 2009, Blackburn et al. 2016), gross domestic product (GDP; Denslow et al. 2009, Kueffer et al. 2010, Moser et al. 2018), and native plant richness (Long et al. 2009, Blackburn et al. 2016). While some studies have observed contradictory relationships to those described above (Guo 2014, Blackburn et al. 2016), the biogeographic characteristics that influence establishment of non-native plants on islands have largely been well-defined.
Although it is useful to understand the drivers of non-native plant establishment on islands, invasive species are notably different in that they have negative environmental, economic, and cultural impacts (Pimentel et al. 2005, Pfeiffer and Ortiz 2007, Vilà et al. 2011, IPBES 2023), and the characteristics of islands related to invasion may be different from those related to establishment (Box 1). For example, propagule pressure likely increases establishment success of non-native plants (Lockwood et al. 2009, Catford et al. 2011) because it increases the likelihood that a given individual is introduced to a suitable habitat (Lockwood et al. 2009, Duncan et al. 2014). However, this mechanism may not promote invasion success in the same way that it promotes establishment. A few studies have evaluated characteristics that promote invasion, not just establishment, on islands (Kueffer et al. 2010, Essl et al. 2019), but this topic remains underrepresented in the literature. Perhaps because of this, existing results have been inconsistent. For example, Essl et al. (2019) found that island isolation had a strong impact on invasive plant richness, but Kueffer et al. (2010) found that isolation was not a significant predictor. Therefore, large discrepancies still remain in understanding the factors that influence invasibility on islands, demonstrating a clear knowledge gap for further investigation.
Here, we address the question, ‘which geographic, climatic, and anthropogenic covariates influence invasive plant richness on islands?' We evaluate a suite of 16 explanatory covariates to determine which characteristics of islands are most closely associated with increased invasive plant richness (Table 1). For comparison, we also evaluate the covariates that are associated with native plant richness and established plant richness on the same islands.
| Cat. | Covariate | Hypothesized relationship with established plant species richness | Hypothesized relationship with invasive plant species richness | Rationale | Key references |
|---|---|---|---|---|---|
| B | Native plant species richness* | positive | positive | Greater diversity of environments | Blackburn et al. 2016 |
| G | Area | positive | positive | MacArthur and Wilson 1967 | |
| G | Maximum elevation | positive | positive | Whittaker et al. 2008, Essl et al. 2019 | |
| G | Age | hump-shaped | hump-shaped | Environmental heterogeneity should peak at intermediate ages | Whittaker et al. 2008, Barajas-Barbosa et al. 2020 |
| G | Distance to nearest continent† | positive | positive | Anthropogenic activity ‘overrides' isolation; competitively weaker species on more isolated islands | Helmus et al. 2014, Moser et al. 2018, Essl et al. 2019 |
| G | Distance to nearest landmass*† | positive | positive | ||
| A | Mean human modification | positive | none | Greater propagule pressure | Lockwood et al. 2009, Catford et al. 2011, Essl et al. 2019 |
| A | Gross domestic product (GDP) per capita | positive | none | ||
| A | Human population | positive | none | ||
| A | Year of modern anthropogenic colonization | positive | none | ||
| C | Latitude of island centroid | negative | negative | High susceptibility in abiotically favorable environments (including the tropics) | Zefferman et al. 2015, Chong et al. 2021 |
| C | Maximum monthly temperature | positive | positive | ||
| C | Minimum monthly temperature* | negative | negative | ||
| C | Range of monthly temperatures | negative | negative | ||
| C | Coefficient of precipitation seasonality‡ | negative | negative | ||
| C | Total annual precipitation | positive | positive |
Hypotheses
The ‘Theory of Island Biogeography' posits that increases in habitat heterogeneity are likely to increase native species richness (MacArthur and Wilson 1967). Therefore, we expected island area and maximum elevation to be positively correlated with the number of native plant species (Ricklefs 1977, Hortal et al. 2009). We also expected these variables to be positively associated with established (Blackburn et al. 2016, Irl et al. 2021) and invasive plant richness (Kueffer et al. 2010), since greater habitat diversity means there are potentially suitable habitats for larger numbers of established and invasive species.
While habitat heterogeneity is almost always associated with increased species richness, the expected effects of island age on plant species richness are less obvious. Older islands with longer evolutionary histories may have higher numbers of competitively superior native plant species which are invasion resistant (as per the evolutionary imbalance hypothesis; Levine and D'Antonio 1999, Fridley and Sax 2014). However, phylogenetic diversity would be a more precise measure of the competitive strength of native flora instead of island age (Fridley and Sax 2014). Another more likely possibility is that the effects of island age on native, established, and invasive plant richness will be indirect, operating mainly through habitat heterogeneity. In this scenario, islands of intermediate age should have the highest heterogeneity and therefore the highest richness of established and invasive plants, with younger and older islands supporting fewer numbers of non-native plants (Whittaker et al. 2008, Barajas-Barbosa et al. 2020).
As described by the ‘Theory of island biogeography' (MacArthur and Wilson 1967), we expect that island isolation (distance to the nearest continent or landmass) will be negatively correlated with native plant richness. However, we expect island isolation will be positively correlated with established and invasive plant richness. Anthropogenic relocation of non-native species should not be as restricted by geographic isolation as natural dispersal of plants (as has already been demonstrated for other taxa; Helmus et al. 2014). Furthermore, once a non-native plant is introduced, it should be able to establish and invade more easily on more isolated islands, since the native flora on these islands should be weaker competitively than less-isolated islands (Denslow et al. 2009, Helmus et al. 2014, Moser et al. 2018).
We expect that covariates representing propagule pressure (mean human modification, per capita GDP, human population, and year of modern anthropogenic colonization) will be positively correlated with established plant richness, since more introductions usually result in higher likelihood of establishment (Lockwood et al. 2009, Catford et al. 2011). However, we do not expect higher propagule pressure to result in increased invasive plant richness. Previous studies have proposed that anthropogenic influences and propagule pressure mainly act on the introduction and establishment phases, but are less important for determining invasion (Williamson 2006, Daly et al. 2023).
Finally, we hypothesize that low-stress climate conditions (like those present in the tropics) will be associated with higher richness of both established and invasive plants (Table 1). More specifically, these low-stress climate conditions are: lower latitudes, warmer maximum and minimum monthly temperatures, smaller ranges of monthly temperatures, smaller coefficients of precipitation seasonality, and greater total annual precipitation. These conditions should promote establishment since climate suitability of the non-native range is an important driver of establishment success (Fristoe et al. 2023), and conditions that are consistent and broadly favorable should be suitable for a larger variety of introduced non-native plants (Zefferman et al. 2015). Additionally, we expect these climatic conditions to be associated with higher establishment and invasion because regions with these climate conditions (i.e. the tropics) often have high resource availability (Chong et al. 2021) that could become available to invasive plants through niche vacancies created by various global change factors including anthropogenic disturbance, climate change, and other invasive species (Simberloff and Von Holle 1999, Tilman and Lehman 2001).
Methods
Study area
We focused on islands and archipelagos that had richness of established and invasive plants recorded in Pfadenhauer and Bradley (2024; n = 66). We excluded islands that were geologically continental (i.e. islands that were previously connected to a larger continent; n = 20) according to ‘Wikipedia' and ‘Encyclopedia Britannica' (see the Supporting information for a list of excluded continental islands). We excluded geologically continental islands to create a more explicit test of colonization patterns on islands that have been separate from mainlands throughout geologic time. Previous studies of non-native plant establishment have similarly restricted their analyses to oceanic islands (Blackburn et al. 2016). We compiled geographic, climatic, and anthropogenic covariate data for the resulting list of islands and archipelagos (Fig. 1).
Map of study islands. We modeled native, established, and invasive plant species richness on 46 distinct islands and archipelagos. We only used islands that had established plant species richness and invasive plant species richness recorded in at least one of the following sources: the ‘Global Naturalized Alien Flora' database (van Kleunen et al. 2019), the ‘Centre for Agriculture and Bioscience International Invasive Species Compendium' (CABI 2022), the ‘Global Plant Invaders' database (Laginhas and Bradley 2022), and the ‘Global Register of Introduced and Invasive Species' (Pagad et al. 2018). We only used islands of oceanic origin to remain consistent with the application of island biogeography theory.
Plant richness (response) variables
We compiled species richness for native, established, and invasive plants on each study island. We chose to use species richness as our metric of invasion instead of abundance (or another metric) since species presence data were available for the largest number of islands. To determine native plant richness on each island, we used the ‘World Checklist of Vascular Plants' (WCVP ver. 11; Govaerts et al. 2021). We summed species that were listed as native to each island. To obtain established and invasive plant richness on each island, we summed species listed as either established or invasive on each island according to Pfadenhauer and Bradley (2024). This method was island-specific; species that had negative impacts on one island but were established without having negative impacts on another island were listed as invasive on the former island and established on the latter. Occurrences of established and invasive plants within Pfadenhauer and Bradley (2024) were primarily sourced from the ‘Global Naturalized Alien Flora' database (GloNAF; van Kleunen et al. 2019), the ‘Centre for Agriculture and Bioscience International' (CABI) Invasive Species Compendium (CABI 2022), the ‘Global Plant Invaders' database (Laginhas and Bradley 2022), and the ‘Global Register of Introduced and Invasive Species' (GRIIS; Pagad et al. 2018). To ensure the reliability of our data, we confirmed that all sources used compatible definitions of established and invasive (Box 1). While there are still global reporting biases for invasive plant species (Laginhas et al. 2022) and discrepancies in the use of the terms ‘invasive' and ‘established' (Soto et al. 2024), our filtering and standardization steps resulted in a list of native, established, and invasive species with minimal classification uncertainty.
Island covariates
We compiled 16 covariates to characterize geographic, climatic, and anthropogenic conditions on each island (Table 1, Supporting information). Many of our study islands spanned multiple climates, so we categorized each one based on the Köppen climate with the greatest areal coverage (Kottek et al. 2006). By calculating single values for each covariate on each island or archipelago, we inevitably overlook some of the microhabitats and fine-scale variation in biogeography that is present. However, because our plant richness data are also reported at the scale of entire islands and archipelagos, this scale was also most appropriate for evaluating covariates.
Modeling plant richness
Modeling approach
Structural equation models (SEMs) are a family of modeling techniques that confirm the presence of relationships between large networks of variables by estimating covariances. Piecewise structural equation models (pSEMs) differ from traditional SEMs in that they estimate relationships for exogenous (explanatory) and endogenous (response) variables individually and then subsequently combine them, rather than estimating all covariances simultaneously (Shipley 2000, Lefcheck 2016). Because of this, pSEMs require less data to achieve proper fit when compared to traditional SEMs, and are able to model response variables with different underlying distributions without requiring transformations. Our plant richness variables were discrete counts (best fit using negative binomial distributions) and most of our covariates were continuous (best fit with Gaussian distributions), making pSEMs an ideal choice because of their flexibility. This approach also allowed us to differentiate between direct and indirect effects. Direct relationships control for the variation in other explanatory variables, whereas indirect effects include the variation of one or more mediating variables.
Model building
To determine the covariates that are most closely associated with plant richness on islands, we built pSEMs, each composed of multiple generalized linear models (GLMs) using the ‘piecewiseSEM' package in R (www.r-project.org, Lefcheck et al. 2023). We built three separate pSEMs (one for each response variable: native, established, and invasive plant richness) to test the effects of geographic, climatic, and anthropogenic covariates. Our processes of building, testing, and interpreting pSEMs followed the modeling framework presented in Blackburn et al. (2016).
For each of the three pSEMs, we started by building an a priori hypothesis model based on commonly reported relationships between response variables and covariates in the literature (Table 1 and Supporting information). We fit the hypothesized model using the psem() function and then manually removed covariates whose standardized estimates had bootstrapped confidence intervals that included zero (standardized estimates are converted to the same scale for comparison; unstandardized estimates retain the original units of each covariate, and therefore cannot be easily compared). We used parametric bootstrapping with 10000 resamples in the ‘semEff' package to generate standardized coefficients (as well as their respective 95% confidence intervals) for the direct and indirect effects of significant covariates on the response variables (Murphy 2022). We continued this process until all standardized coefficients did not overlap zero. We refer to these variables as ‘significant' although no p-values were used in their evaluation.
While fitting each of the three pSEMs, we avoided using combinations of covariates that were correlated (Pearson's r > 0.7) or combinations of covariates that were inherently similar because highly correlated covariates can distort coefficient estimates and variable selection (Dormann et al. 2013). In our dataset, we found three such pairs of covariates that fit this criteria: maximum elevation and minimum monthly temperature (Pearson's r = 0.73), human population and native plant richness (Pearson's r = 0.80), and distance to nearest continent and distance to nearest landmass (inherently similar). We ultimately used maximum elevation, human population, and distance to the nearest continent, since these variables explained more variation in the plant richness response variables than their highly-correlated counterparts.
Testing model fit
To ensure the reliability of model results, we tested the fit of each component generalized linear model (GLM) within each pSEM. For each GLM, we compared the AIC values for all nested models against that of the fitted model to ensure that the fitted model had the lowest value. We verified that GLMs for native, established, and invasive plant richness were not overdispersed, and that the results of residual simulations using the ‘DHARMa' package (Hartig and Lohse 2022) for these models did not significantly deviate from expectations. Finally, we calculated the proportion of deviance explained, and plotted residuals against fitted values.
Model interpretation and visualization
Once each pSEM and its component GLMs satisfied the model fit criteria, we compared the relative strength of explanatory variables within each pSEM using the standardized coefficients. We considered effects to be weak if the mean standardized coefficient was close to zero, whereas we considered covariates to have strong effects if the mean standardized coefficient was far from zero. We also extracted unstandardized coefficients for the direct effects, which we used to determine the expected change in each response variable resulting from a one-unit change in each covariate's original scale. To visualize these relationships, we generated added-variable plots for unstandardized coefficients using the ‘car' package (Fig. 3; Fox et al. 2023).
Results
Summary of islands dataset
Our final dataset included 46 study islands and archipelagos (Fig. 1). Nearly all of our 46 study sites consisted of more than one physical island, but we refer to the study sites collectively as ‘islands' throughout the manuscript for simplicity (even when ‘archipelagos' would also be appropriate). Islands ranged in size from 29 km2 (Nauru) to 940 000 km2 (Fiji) with a median of 1147 km2. Islands ranged in isolation from 27 km (Aruba) to 5229 km (Line Islands) to the nearest continent, with a median of 1002 km. Islands ranged in maximum elevation from 2 m a.s.l. (Maldives) to 4205 m a.s.l. (Hawaii) with a median of 923 m a.s.l.
A total of 4308 established plant species were present across all study islands. The Faroe Islands had the lowest reported established plant richness (12), while Hawaii had the highest (1573), with a median of 261. Invasive plant richness ranged from 1 (Ascension Island) to 411 (Hawaii), with a median of 38. Invasion rates (the percentages of established plant species that were identified as invasive) ranged from 0.6% (1/155; Ascension Island) to 96.4% (81/84; Comoros), with a median of 17%. South Georgia Island had the lowest reported native plant richness (48), while the Philippines had the highest (12231), with a median of 1462.
The majority of our study islands (n = 31) were primarily tropical (Köppen climate A), five islands were primarily arid (Köppen climate B), seven islands were primarily temperate (Köppen climate C), and three islands were primarily polar (Köppen climate E; Fig. 1). None of our study islands were classified as Köppen climate D (continental).
Plant richness models
Model fit
All of our component GLMs within the three pSEMs had the lowest AIC values out of all nested models (with all ΔAIC values > 6). The GLM that modeled native plant richness was slightly overdispersed but additional data simulations revealed that the extent of overdispersion was unlikely to affect covariate significance or coefficient estimates (see the Supporting information and code for data simulations on GitLab). Aside from this minor deviation, all models met criteria for proper fit.
Native plant richness
A combination of geographic and climatic covariates were the most closely associated with native plant richness (Fig. 2a). Native plant richness correlated positively with maximum monthly temperature (direct), maximum elevation (direct), and island area (indirect; Table 2). Native plant richness correlated negatively with distance to the nearest continent (direct), island age (indirect), and latitude (indirect). The covariates with the strongest overall effects on native plant richness were maximum elevation (total standardized coefficient [TSC] of 0.514), maximum monthly temperature (TSC = 0.346), and distance to continent (TSC = −0.343).
Resultant path diagrams for each piecewise structural equation model (pSEM) with standardized and unstandardized coefficients (in parentheses). Standardized coefficients were generated using parametric bootstrapping. Panels correspond to response variables, which are shown in black central ovals(a) native plant species richness on islands, (b) established plant species richness, and (c) invasive plant species richness. White ovals and boxes represent covariates. Boxes represent exogenous variables (i.e. variables with no explanatory covariates), ovals represent endogenous variables (i.e. variables with explanatory covariates). Black (solid) arrows indicate positive direct effects; red (dashed) arrows indicate negative direct effects. Indirect effects are not displayed, but they are described in Table 2.
| Endogenous variable ~ Exogenous variable | Direct or Indirect | Total standardized coefficient (bounds of 95% CI) |
|---|---|---|
| pSEM #1: Native plant species richness | ||
| Native richness ~ Elevation | direct | 0.514 (0.339, 0.686) |
| Native richness ~ Max. monthly temp. | direct | 0.346 (0.193, 0.750) |
| Native richness ~ Distance to continent | direct | −0.343 (−0.494, −0.232) |
| Native richness ~ Age | indirect | −0.263 (−0.405, −0.132) |
| Native richness ~ Latitude | indirect | −0.258 (−0.617, −0.121) |
| Native richness ~ Area | indirect | 0.154 (0.078, 0.256) |
| Elevation ~ Age | direct | −0.511 (−0.692, −0.233) |
| Elevation ~ Area | direct | 0.300 (0.161, 0.444) |
| Max. monthly temp. ~ Latitude | direct | −0.744 (−0.872, −0.411) |
| pSEM #2: Established plant species richness | ||
| Established richness ~ Elevation | direct | 0.427 (0.179, 0.651) |
| Established richness ~ Human modification | direct | 0.319 (0.098, 0.535) |
| Established richness ~ Age | indirect | −0.218 (−0.407, −0.076) |
| Established richness ~ Area | indirect | 0.128 (0.048., 0.248) |
| Elevation ~ Age | direct | −0.511 (−0.692, −0.233) |
| Elevation ~ Area | direct | 0.300 (0.161, 0.444) |
| pSEM #3: Invasive plant species richness | ||
| Invasive richness ~ Max. monthly temp. | direct | 0.445 (0.081, 0.724) |
| Invasive richness ~ Elevation | direct | 0.281 (0.085, 0.492) |
| Invasive richness ~ Latitude | indirect | −0.331 (−0.622, −0.039) |
| Invasive richness ~ Age | indirect | −0.144 (−0.294, −0.038) |
| Invasive richness ~ Area | indirect | 0.084 (0.025, 0.178) |
| Max. monthly temp. ~ Latitude | direct | −0.744 (−0.872, −0.411) |
| Elevation ~ Age | direct | −0.511 (−0.692, −0.233) |
| Elevation ~ Area | direct | 0.300 (0.161, 0.444) |
Established plant richness
Established plant richness correlated positively with maximum elevation (direct), human modification (direct), and island area (indirect; Fig. 2b). Established plant richness correlated negatively with island age (indirect). The covariates with the strongest overall effects on established plant richness (Fig. 3) were maximum elevation (TSC = 0.427) and human modification (TSC = 0.319).
Added-variable plots for covariates with significant direct effects in the piecewise structural equation models (pSEMs) that modeled established plant species richness (orange lines, (a) and (b)) and invasive plant species richness (red lines, (c) and (d)).
Invasive plant richness
Invasive plant richness correlated positively with maximum monthly temperature (direct), maximum elevation (direct), and area (indirect; Fig. 2c). Invasive plant richness correlated negatively with latitude (indirect) and island age (indirect). The covariates with the strongest overall effects on invasive plant richness were maximum monthly temperature (TSC = 0.445) and maximum elevation (TSC = 0.281). Temperature, elevation, area, latitude, and age represent the climatic and geographic categories of our covariates; none of the covariates from the anthropogenic category were significantly associated with invasive plant richness (Table 2).
Theory of island biogeography
Island area
Focusing specifically on pSEM results for measures of island biogeography, island area had a significant, indirect positive effect on native, established, and invasive plant richness (Fig. 2). In each model, the effect of island area was the weakest of any covariate in the final model (Table 2, Supporting information), and the effect of island area on plant richness was always mediated by maximum elevation.
Island isolation
Island isolation (distance to the nearest continent) had a significant, negative, direct effect on native plant richness (Fig. 2a). Distance to the nearest continent ranked third out of six covariates with significant effects on native plant richness (Table 2). Island isolation had no significant effect on established and invasive plant richness (Table 2).
Discussion
Understanding the factors that are associated with plant richness on islands has been a fundamental goal of ecology since the publication of the ‘Theory of island biogeography' (MacArthur and Wilson 1967). We examined the significance of geographic, climatic, and anthropogenic variables on native, established, and invasive plant richness on oceanic islands. By evaluating consecutive stages of the invasion process, we differentiated between the biogeographic characteristics associated with non-native plant establishmentversus those associated with non-native plant invasion. Notably, human modification was positively correlated with the establishment, but not invasion, of non-native plants on islands.
Revisiting hypotheses
Habitat heterogeneity
Habitat heterogeneity covariates were positively related to native, established and invasive plant richness (Fig. 2). Other studies have similarly found that habitat heterogeneity increases the diversity of both native and non-native species that an island can support (MacArthur and Wilson 1967, Sax and Gaines 2008, Essl et al. 2019) and also increases the likelihood that there will be vacant niches for invaders across a range of habitats (Simberloff 1995).
Maximum elevation was positively and directly associated with native, established, and invasive plant richness. Island area was also positively correlated with each type of plant richness, but in each pSEM, its effects were mediated by maximum elevation. This suggests that maximum elevation is a better proxy for habitat heterogeneity than island area. Several important abiotic variables have elevation-dependent relationships, including atmospheric pressure, temperature, humidity, and UV radiation (Körner 2007), leading to a higher likelihood of high habitat heterogeneity on islands with higher elevations than on larger-sized islands. These results are supported by other analyses which have also shown elevation to be a strong predictor of native and established plant richness on islands (Whittaker et al. 2008, Denslow et al. 2009, Kallimanis et al. 2010, Steinbauer et al. 2012).
Maximum elevation on islands also appears to be a better proxy for habitat heterogeneity than native plant richness. Previous studies have found positive relationships between native plant richness and established plant richness at broad spatial scales (Stohlgren et al. 2006, Sax and Gaines 2008), but these studies have suggested that this relationship is correlational rather than causal with both native and established plant richness being positively related to habitat heterogeneity. Our analysis supports this interpretation. While native plant richness consistently correlates with lower rates of establishment and invasion (Levine et al. 2004, Beaury et al. 2020), this effect is not evident at the spatial scale of an entire island because biotic resistance operates at much smaller scales of species interactions (Shea and Chesson 2002).
Island age
We hypothesized that the effect of island age on native, established, and invasive plant richness would be hump-shaped, indirect, and mediated by habitat heterogeneity. Our results were mostly consistent with this hypothesis. We found that island age had an indirect negative effect on native, established, and invasive plant richness, and in each case, the effect of island age was mediated by maximum elevation. Since maximum elevation strongly influenced native, established, and invasive plant richness, and older islands in our dataset are generally flatter than younger islands, it is likely that loss of habitat heterogeneity is the main mechanism through which geologic age affects species richness for the oceanic islands in our dataset. This finding is largely consistent with previous findings that island age determines habitat heterogeneity, which in turn influences species richness (Barajas-Barbosa et al. 2020). Indeed, Whittaker et al. (2008) proposed a ‘hump-shaped' trend in island carrying capacity over time, which was later empirically demonstrated by Steinbauer et al. (2012). Many of our study islands are much older than the four archipelagos used by Whittaker et al. (2008; Hawaii, Canary, Galapagos and Azores). Therefore, it is plausible that most of our islands have transitioned to the latter half of the ‘hump,' with lower-than-expected plant richness values driven mainly by loss of elevation and associated habitat heterogeneity. Since the relationship between island age and carrying capacity is roughly linear after it peaks (on the ‘descent'), this would help to explain why our pSEMs demonstrated proper fit despite fitting linear relationships to a variable (age) that has previously displayed nonlinearity.
Human modification
Human modification on islands could be a proxy for propagule pressure and/or human disturbance. Propagule pressure increases the number of non-native species introduced to an island (as well as the number of individuals within each non-native species), while disturbance creates new empty niches and opportunities for invasion. Because these variables are associated with human time scales (i.e. operating over the course of years to centuries) rather than evolutionary time scales (i.e. millenia), there is no expectation that they would influence native plant richness on islands; however, they may influence non-native species establishment and invasion (Britton-Simmons and Abbott 2008, Eschtruth and Battles 2009). The significant positive direct effect of human modification on established plant richness is consistent with much of the existing literature (Jeschke and Starzer 2018, Nordheimer and Jeschke 2018), with both propagule pressure and disturbance potentially increasing non-native plant establishment. Association with human populations and anthropogenic disturbance have also been shown to increase the likelihood of a native plant species being selected for introduction elsewhere (Fristoe et al. 2023). Therefore, anthropogenic forces may similarly increase the likelihood of establishment through their effects in both the native and non-native ranges of a plant species.
Interestingly, we found no significant relationship between human modification and invasive plant richness. The absence of this relationship is noteworthy because it could suggest that propagule pressure and disturbance do not influence plant invasion on islands as strongly as previously assumed. We hypothesized that increased propagule pressure would increase a new population's likelihood of finding suitable habitat and reduce the impacts of founder effects; these results would, in turn, improve establishment rates without affecting invasion rates. Our results support this hypothesis, which is derived from previous studies that have proposed that anthropogenic forces and propagule pressure should be more important at the earlier stages of the invasion process (Williamson 2006, Theoharides and Dukes 2007, Daly et al. 2023). However, these results conflict with some recent analyses that have suggested that propagule pressure and disturbance may be equally or more important for invasion than for naturalization (Lavoie et al. 2016, Essl et al. 2019). All of these analyses have been conducted at coarse spatial scales (island-wide or state-wide). Finer-scale analyses may be useful for clarifying the mechanisms that underlie the role of propagule pressure and disturbance in establishment and invasion.
Tropical susceptibility
We found that maximum monthly temperature had a positive, direct effect on native plant richness, a result which reflects well-known patterns of native biodiversity as it pertains to latitude and islands. The tropics are well-known for hosting the majority of the Earth's biodiversity (Raven 1988), and islands are similarly well-known as hotspots of endemism (Kier et al. 2009), creating an expectation of higher plant richness on islands with warmer temperatures.
We also found a positive direct effect of maximum monthly temperature on invasive plant richness, which also aligned with existing expectations. At large spatial scales, non-native plant richness often correlates with native plant richness (Stohlgren et al. 2006, Sax and Gaines 2008). Warmer climates present lower abiotic stress and higher resource availability. These patterns compound the effect of habitat heterogeneity such that tall tropical islands have both high resource availability and a variety of habitats, further increasing the likelihood that non-native species can find and exploit vacant niches to establish and invade. Although warmer climates also are positively associated with higher native plant richness, which should create biotic resistance, that process is more likely to be evident at the plot scale rather than the whole island scale (Shea and Chesson 2002, Levine et al. 2004, Beaury et al. 2020).
We did not find a significant effect of maximum monthly temperature on established plant richness, a finding that appears to contradict our hypothesis. We expected maximum monthly temperature to impact established plant richness in the same way it impacts invasive plant richness. Previous work has suggested that climate matching between the native and introduced ranges strongly influences establishment success (Theoharides and Dukes 2007, Richardson and Pyšek 2012, Fristoe et al. 2023). We did not consider the native ranges of each of the non-native plant species in our dataset, which might explain why all climatic variables were insignificant in the pSEM that modeled established plant richness. However, it is possible that once a species has already become established, the subsequent progression to invasion is determined more strongly by the overall favorability of the climate conditions (i.e. the maximum monthly temperature) in the non-native range, rather than the degree to which the recipient climate matches a species' native climate. This could explain why maximum monthly temperature was only significant in the model of invasive plant richness.
Theory of island biogeography
Although the species–area relationship was consistent for native, established, and invasive plant richness (Table 2, Fig. 2), the effect of isolation was only significant in the model that predicted native plant richness. The Theory of island biogeography was built on the assumption that colonizing species would arrive naturally (MacArthur and Wilson 1967), and indeed, our results suggest that isolation decreases the number of native plant species on islands. However, in the Anthropocene, this assumption is frequently violated, as humans greatly influence plant and animal dispersal (Helmus et al. 2014, Moser et al. 2018). For this reason, and because of our expectation that isolated flora would be evolutionarily naive, we hypothesized that more isolated islands would have more established and invasive plant species, but this hypothesis was not supported by our model results. Kueffer et al. (2010) similarly found a lack of relationship between invasive plant richness and island isolation, and noted that previous growth comparisons of native and non-native plants on isolated islands have failed to observe meaningful differences in competitive ability (Schumacher et al. 2009). Therefore, positive relationships between non-native plants and isolation reported in other studies (Moser et al. 2018, Essl et al. 2019) may be underlain by mechanisms other than competitive differences conferred by divergent evolutionary histories.
Limitations and future directions
Due to the coarse resolution of our spatial data and species data, as well as the macroscale approach of our analyses, there are some processes that we were unable to examine but warrant further investigation. One such topic is the evolutionary imbalance hypothesis (Fridley and Sax 2014), which has recently received strong empirical support for non-native plants (Fristoe et al. 2023). The difference in phylogenetic diversity between source ecosystems and recipient ones may be an underlying driver that was not considered in our analysis. Additionally, the negative impacts demonstrated by invasive plants vary widely (Pimentel et al. 2005, Vilà et al. 2011); further refinement of the invasive species pool into subcategories based on type of impact could generate novel insights in future analyses.
Conclusion
Broadly, our results confirm the significant influence of anthropogenic forces on the establishment of non-native species on islands. Notably, however, we find that human modification does not influence invasion to the same extent as establishment. Whether a given established species will invade on islands is mainly determined by the habitat heterogeneity available to the established species pool and the abiotic climatic conditions of the environment.
Acknowledgements
– We thank Jesse Bellemare and Michael Nelson for their contributions to discussions which led to this manuscript. We also thank the members of the Spatial Ecology Lab at UMass, especially Annette Evans, for their feedback on the analyses presented here. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. This article has been contributed to by U.S. Government employees and their work is in the public domain in the USA.
Funding
– This work was funded by a U.S. Geological Survey Northeast Climate Adaptation Science Center graduate fellowship, grant G19AC00091 (WGP).
Author contributions
William Pfadenhauer: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Funding acquisition (supporting); Investigation (lead); Methodology (equal); Writing – original draft (lead); Writing – review and editing (equal). Graziella DiRenzo: Data curation (supporting); Formal analysis (supporting); Writing – review and editing (equal). Bethany Bradley: Conceptualization (equal); Funding acquisition (lead); Methodology (equal); Writing – review and editing (equal).
Open Research
Transparent peer review
The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/ecog.07379.
Data availability statement
The data and code that support the findings of this study are openly available on GitLab at https://doi.org/10.5066/P9XES5OI.
