Invasive species are considered one of the main threats to native ecosystems, because they can cause negative impacts on the stability and evolutionary trends in any community via changes in ecological, behavioral and genetic interactions (e.g., hybridization and introgression; Mooney and Cleland, 2001; Cox, 2004; Simberloff, 2010). To make matters worse, the number of anthropogenic-mediated invasions increased in recent years due to globalization (Perrings et al., 2005; Poulin et al., 2011). The process of biological invasion is fairly complex and can be affected by several factors. The first factor is the capacity of the potential invader to transpose environmental filters, which can interact with the number and frequency of invading propagules (Colautti and McIsaac, 2004). The second factor is the species’ capacity to establish itself and spread in novel environments. High adaptability to new conditions is a crucial trait at this stage, because if they cannot adapt to local conditions they can neither establish themselves nor invade (Jímenez-Valverde et al. 2011; Dobrzycka-Krahel and Graca, 2014; Dobrzycka-Krahel et al., 2015). The third and last factor is the susceptibility of the ecosystem to invasive species (Colautti and McIsaac, 2004; Cox, 2004; Colautti et al., 2006). Due to this inherent complexity, conservation measures employed to mitigate the threat of invasive species need to cover all steps of the invasion process. Consequently, these measures quickly become expensive, highlighting the need for cost-effective measures.
According to Hulme (2006), the most cost-effective measure is preventing the potential invader from establishing itself in the novel environment; these measures are relatively cheap and have a high chance of success, if employed correctly. One tool that can be used to help guide the conservation efforts is species distribution modelling (SDM) (Araujo and Guisan, 2006; Jiménez-Valverde et al., 2008). SDM techniques calculate the environmental space (‘climatic niche’) of the species based on the environment in which it currently occurs (occurrence records), thus being a portion of the realized niche (Soberón, 2007). This climatic niche can then be projected on to the geographic space, including areas in which the species is not currently found (Jiménez-Valverde et al., 2011). If used in the context of biological invasions, SDM can indicate geographical regions that are suitable for a given species, potentially to establish itself and have a positive growth rate (Jiménez-Valverde et al., 2011). Therefore, it offers a framework for conservation strategies to be drawn up more effectively (Guisan et al., 2013). This type of modelling has been used extensively in recent years and has become a useful tool to study invasion ecology (e.g. Peterson, 2003; Nori et al., 2011; Palaoro et al., 2013; Campbell et al., 2015; Marcantonio et al., 2016). Another source of evidence we can use to predict the impact of the invader is to compare the invader’s climatic niche with that of a native species. If native and invasive species have similar climatic niches, it is possible that they may compete for resources on a smaller scale, creating further disturbances for native species (Palaoro et al., 2013).
Neotropical biodiversity has been considered vulnerable to invasive species, with many groups of native plants and animals - especially freshwater organisms - threatened by extinction due to the activity of invasive species (Rodríguez, 2001; Clavero and García-Berthou, 2005; Speziale and Lambertucci, 2010). A particularly striking example is the thermophilic сladoceran Daphnia (Ctenodaphnia) lumholtzi Sars, 1885 (Crustacea: Cladocera: Daphniidae). This species complex is native to southern Asia, Australia and Africa (Kotov and Taylor, 2014), but it successively invaded North America in the 1990’s (Havel and Hebert, 1993; Havel et al., 2000; Havens et al., 2000). Although Havens et al. (2012) indicated no evidence for competitive exclusion of native cladocerans by invasive species in lakes of North America, experimental studies showed that D. lumholtzi is able to tolerate a wide range of temperature variation (Brown, 1999; Johnson and Havel 2001; Fey and Cottinghan, 2012) and that, in the presence of predators, D. lumholtzi increases its investment in reproduction, becoming a stronger competitor when compared to native species (Dzialowski et al., 2003). In invaded habitats, D. lumholtzi might suppress native species at least seasonally (Dobberfuhl and Elser, 2002). So far, there is no consistent information on the influence of D. lumholtzi on the provision of ecosystem services in invaded habitats, or of economic impacts of the invasion process.
After the establishment of the first populations in North America (Havel and Hebert, 1993), a rapid colonization proceeded (Dzialowski et al., 2003). Soon after that, D. lumholtzi was reported in Mexico (Eliás-Gutierréz et al., 2008; Silva-Briano et al., 2010), Brazil (Zanata et al., 2003; Simões et al., 2009) and, more recently, Argentina (Kotov and Taylor, 2014). A darker scenario emerges when we realize these records are distant from each other, both in space and time - thus not enough sampling has been done, and the situation in the Neotropics may be significantly underestimated. Of all Daphnia species with exclusive occurrence in the Neotropics (see Kotov et al., 2013), only D. gessneri Herbst, 1967 has a wide distribution range. At least five other Daphnia species are endemic to a few water bodies or even a single one (Hann, 1986; Valdivia-Vilar and Burger, 1989; Kotov et al., 2010), thus being considered geographical relicts (sensu Grandcolas et al., 2014). This is troubling, since a high level of endemism usually correlates with environmental specificity, which increases the risk of functional and local extinction (Carlton et al., 1999; Sodhi et al., 2009). Therefore, the invasion of D. lumholtzi increases the potential risk of extinction for the Neotropical Daphnia, which makes understanding the invasive potential of D. lumholtzi even more necessary.
In this study, we used SDM to predict the invasion scenario for D. lumholtzi in the Neotropics. We used occurrence records of this species (including both native and invaded areas) to build a model and then project it in the Neotropics. In addition, we checked if D. lumholtzi and the endemic Neotropical species of Daphnia occupy similar, or significantly overlapping, climatic niches.
We found 137 occurrence records of D. lumholtzi in different locations throughout the world (Fig. 1). These data were obtained from the Global Biodiversity Information Facility (GBIF) (http://www.gbif.org), United States Geological Survey (USGS, 2012) and from different previously published literature sources (Supplementary References). We did not include records from the Caspian region, because it represents a separate species initially described as Daphnia lumholtzi palaeartica Behning, 1928. It has obvious morphological differences from D. lumholtzi and has a distinct life-history, i.e. it is adapted to cooler temperatures than its congeners (Behning, 1941).
To build the SDM, we did not distinguish between records of native and invaded populations, because invasive species may violate two premises of SDMs: i) the invasion process may not be complete; ii) the niche may change in novel habitats (Peterson, 2003; Guisan and Thuiller, 2005; Broennimann et al., 2007; Jiménez-Valverde et al., 2011). The recommended method to overcome these issues is to include both the native and invaded ranges, since these models how better performances than models using only the native range (Broennimann and Guisan, 2008; Beaumont et al., 2009; Capinha et al., 2011). The SDM was performed using eight climatic variables taken from the Wordclim database (Hijmans et al., 2005) (download: http://www.worldclim.org) and one topographic variable (Breakline emphasis GMTED2010) at a spatial resolution of 30 arc-seconds (Tab. 1). These variables were chosen because they were not highly correlated (Pearson’s r <|0.9|). According to Jiménez-Valverde et al. (2011), it is also necessary to consider the biological relevance of the variables chosen. Thus, all variables used to build the SDM are related with features from the biology or ecology of the cladoceran species (Tab. 1).
Maximum entropy modelling (MaxEnt) is a machine learning method that is considered the most robust method for predicting environmental suitability using presence-only data. The advantages of this method are: i) mathematical functions are concise and the results are easily interpreted (Phillips et al., 2006; De Marco and Siqueira, 2009); and ii) MaxEnt can make accurate predictions using a reduced set of occurrence records (Costa et al., 2010). When compared to other SDM techniques, Max-Ent was among the most effective when using presence-only data, with a performance comparable to techniques that use presence and absence data (Elith et al., 2006; Costa et al., 2010). Thus, we chose MaxEnt to model the environmental suitability of D. lumholtzi in the Neotropics. We used occurrence records from both native and invaded areas to calibrate the model, which was then projected to the Neotropical region, following the suggestion of Peterson (2003) and Jimenéz-Valverde et al. (2011). A cross-validation with 10 replications was used to estimate errors of adjusted functions and the predictive power of data held for testing (Palaoro et al., 2013). The SDM for D. lumholtzi was assessed using the area under the receiver-operating characteristic curve (AUC). The AUC provides a threshold independent measure of model performance compared to null expectations. It has been used to assess SDMs due to the evaluation of specificity (absence of commission errors) and sensitivity (absence of omission errors) (Elith et al., 2006). If AUC values are close to 1, they indicate a good performance of the SDM. To help in the interpretation of results we followed the suggestion of Palaoro et al., (2013), as follows: AUC >0.9=excellent; 0.8 >AUC <0.9=good; 0.7 >AUC <0.8=fair; 0.6 >AUC <0.7=bad; AUC <0.6=failed.
To observe if the climatic niche of D. (Ctenodaphnia) lumholtzi may overlap with that of native species of the genus, we obtained the occurrence records of D. (Daphnia) gessneri Herbst, 1967, D. (D.) peruviana Harding, 1955, D. (D.) marcahuasensis Valdivia-Villar & Burger, 1989, D. (Ctenodaphnia) chilensis (Hann, 1986), D. (C.) dadayana (Paggi, 1999), D. (C.) inca Korinek & Villalobos, 2003, D. (C.) menucoensis Paggi, 1996, D. (C.) ornithocephala Baribén, 1953, D. (C.) paggii Kotov, Sinev & Berrios, 2010, D. (C.) spinulata Baribén, 1917. Due to the uncertainty of several occurrence records in the literature, we checked all native species on the List of Cladocera of the World to ascertain that they are only endemic in the Neotropics (Kotov et al., 2013).
After that, we gathered the environmental variable values for each occurrence record of the native species and D. lumholtzi and transformed the data using ranging transformation:
Lastly, we performed an ordination analysis (principal components analysis - PCA) using a covariance matrix. Data were analyzed in the Software PAST (Hammer et al., 2001).
The calibrated model had a mean AUC of 0.884±0.03, suggesting a good model (Fig. 2). The projection showed high suitability in the regions connected by the Paraná River Basin between southwestern Brazil, eastern Argentina and Uruguay (Fig. 2). High suitability was also found in South-central Chile, coastal areas of Brazil (Atlantic coast) and Mexico (Fig. 2).
PCA axis 1 (PC1) explained 54.48% of the variance with an eigenvalue of 0.16, while axis 2 (PC2) explained 18.17% of the variance with an eigenvalue of 0.05. The ordination presented a clear gradient related to climatic variables (from the right to the left) with D. spinulata, D. menucoensis, D. dadayana, D. ornithocephala and D. gessneri showing different degrees of overlap with D. lumholtzi (Fig. 3 A,C,D,G,H,K). In contrast, D. lumholtzi does not overlap with species that are endemic to mountainous regions, i.e. D. chilensis, D. inca, D. marcahuasensis, D. paggii and D. peruviana (Fig. 3 A,B,E,F,I,). The most important variables (r >|0.7|) observed were isothermality (r=-0.94), temperature seasonality (r=0.98), temperature annual range (r=0.95) for PC1, and precipitation of the driest quarter (r=-0.73) for PC2.
Our results showed that D. lumholtzi has the potential to invade large portions of the Neotropics that are not yet colonized (or regarded as not yet colonized due to insufficient sampling). We highlight the high environmental suitability of D. lumholtzi in Argentina, Chile and Uruguay, which should increase the probability of invasion in many water bodies in that region (Fig. 2; Sousa et al., 2014). This result is worrying, because D. lumholtzi disperses both adults and resting eggs (in the ephippia) via lotic systems, and uses stagnant water bodies - such as reservoirs and dams - to establish viable populations (Panov et al., 2004: Havel and Medley, 2006). These lowland man-made reservoirs are more susceptible to invasions when compared to natural water bodies, because they function as stepping stones for the invaders (Havel et al., 2005, 2015; Johnson et al., 2008). The main issue is that the Paraná River basin, the most vulnerable area for the invasion of D. lumholtzi according to our model, connects three countries on a large spatial scale by lotic and flood systems with periodic hydrological cycles - also presenting numerous dams and reservoirs. Due to this, the population found in Argentina by Kotov and Taylor (2014) could have come originally from Brazil, since the first reports in South America come from there (Zanata et al., 2003; Simões et al., 2009). Further colonization of the Neotropics could occur by dispersion via water currents (e.g. the Argentinian population) or by waterfowl (i.e., carrying the ephippia on their feathers and through midgut passage; Incagnone et al., 2015).
Daphnia lumholtzi shows a considerable climatic niche overlap with some native lowland species (Fig. 3). Havens et al. (2012) indicate no evidence of the competitive exclusion of native species by D. lumholtzi invaders in lakes of North America. However, some key sources of evidence show why D. lumholtzi may negatively affect native Daphnia species. Daphnia lumholtzi has a relatively higher fitness in warmer waters when compared to other Daphnia species (Kolar et al., 1997; Work and Gophen, 1999; Lenon et al., 2001; Wittmann et al., 2013). In addition, D. lumholtzi increases its reproductive effort when exposed to predators (Dzialowski et al., 2003) and develops a long and strong helmet and tail spine, serving as a mechanism to reduce predation, causing difficulties to invertebrate predators and even fish larvae (Engel and Tollrian, 2009; Engel et al., 2014).
In the Indian subcontinent, Padhye et al., (2016) indicated that the distribution of D. lumholtzi was associated with localities that show warm and stable temperatures. Likewise, the relatively higher temperatures and higher stability of lowland Neotropical ecosystems when compared to temperate ecosystems would thus increase any competitive advantage of D. lumholtzi over the native species (McMahon, 2002). This was supported by the PCA, where isothermality (related to climatic stability) and temperature seasonality (related to climatic seasonality) seem to be responsible for the ordination of D. lumholtzi to D. chilensis, D. inca, D. marcahuasensis, D. paggii, D. peruviana (Fig. 3 A,B,D,E,F,I,J). These five mountainous endemics do not show a climatic niche overlap with D. lumholtzi. However, D. lumholtzi may also influence these native species via niche expansions or shifts that may occur during the invasion process (e.g., predator release effect, niche shifts; Capinha et al., 2011; Petit-pierre et al., 2012).
If widely distributed lowland species are more vulnerable to the invader, endemic species of Daphnia may be protected against such a fate by natural barriers and unusual (extreme) environmental conditions. At least some South American endemic daphniids with very local distribution could be regarded as relicts (Kotov et al., 2010). Relicts, usually occurring in extreme habitats, are extinction-prone due to environmental disturbances caused by human activity (Pitman et al., 2002; Lavergne et al., 2006; Chambers et al., 2008). Paradoxically, we found that locally distributed relicts may be better protected against invaders due to the climatic and geographic singularities in which they occur. Zhang et al. (2016) showed that geographic isolation was an important predictor of the distribution pattern of relict populations of the plant Emmenopterys henryi Oliv. (Rubiaceae). In the case of Neotropical relict Daphnia species, current geographic isolation provided by mountains (especially the Andes) can be an important barrier to prevent propagule pressure of D. lumholtzi from the lowlands. Indeed, the connections between aquatic systems (especially man-made) and their position in the landscape seems to have a primary and important role in the invasion process of D. lumholtzi in North and South American lowlands (Havel and Medley, 2006; Simões et al., 2009). Unfortunately, such barriers are not effective in the case of occasional or deliberate human-mediated dispersion.
Another factor that was not covered in the model is UV radiation, which is very strong in the highlands. Most mountainous endemic species are strongly pigmented (Benzie, 2005; Kotov et al., 2010) to protect themselves against UV radiation. The situation in the Neotropics could be similar to that in water bodies of the African Rift Valley, where ‘aggressive’ American clones of D. cf. pulex replaced the native daphniids in many lowland water bodies (Mergeay et al., 2005a, 2005b, 2006) but did not impact the endemic species, D. izpodvala, which inhabits the Bale Mountains. Daphnia izpodvala Kotov & Taylor, 2010 is strongly pigmented, while D. cf. pulex is not able to produce such melanin pigmentation (Kotov and Taylor, 2010).
Our results represent the first approach using modelling distribution of D. lumholtzi on a large spatial scale. Landscape features, higher temperatures and climatic stability in lowlands may facilitate the invasion process of D. lumholtzi due to the connectivity of lotic systems, presence of man-made water bodies, and higher fitness of the invader in warmer and more stable climates. In the Neotropics, generalist species, such as D. gessneri, might be relatively more threatened by the invasion process than other species. Daphnia lumholtzi and D. gessneri already co-occur in the Paraná River Basin (Zanata et al., 2008; Simões et al., 2009). Meanwhile, the other relict species seem to be protected from the invader due to the geographic isolation and local conditions of their habitats, features that may prevent D. lumholtzi’s establishment in the region. Regarding future studies, in vulnerable regions, such as the coastal regions of Brazil and Mexico, we urge for a continuous monitoring of the cladoceran fauna. In the regions already invaded by D. lumholtzi we recommend further studies so we can fully understand the extent of the current invasion. More studies on how the invader may affect native species, either via ecological (e.g., the interaction between the species) or evolutionary studies (e.g., behavior, genetic studies), are also needed. The co-occurrence of D. lumholtzi and D. gessneri in the Paraná River Basin seems particularly interesting for this purpose. It is also necessary to emphasize that the water bodies where the relict species occur are located in areas that have experienced minimal human disturbance. By keeping them preserved, the risk of occasional introduction of D. lumholtzi ephippia in those environments decreases, thus protecting the relict taxa.