Threatened and invasive reptiles are not two sides of the same coin

Introduction

Anthropogenic activities homogenize biodiversity by increasing extinction risk, while simultaneously facilitating the transition of species through each stage of the invasion pathway (Romagosa et al., 2009; Blackburn et al., 2011; Schmidt et al., 2012). Identifying biological traits that influence a species’ response to anthropogenic activities is thus a shared goal of conservation and invasion biology (Fisher & Owens, 2004; Bradshaw et al., 2008). Indeed, both disciplines are concerned with understanding factors that determine the fate of small populations (Blackburn & Jeschke, 2009; Blackburn et al., 2009). General rules of thumb developed from trait-based analyses can be used to predict extinction or invasion risk of poorly known species, providing a rapid, cost-efficient approach to the allocation of limited conservation resources (Bradshaw et al., 2008; Böhm et al., 2016).

Traits that predispose species to become either threatened or invasive are often assumed to lie on opposite ends of a continuum: trait values that render a species more vulnerable to extinction are thought to be opposite in sign to those that facilitate invasion, and vice versa (Bradshaw et al., 2008; Jeschke & Strayer, 2008; Schmidt et al., 2012). The assumption that threatened and invasive species are ‘two sides of the same coin’ (Jeschke & Strayer, 2008) has strong links to life-history theory (Jeschke & Kokko, 2008), which posits that species can be classified along a fast–slow continuum (i.e. that invasive species have fast life histories and threatened species have slow ones). Most trait-based analyses have examined the characteristics that influence whether a species becomes threatened or invasive in isolation, and thus provide limited inference regarding the validity of that hypothesis (Bradshaw et al., 2008; Jeschke & Strayer, 2008). Furthermore, studies that have jointly analysed traits that influence a species’ fate have produced equivocal results. Studies of plants (Bradshaw et al., 2008; Pandit et al., 2011; Schmidt et al., 2012) and crayfish (Larson & Olden, 2010) support the ‘two sides of the same coin’ hypothesis, whereas studies of fishes and birds (Jeschke & Strayer, 2008; Blackburn & Jeschke, 2009) contradict it. However, these tests have either failed to account for shared evolutionary history among species (Blackburn & Jeschke, 2009), arguably used inappropriate source pools for contrasting invasive with non-invasive species (Bradshaw et al., 2008; Jeschke & Strayer, 2008; Larson & Olden, 2010; Pandit et al., 2011), examined traits individually with univariate techniques (Blackburn & Jeschke, 2009) or have been limited in geographical and taxonomic scope (Larson & Olden, 2010). In addition, the ‘two sides of the same coin’ hypothesis has not been evaluated in reptiles. Independent studies suggest that geographical range size and body size show opposing relationships in threatened (Tingley et al., 2013; Ducatez et al., 2014; Böhm et al., 2016) versus successfully established (Mahoney et al., 2015) reptile species, but these studies differ markedly in the spatial scale at which they were conducted, as well as the species and traits they considered.

Previous comparisons of threatened and invasive species have also failed to account for anthropogenic factors that directly influence the probability that a species will succumb to either fate (Cardillo et al., 2004; Pyšek et al., 2010; Murray et al., 2014). In contrast to species-level traits, we predict that many anthropogenic factors will have similar effects on invasion and extinction risk. For example, species that occupy urbanized landscapes and are sympatric with high human population densities may be introduced in greater numbers (Cassey et al., 2004; Tingley et al., 2010) and may be more likely to establish in human-modified environments where they are typically released. Similarly, species that are associated with urban habitats and large human populations may be at greater risk of extinction in their native ranges due to factors such as exploitation and habitat loss (Cardillo et al., 2004). International trade in species can also reduce the viability of wild populations through overharvesting, while increasing propagule pressure and thus invasion potential (Romagosa et al., 2009). Jeschke & Strayer (2008) and Schmidt et al. (2012) contrasted traits of threatened and invasive species and found that species-level traits were stronger predictors of invasion success than of extinction risk. They speculated that anthropogenic threats may be more important in determining extinction risk than establishment success or pest status, but this hypothesis has not been explicitly tested.

Here we overcome many of the limitations of previous tests of the ‘two sides of the same coin’ hypothesis with the use of data on extinction risk and invasiveness of reptile species from across the globe. Unlike most previous tests of this hypothesis (Bradshaw et al., 2008; Jeschke & Strayer, 2008; Larson & Olden, 2010; Pandit et al., 2011), we directly compare successful and unsuccessful introduced species, allowing us to decouple characteristics associated with establishment from those linked with earlier and later stages of the invasion pathway (Blackburn & Jeschke, 2009). Additionally, we examine the related hypothesis that anthropogenic factors that facilitate invasion success are the same as those that lead to a high risk of extinction.

Methods

Results

Species-level traits

Comparisons of trait values between threatened and invasive reptiles revealed limited support for the ‘two sides of the same coin’ hypothesis (Figs 1 & 2). Seventy-three per cent (8/11) of model coefficients describing 10 aspects of species life history, ecology, biogeography and environmental niche breadth were opposite in sign between the two groups; however, most effect sizes for variables that were opposite in sign were small and/or highly uncertain (Fig. 3). The only variable that showed somewhat strong and opposing relationships with threat status versus invasion success was SVL. Larger-bodied species were more likely to be threatened, whereas smaller-bodied species were more likely to be invasive (although 95% CIs for the latter coefficient only excluded zero when propagule pressure was not included as a covariate; Appendix S3). Furthermore, the positive relationship between body size and threat status was only apparent after accounting for taxonomy and additional confounding characteristics (cf Figs 2 & 3).

Variation in precipitation across a species’ native geographical range (a measure of environmental niche breadth) and native geographical range size were negatively correlated with threat probability. This latter finding is expected, given that IUCN uses range size to assign threat status. Native geographical range size was also negatively correlated with invasion success (but only after controlling for propagule pressure), whereas parthenogenicity, oviparity and herbivory (relative to carnivory) were all positively correlated with invasion. However, 95% CIs for the latter two coefficients slightly overlapped zero. Effects of parthenogenicity, oviparity and herbivory were much more certain when propagule pressure was excluded from the model (Appendix S3).

Discussion

We did not find strong support for the hypothesis that trait values that predispose species to become threatened or invasive lie on opposite ends of a continuum. Although nearly 75% of the species-level traits that we considered showed opposite relationships with threat status versus invasion success, effect sizes for those traits were generally small and/or uncertain. These findings are consistent with the results of previous analyses of Holarctic fish and bird species; both Jeschke & Strayer (2008) and Blackburn & Jeschke (2009) found that c. 50% of traits showed opposite relationships in comparisons of trait values between species listed as threatened by the IUCN and species that have established non-native populations. Additionally, effect sizes in both previous studies were weak for traits that differed between threatened and established species, as we observed here. The number of traits that have opposite relationships with threat status versus invasion success, as well as the magnitude of the observed effects, will clearly depend on the specific traits that are considered. Furthermore, the broad taxonomic and geographical diversity of our analysis, coupled with the coarse resolution of some of the input variables used (e.g. geographical range maps and associated variables), probably contributed to some of the uncertainty in our analyses. Nonetheless, the fact that we contrasted species at opposite ends of the threat continuum (i.e. Least Concern versus Endangered/Critically Endangered), means that our analyses had a high sensitivity to detect trait differences between threatened and non-threatened species – yet we still find little support for the ‘two sides of the same coin’ hypothesis. The overall consistency between our results and those of Jeschke & Strayer (2008) and Blackburn & Jeschke (2009) suggest that trait values that lead to vertebrate endangerment are not simply the opposite of those that facilitate invasion success.

Studies of plants (Bradshaw et al., 2008; Pandit et al., 2011; Schmidt et al., 2012) and crayfish (Larson & Olden, 2010), on the other hand, provide stronger support for the ‘two sides of the same coin’ hypothesis. The extent to which this disparity reflects true taxonomic differences in the validity of the hypothesis is questionable, however, as studies have applied very different statistical approaches (e.g. univariate t-tests versus multivariate machine learning algorithms), compared different response variables (e.g. establishment success versus pest status) or used different source pools to contrast invasive and non-invasive taxa (e.g. all species within a family or region versus only introduced species). As highlighted by Cassey et al. (2004) and Blackburn & Jeschke (2009), the latter consideration is critical. Established or alien pest species typically represent a biased subset of all species in a region or taxonomic group (Tingley et al., 2010; García-Díaz & Cassey, 2014), and thus tests of the ‘two sides of the same coin’ hypothesis will be most informative when studies can control for characteristics associated with earlier or later stages of invasion, as we have done here by comparing successful versus unsuccessful introduced species.

The trait that differed most between threatened and invasive species in our analysis was body size. Large species were more likely to be threatened, whereas small species were more likely to be invasive. Similar positive relationships between body size and threat status have been found in a wide array of vertebrate taxa, including amphibians (Sodhi et al., 2008), mammals (Cardillo, 2003), birds (Bennett & Owens, 1997) and squamate reptiles (Tingley et al., 2013; Böhm et al., 2016). Body size has also been found to correlate negatively with establishment success in insects (Lester, 2005), fishes (Ruesink, 2005), mammals (Jeschke & Strayer, 2006) and herpetofauna (Allen et al., 2013; Mahoney et al., 2015). Blackburn & Jeschke (2009) found that measures of body size (mass and length) had opposite effects on threat status and invasion success in birds and fishes as well (although effects were relatively weak for at least one response variable). Collectively, these findings suggest that although there is little support for the ‘two sides of the same coin’ hypothesis in vertebrates overall, there do appear to be weak (but consistent) diametric relationships with body size.

Small body size may facilitate invasion success by reducing the probability that a species is detected by biosecurity personnel, by increasing propagule pressure or by increasing niche opportunities upon introduction (Mahoney et al., 2015). Alternatively, differences in body size between threatened versus invasive species may reflect the fact that large species typically exhibit lower reproductive rates, slower growth rates and consequently lower population growth rates than small species (Pimm et al., 1988; Cardillo, 2003). Thus, large species may be slower to recover from anthropogenic disturbance, and have a reduced likelihood of establishing populations from low numbers of founders (Pimm et al., 1988; Owens & Bennett, 2000). This rationale is consistent with the idea that threatened and invasive species lie at opposite ends of a ‘fast–slow continuum’, in which invasive species have fast life histories and threatened species have slow ones (Jeschke & Strayer, 2008). Nonetheless, other aspects of the fast–slow continuum that we predicted would be important, such as annual fecundity, diet, reproductive mode and the presence or absence of parthenogenesis, did not conform to the predictions of the ‘two sides of the same coin’ hypothesis (although the latter three variables were correlated with invasiveness). Using a similar dataset, Mahoney et al. (2015) also found that establishment success in reptiles was higher for species with parthenogenesis; however, our findings regarding fecundity and oviparity are at odds with the results of Mahoney et al. (2015). This disparity may be due to the fact that we aggregated introductions at the species level, whereas Mahoney et al. (2015) treated individual introductions as the unit of analysis. Furthermore, our study considered different covariates, which may have influenced model coefficients and estimates of uncertainty.

Unlike previous tests of the ‘two sides of the same coin’ hypothesis, we also examined whether anthropogenic variables influenced threat and invasion probabilities. In line with our prediction, human population density had strong, positive effects on probabilities of both outcomes – species with high human densities in their native ranges were more likely to be both threatened and invasive than species from less populated areas. High human population densities have been shown to correlate with a high risk of extinction in global studies of mammals (Cardillo et al., 2004, 2008), birds (Davies et al., 2006), and amphibians (Sodhi et al., 2008) and in New Zealand reptiles (Tingley et al., 2013). Additionally, our finding that invasive reptile species tend to originate from regions with high human population densities accords with the results of Jeschke & Strayer (2008), who found that species affiliated with humans were more likely to be invasive. Positive effects of human population density on reptile extinction risk likely reflect more direct impacts of habitat loss, pollution and/or over-exploitation. Similarly, in the context of invasions, human population density plausibly integrates more proximal factors, such as propagule pressure and introduction pathways (e.g. pet trade activity) (Pyšek et al., 2010). It is also important to note that potential filtering of species pools caused by historic human activities means that our results likely underestimate the effects of human population density on extinction risk.

Our analyses also revealed effects of additional anthropogenic variables on threat and invasion probabilities. Firstly, we found that GDP within a species’ native range was negatively correlated with its threat status. Several studies have found that regions with higher GDP host fewer threatened plant and vertebrate species (Davies et al., 2006; Vamosi & Vamosi, 2008; Holland et al., 2009). Our results build upon these earlier studies by demonstrating that effects of GDP can also be apparent at the level of individual species. This correlation may reflect a greater focus on economic growth at the cost of environmental protection in developing countries or, conversely, the fact that many species that are sensitive to development have already been extirpated from regions with high GDP. Secondly, we found that the extent of urban area within a species’ native range was negatively correlated with its invasion success. Thus, sufficient amounts of native habitat may be required to provide enough propagules for species to establish successfully. The fact that GDP and urban extent did not have strong effects on both response variables, however, contradicts our hypothesis that anthropogenic variables have similar effects on threat and invasion probability.

Incorporating anthropogenic variables can not only improve the explanatory power of threat and invasion models (as we have shown here), but may also help resolve contradictions between studies testing the ‘two sides of the same coin’ hypothesis (Schmidt et al., 2012). Species-level traits may be confounded by anthropogenic variables, or traits may interact with anthropogenic variables to create context-dependent pathways to decline or invasion (Cardillo et al., 2004; Murray et al., 2011). Böhm et al. (2016), for example, found that effects of geographical range size on extinction risk depended on a species’ proximity to a large population centre. Although we found no evidence of interactions between range size and anthropogenic variables, we did find that accounting for propagule pressure shifted coefficient estimates for some species-level traits and generally increased coefficient uncertainty. Thus, the validity of the ‘two sides of the same coin’ hypothesis may be sensitive to the inclusion of propagule pressure, and future studies should seek to control for its effects. Furthermore, our results demonstrate that while anthropogenic variables, such as GDP and human population density, are not causally related to invasion and extinction processes, they can serve as useful surrogates of anthropogenic impact in the absence of more detailed data on introduction effort and key threatening processes (Cardillo et al., 2004; Pyšek et al., 2010).

By jointly analysing correlates of both threat and invasion probability, we have shed new light on the factors that lead to biotic homogenization in a diverse, but poorly understood, vertebrate group. Our study also builds on previous tests of the ‘two sides of the same coin’ hypothesis by considering a broad range of species characteristics in a multivariate context, while controlling for shared evolutionary history among species, anthropogenic factors and biases involved in the introduction stage of the invasion pathway. Despite these statistical and methodological advances, our results accord with those of previous studies of other vertebrate taxa in failing to support the ‘two sides of the same coin’ hypothesis. Thus, the assumption that threatened and invasive species lie on opposite ends of a continuum, while consistent with life-history theory, appears to be an oversimplification. However, our results strongly support the findings of previous studies demonstrating that anthropogenic variables can be important predictors of a species’ fate (e.g. Murray et al., 2014). Measures of anthropogenic impact should therefore be more routinely incorporated in trait-based analyses of extinction and invasion risk.