Introduction

Human activities have strongly altered species composition and richness of the freshwater ecosystems over the world (Vitousek 1997; Butchart et al. 2010; Leprieur et al. 2008a). The increase of global trade and economic activities during the last two centuries have promoted intentional and unintentional dissemination of species throughout the world (Taylor and Irwin 2004; La Sorte and McKinney 2007; Leprieur et al. 2008a). At the end of the twentieth century at least 624 freshwater fish species have been established into areas located outside their natural range (Gozlan et al. 2010), due of trading, aquaculture and game fishing purpose (Rahel 2007). As a consequence, more than half of the river basins in the world currently host at least one non-native fish species (Leprieur et al. 2008a), with consequent changes of large scale faunal dissimilarities among fish species assemblages (Villéger et al. 2011).

Native species extirpations have also modified the species composition of river basins throughout the world (Rahel 2007), but to a lower extent than introductions of non-native species (Olden et al. 2011; Villéger et al. 2011). Indeed, in most regions the number of non-native species introduced exceeds the number of extirpations, which ultimately leads to an increase in local species richness (Sax and Gaines 2003; Cassey and Lockwood 2007; Cadotte et al. 2009).

Besides this change in alpha diversity among river basin assemblages, the addition or deletion of species also alters the compositional dissimilarity between assemblages, i.e. beta-diversity (Olden et al. 2006; Rahel 2007), either inducing a decrease of dissimilarity (i.e. homogenization) or an increase of dissimilarity (i.e. differentiation). Numerous studies on changes in dissimilarity between the river basin assemblages in freshwater fish fauna have reported a global trend toward homogenization (McKinney and Lockwood 1999; Leprieur et al. 2008a; Olden et al. 2008; Villéger et al. 2011), although differentiation has also been reported in a non-negligible number of river basins (Villéger et al. 2011). The direction of change in dissimilarity (i.e. homogenization or differentiation) and its intensity is mainly determined by the historical richness of the assemblages, their historical dissimilarity and the number of the non-native species introduced (Villéger et al. 2011; Toussaint et al. 2014).

In most studies on freshwater fish, all non-native species were pooled to compute the changes in alpha- and/or beta-diversity (Alcaraz et al. 2005; Blanchet et al. 2010; Gozlan et al. 2010; Britton et al. 2011; Clavero 2011). However, at the global scale, the biogeographic regions (also called realms) are characterized by distinct freshwater fish fauna (Lévêque et al. 2008) and the effects of non-native species on change in dissimilarity is likely to be uneven among them within a realm since the most occurring ones may exhibit a higher probability of being transplanted and since their origin may favour their spread (Villéger et al. 2011; Toussaint et al. 2014; Villéger et al. 2015). Indeed, two types of non-native species can be considered based on their realm of origin: (1) ‘exotic’ species when the species introduced is native from another realm or (2) ‘translocated’ species when the species is native from the realm. For instance, Leprieur et al. (2008b) showed for European freshwater fish assemblages that translocated species drove assemblages towards homogenization whereas exotic species promoted differentiation.

In the current context of increasing non-native species occurrences, it is urgent to go one step further and we here discriminate the non-native species according to their biogeographic origin, and identify the non-native species that contribute the most to the overall trend toward homogenization of freshwater fish faunas. To this aim, we assessed the role played by each introduced species on change in dissimilarity among fish assemblages. Then, we quantified how their introduction type (i.e. ‘exotic’ or ‘translocated’) status and the number of times they were introduced explained the effect of each non-native species on change in dissimilarity in each biogeographic realm.

Methods

Database

We used a global database of 1054 river basins across the world (Brosse et al. 2013) for which the native and non-native status in each basin has been recorded for 9722 freshwater fish species. Therefore, historical (pre-industrial period, before human-mediated species introductions) and current compositions of assemblages (historical composition + non-native species − extirpated species) are known (Leprieur et al. 2008a; Villéger et al. 2011).

For each non-native species, we determined its realm of origin, i.e. the realm where the species was historically present. We studied changes in taxonomic dissimilarity among fish assemblages in the six biogeographic realms: Afrotropical, Australian (including Oceania), Nearctic, Neotropical, Oriental, and Palearctic (Lévêque et al. 2008). In addition, each occurrence of a non-native species was classified as ‘translocated’ if the species was historically present in another river basin from the same realm, or as ‘exotic’, if the species did not historically occur in the realm.

Measuring change in taxonomic dissimilarity

For each pair of river basins belonging to the same realm, we computed dissimilarity using the Jaccard dissimilarity index (Jaccard 1901) as follow:

$$\beta_{Jaccard } = \frac{{{\text{Number}}\,{\text{of}}\,{\text{unshared}}\,{\text{species}}}}{{{\text{Total}}\,{\text{number}}\,{\text{of}}\,{\text{species}}}} = \frac{b + c}{a + b + c}$$

with b and c the number of species present only in each river basin assemblages and a the number of species shared by the two assemblages (Villéger and Brosse 2012).

These measures were assessed for historical and current periods to compute a change in taxonomic dissimilarity from historical to current period. A negative change in taxonomic dissimilarity reveals that the species assemblages of a pair of river basins become more similar from historical to current period. This change is called homogenization. On the contrary, the differentiation refers to basin pairs that become more dissimilar with time. For each realm we then computed the mean change in dissimilarity among all pairs of basins. These analyses were performed using the betapart package (Baselga and Orme 2012).

Assessing the effect of each non-native species

We assessed the effect of each non-native species on taxonomic dissimilarity independently of other species introductions; this was done for each pair of river basins as the change in Jaccard’s dissimilarity between the historical situation and a hypothetical current situation where only the occurrences of the focal non-native species were kept (i.e. introductions of other non-native species and extirpations were ignored). Then, we computed the average change in dissimilarity among all river basin pairs in a realm to produce an indicator translating the independent effect of each non-native species in a realm. The sign of the effect gives the direction of the change due to the introduction of each non-native species. A positive effect indicates that the non-native species drives the overall change in dissimilarity toward differentiation whereas a negative value indicates that the non-native species drives the overall change in dissimilarity toward homogenization.

Determinants of the effect of each non-native species

We performed linear models to measure the contribution of the following drivers to the effect of each non-native species on change in dissimilarity in each biogeographic realm. The drivers we considered were (1) the number of river basins of the realm in which the non-native species was historically present (i.e. the occurrence as native, which is positive for a translocated species and nil for an exotic species), (2) the introduction type (i.e. translocated or exotic), and (3) the number of river basins in which the non-native species was introduced (i.e. the occurrence of introduction). This analysis was run independently for each biogeographic realm, and both simple effects and pairwise interaction between drivers were considered. For each biogeographic realm, influential outliers, defined according to the Cook’s distance cut-off (Cook 1977) defined as D = 4/(n − k − 1), where n is the number of observations and k is the number of variables in the model, were removed prior analysis to prevent from biased model outputs (Hair et al. 2009). All analyses were carried out using R, v. 3.0.2 (R Core Team 2014).

Results

The 20 most frequently introduced species in each realm represented only a small fraction of the introduced species in the Nearctic and Palearctic realms (8.9 and 11.8 %, respectively), but account for more than half of the introduction events. Indeed, those 20 species account for 52 and 60 % of the total number of non-native introductions within the Nearctic (691/1330 introduction events) and Palearctic (851/1407) realms, respectively. In the four remaining realms, the 20 most frequently introduced species account for 28.6 (in the Neotropics) to 45.5 % (in the Afrotropics) of the number of species introduced in each realm. As for northern realms, the 20 species account for more than half of the introduction events. This proportion reached 74 % (691/1222) of the introduction events in the Neotropics, 76 % (113/149) in Oriental, 82 % (128/156) in Afrotropical and 91 % (369/406) in Australian realms. These introductions of non-native species have impacted the dissimilarity of more than 60 % of river basin pairs in Afrotropical (64 %) and Neotropical (62 %) realms, and more than 80 % of basin pairs in Australian (90 %), Nearctic (83 %), Oriental (80 %) and Palearctic (81 %) realms (Fig. 1).

Fig. 1
figure 1

Freshwater fish species introduced in the six biogeographic realms. Species are ranked according to their decreasing frequency of occurrence as non-native in the realm. The bars represent the percentage of river basin pairs where the considered species is introduced. Only the twenty most frequently introduced species in each realm are represented for clarity. ‘Exotic’ species (i.e. native from another realm) are in grey, and ‘translocated’ species (i.e. native from the same realm) are in black (e.g. carp, Cyprinus carpio, occurs as translocated species in 55 % of the Palearctic river basin pairs). The dotted line represents the percentage of basin pairs which have received at least one non-native species in each realm. The number of river basins considered in each realm is indicated under brackets

Full size image

The effect of each of the twenty most frequently introduced non-native species on changes in dissimilarity was highly variable between species and realms both in terms of direction of change and intensity (Fig. 2). A non-native species introduced in different realms can have opposite effects, either towards homogenization or differentiation (Fig. 2). For instance, the brown trout (Salmo trutta) drove Australian and Neotropical realms toward homogenization whereas it was responsible from differentiation in the Oriental and Nearctic realms (Fig. 2).

Fig. 2
figure 2

Effect of the introduced species on change in dissimilarity in each realm. Species are ranked according to their decreasing frequency of occurrence as non-native in the realm. Vertical bars represent the change in dissimilarity caused by each non-native species (i.e. if it has been the only non-native species introduced, see “Methods” section for details). The observed value of the percentage of change in dissimilarity while considering all the non-native species (obs) is given for each biogeographic realm (mean ± standard deviation). Note that the obs value can be lower than the individual species effect sum. Positive values indicate an increase of dissimilarity (differentiation), whereas negative values indicate a decrease (homogenization). Bars for translocated and exotic species are filled in black and grey, respectively

Full size image

In the Australian, Neotropical and Palearctic realms, the two most introduced species contributed strongly to homogenization (Fig. 2). On the contrary, in the other three realms, the non-native species having the strongest effect on change in dissimilarity were not the one with the highest number of introductions. For instance, in the Oriental realm, the eastern mosquitofish (Gambusia holbrooki) had the strongest homogenization effect (−0.14 %, Fig. 2), while being less frequently introduced than the common carp (Cyprinus carpio) or the Mozambique tilapias (Oreochromis mossambicus) that accounted for only 0.001 and 0.01 % of change in dissimilarity (Fig. 2). A similar trend was found in the Nearctic realm where the rainbow trout (Oncorhynchus mykiss) has a strong homogenization effect (−0.23 %) although being less frequently introduced than the common carp that hardly affected changes in dissimilarity (−0.0001 %, Fig. 2). This was confirmed by Table 1 which shows that the occurrence of introduction was not the only driver explaining the homogenization effect of each non-native species. Indeed, in all the biogeographic realms except the Australian, the introduction type (i.e. translocated or exotic) and/or the native occurrence of the species significantly affected (directly or interacting with the occurrence of introduction) the homogenization effect of each non-native species. Moreover, the native occurrences of the species in the realm has a significant effect, or interact with the occurrence of introduction, in all those 5 realms (Table 1), testifying for the strong homogenization effect of translocated species (Fig. 2).

Table 1 Analysis of variance testing the influence of the occurrence of the species as native, as introduced and of the type of introduction (translocated vs. exotic) on the change in dissimilarity due to each non-native species in the six biogeographic realms
Full size table

Discussion

Among the 624 freshwater fish species introduced over the world (Rahel 2007; Gozlan et al. 2010; Lauzeral et al. 2011), fourteen have been introduced into at least one river basin in the six realms, and 90 non-native species have been introduced in realms other than their native one (Rahel 2007; Villéger et al. 2015). Leprieur et al. (2008a) emphasized the role of human activities, which promoted the introduction of species of economic interest, leading to the introduction of a few species over large areas. Those widespread processes have received attention in previous local studies on homogenization patterns reporting the important role of a small pool of species of interest as part of human activities (Rahel 2000; Clavero and García-Berthou 2006; Rahel 2007). Although a few non-native species can synergistically contributed to the global trends of freshwater fish homogenization, we showed here that the overall pattern could hide contrasted species individual contributions.

Indeed, analysing species individual effects enabled to highlight strong discrepancies among species and among realms, suggesting that non-native species contributed differently to the homogenization process. Among the 61 most frequently introduced species in the world (i.e. the global pool made by the 20 most frequently introduced species in each of the six realms, that account for 69 % of the global number of introductions, Fig. 1), only 10 non-native species (Carassius auratus, Cyprinus carpio, Gambusia holbrooki, Micropterus salmoides, Oncorhynchus mykiss, Oreochromis mossambicus, Oreochromis niloticus, Perca fluviatilis, Salmo trutta, Tilapia zillii) had an effect toward homogenization higher than 10 % of the realm homogenization level. Those species are all known to be highly invasive, often reaching high abundances (Gherardi and Acquistapace 2007; Rahel 2007; Lauzeral et al. 2011), but we nevertheless found a low correlation between the number of introductions and the effect of each of these species on change in dissimilarity (Table 1). This suggested that the number of introduction was not the only determinant of the potential effect on change in dissimilarity. This was particularly striking for common carp (Cyprinus carpio), which although being the most frequently introduced species in Oriental, Nearctic and Palearctic realms, had only a weak impact on change in dissimilarity in Nearctic and Oriental realms. On the contrary, this species had a strong effect toward homogenization in the Palearctic realm. A similar pattern was observed for several species (e.g. Oreochromis mossambicus or Tilapia sp.), suggesting that the identity or the occurrence of the species itself did not explain its role in the change in dissimilarity.

Indeed, the type of introduction (i.e. translocated vs. exotic) was a significant predictor of the changes in dissimilarity in five out the six biogeographic realms. The translocated species had a strong effect on homogenization patterns because they were historically already present as native species. Hence, their introduction in more basins of the realm could only lead to homogenization for all the basin pairs where the species was historically present. Such a situation is one of the major drivers of homogenization worldwide (Olden and Poff 2004; Toussaint et al. 2014).

In contrast, exotic species were often introduced in a few basins and some of them then spread in nearby basins (Arthington 1991; Lintermans 2004). Exotic species hence tend to favour differentiation at the early stages of invasion, as shown by Leprieur et al. (2008b) for Western Europe. Nevertheless, when those species expand over their invasion area, the number of river basin pairs where they occur increases and can contribute to homogenization (Toussaint et al. 2014). This reversal effect was observed, for instance, for eastern mosquitofish (Gambusia holbrooki) that quickly invaded large areas in Australian and Oriental realms since its initial introduction in the early twentieth century (Welcomme 1988; Milner and Petts 1994; Pyke 2008; Ayres and Clunie 2010), and hence currently has a strong homogenization effect in those realms (Fig. 2).

Focusing conservation efforts in controlling a few species that have a prominent effect on the global homogenization process (e.g. Carassius auratus, Cyprinus carpio, Gambusia holbrooki, Micropterus salmoides, Oncorhynchus mykiss, Oreochromis mossambicus, Oreochromis niloticus, Perca fluviatilis, Salmo trutta, Tilapia zillii) may be more relevant to counteract the current homogenization trend and its local detrimental effects on species community integrity. We should nevertheless be aware that the homogenization rate of an area results from complex combinations of individual species homogenization and differentiation effects. Therefore, the homogenization strength measured here at the species individual level did not represent the complete species effect when considering other non-native species introductions. Nevertheless, our approach and results provide interesting perspectives to optimize management and conservation efforts by focusing on a limited number of species influencing global homogenization patterns.

Overall, our results confirm and generalize to a global scale the previous local or regional studies on the consequences of fish introductions on changes in dissimilarity (Rahel 2000, 2007; Clavero and García-Berthou 2006; Leprieur et al. 2008b). We also pointed out that a few freshwater fish species widely introduced across the world have largely contributed to the observed changes. Those species are characterized by their large distribution over considered areas, and are hence either the species introduced for a long time, such as common carp in the Palearctic (Welcomme 1988; Balon 2004), or brown trout in the Nearctic (Crawford and Muir 2008), or recent introductions of fast spreading species, such as Gambusia spp. in Australia (Welcomme 1988; Milner and Petts 1994; Pyke 2008; Ayres and Clunie 2010).

Intercontinental fish exchanges should decrease owing to the development of national and international laws regulating the transport and introduction of exotic species (Costello et al. 2005; Hulme et al. 2009). But under the current accelerated rates of non-native species spread within realms (Britton and Gozlan 2013), exotic species are expected to establish over larger areas in the near future, and then turn their differentiation effect toward homogenization. Thus, this would favour a global homogenization, as forecasted by Villéger et al. (2015) using mathematical simulations. From a conservation point of view, the origin, as well as the occurrence of the non-native species should be considered to better control the forthcoming changes in taxonomic dissimilarity across realms. Controlling the spread of a few species might significantly reduce the global homogenization trend. Such actions are nevertheless pending for a better assessment of the ecological consequences of homogenization, that to date, remain largely unknown (France and Duffy 2006; Olden 2006).