Importance of geographic origin for invasion success: A case study of the North and Baltic Seas versus the Great Lakes–St. Lawrence River region

Abstract Recently, several studies indicated that species from the Ponto‐Caspian region may be evolutionarily predisposed to become nonindigenous species (NIS); however, origin of NIS established in different regions has rarely been compared to confirm these statements. More importantly, if species from certain area/s are proven to be better colonizers, management strategies to control transport vectors coming from those areas must be more stringent, as prevention of new introductions is a cheaper and more effective strategy than eradication or control of established NIS populations. To determine whether species evolved in certain areas have inherent advantages over other species in colonizing new habitats, we explored NIS established in the North and Baltic Seas and Great Lakes–St. Lawrence River regions—two areas intensively studied in concern to NIS, highly invaded by Ponto‐Caspian species and with different salinity patterns (marine vs. freshwater). We compared observed numbers of NIS in these two regions to expected numbers of NIS from major donor regions. The expected numbers were calculated based on the available species pool from donor regions, frequency of shipping transit, and an environmental match between donor and recipient regions. A total of 281 NIS established in the North and Baltic Seas and 188 in the Great Lakes–St. Lawrence River. Ponto‐Caspian taxa colonized both types of habitats, saltwater areas of the North and Baltic Seas and freshwater of the Great Lakes–St. Lawrence River, in much higher numbers than expected. Propagule pressure (i.e., number of introduced individuals or introduction effort) is of great importance for establishment success of NIS; however in our study, either shipping vector or environmental match between regions did not clarify the high numbers of Ponto‐Caspian taxa in our study areas. Although we cannot exclude the influence of other transport vectors, our findings suggest that the origin of the species plays an important role for the predisposition of successful invaders.


| INTRODUCTION
Anthropogenic introductions of species to new areas increase due to globalization and climate change, leading to homogenization of biodiversity worldwide (Capinha, Essl, Seebens, Moser, & Pereira, 2015;Hellmann, Byers, Bierwagen, & Dukes, 2008;Hulme, 2009;Olden, Poff, Douglas, Douglas, & Fausch, 2004). Species are incidentally transported with commercial travel and trade, such as in ships' ballast water, wood packing materials, and horticultural soils, or are intentionally introduced like for games or biocontrol (Briski et al., 2013;Hulme et al., 2008;Lockwood, Hoopes, & Marchetti, 2007). Many species fail to establish a viable population after arriving to a new environment, but those that succeed may have significant consequences for local communities, ecosystem functioning and/or services to human society (Carlton & Geller, 1993;Chapin et al., 2000;Olden et al., 2004;Simberloff et al., 2013). Though empirical and statistical evidence suggests that propagule pressure (i.e., number of introduced individuals) is of crucial importance for establishment success (Hayes & Barry, 2008;Simberloff, 2009), population characteristics such as phenotypic plasticity and preadaptation to cope with changeable environmental conditions may keep a high propagule pressure of species while passing through the stages of the invasion process (i.e., transport, introduction, establishment, and spread; Colautti & MacIsaac, 2004;Lande, 2015). Moreover, species evolved in regions known as more geologically and environmentally disturbed and challenged may possess life-history traits, higher phenotypic plasticity, or adaptational and evolutionary capacity which would enable them to be more successful invaders (Reid & Orlova, 2002). If species from certain area/s are proven to be better colonizers, management strategies to control transport vectors coming from those areas must be more stringent, as prevention of new species introductions is a cheaper and more effective strategy than eradication or control of established NIS populations (Hulme et al., 2008;Lockwood et al., 2007;Lodge et al., 2006).
After the opening of canals that link the North and Baltic Seas with the Black and Caspian Seas (the Rhine-Main-Danube, Volga-Don and Volga-Baltic Canals), species from the Ponto-Caspian region (i.e., Black, Azov and Caspian Seas; Figure 1) spread and became abundant in freshwater and estuarine ports of northern Europe (Leppäkoski et al., 2002;Ricciardi & MacIsaac, 2000). The invasion history of the Laurentian Great Lakes tells a more intriguing story, with many Ponto-Caspian species establishing in the region after invasion of Europe (Leppäkoski et al., 2002;Ricciardi & MacIsaac, 2000). Shipping is a leading mechanism for the spread of aquatic nonindigenous species (NIS) globally (Molnar, Gamboa, Revenga, & Spalding, 2008;Ricciardi, 2006), and as ship transit between the North and Baltic Seas and the Great Lakes-St. Lawrence River is relatively high and of similar intensity in both directions (Kaluza, Kölzsch, Gastner, & Blasius, 2010), one would expect a similar ratio of NIS from the Great Lakes-St. Lawrence River in the North and Baltic Seas and vice versa. However, recent studies stated that the transfer of species has been asymmetrical, with only a small number of species from the Great Lakes having invaded Northern European waters (Leppäkoski, Gollasch, & Olenin, 2010;Reid & Orlova, 2002).
The North and Baltic Seas and the Great Lakes-St. Lawrence River regions are intensively explored systems, and probably the most studied areas with regard to aquatic NIS globally (AquaNIS, 2015;DAISIE, 2015;Gollasch, Haydar, Minchin, Wolff, & Reise, 2009; Great Lakes Aquatic Nonindigenous Information System (GLANSIS) database, 2014; Pyŝek et al., 2008;Reise, Gollasch, & Wolff, 1999;Ricciardi, 2006). Both regions are geologically young water bodies formed by glaciations (Leppäkoski et al., 2002;Reid & Orlova, 2002). Their habitat types represent an interesting inverse mirror image with the North and Baltic Seas being mostly marine ecosystem with several large brackish F I G U R E 1 Salinity of the Great Lakes-St. Lawrence River, North Atlantic Ocean, North, Baltic, Mediterranean, Black, Azov, and Caspian Seas, constructed using average annual salinity data with a 1° x 1° spatial resolution from the World Ocean Atlas database (Antonov et al., 2006) (a). Close-up maps of the North and Baltic Seas (b), the Great Lakes-St. Lawrence River (c), and the Black, Azov, and Caspian Seas (d) are shown, as well. The green lines mark the boundaries of the studied areas. Although, the salinity of the Great Lakes is shown in the range from 0.0 to 4.1 (i.e., dark blue), the salinity of the Great Lakes is under 0.5 ppt (i.e., freshwater) to freshwater estuaries, while the Great Lakes-St. Lawrence River region is predominantly a freshwater environment with a huge brackish to saline St. Lawrence River estuary and Gulf of St. Lawrence (Figure 1; Antonov, Locarnini, Boyer, Mishonov, & Garcia, 2006;Environment Canada, 2013;Pocklington, 1986;Reid & Orlova, 2002). Both systems are marginal water bodies of the North Atlantic Ocean (Pocklington, 1986). Despite their opposing salinity patterns, some parts of the systems are rather alike: in particular, the Baltic Sea and the Lower St. Lawrence River. The Baltic Sea is a large semi-enclosed brackish water area characterized by a strong salinity gradient ranging between 2 and 24 ppt (Leppäkoski et al., 2002), while the salinity of the St.
Lawrence River, though freshwater in large part of the river stretch, starts to increase from Quebec City (5 ppt) reaching 24-32 ppt in the Gulf of St. Lawrence (Environment Canada, 2013;Pocklington, 1986).
The climate in the Baltic Sea and St. Lawrence River is also similar, ranging from maritime temperate to continental subarctic climate (Pocklington, 1986

| Observed numbers of NIS, their origin, and taxonomic composition
Lists of aquatic NIS were compiled for the North and Baltic Seas, and the Great Lakes-St. Lawrence River region, respectively. The North and Baltic Seas NIS list (Appendix S1) was assembled using data from AquaNIS-the information system on aquatic nonindigenous and cryptogenic species (AquaNIS, 2015), Reise et al. (1999) America (inland freshwaters), Ponto-Caspian region and unknown region. If a species was native to two or more regions, its contribution was counted as a ratio of "one" over the number of regions that the species was native to. For example, if a species was native to two regions, the value of 0.5 has been assigned to each region. However in

| Expected numbers of NIS
To calculate expected numbers of NIS from major donor regions for the North and Baltic Seas and Great Lakes-St. Lawrence River region, we first estimated average species richness for major donor regions using derived global species richness data from Tittensor et al. (2010).
We calculated average species richness for a particular donor region by adding derived species richness of all coastal grids (880-km resolution equal-area grid) of that region, and then divided this total derived species richness with the number of coastal grids in that region. The average species richness of coastal grids was used to avoid overestimation of species richness due to a potential overlap of the same species from Lawrence River region compatible to the rest of our data (i.e., per 880km resolution equal-area grids). The correction was necessary as total species richnesses per regions were approximately ten times higher than species richness per 880-km resolution equal-area grids.
In the second step, the obtained estimated average species richness was multiplied by the probability of invasion between regions to get the expected numbers of NIS transported from a donor to a recipient region. The expected number of NIS was used as a null model of NIS exchanges irrespective of species' traits, which can be compared with the observed NIS exchanges. The invasion probabilities were calculated using the statistical model of Seebens, Gastner, and Blasius (2013). The model integrates global ship movement data, biogeographical similarity, and environmental conditions of ports worldwide to obtain the likelihood that a NIS is transported in ballast water from a donor port, released in a recipient port and able to establish a new population there.
According to Seebens et al. (2013), the model consists of three independent probabilities each denoting an important step of the invasion process: first, the ballast water released at site j may contain species from all regions previously entered by the ship including NIS but also species which are native to the recipient site j. This is accounted for by the probability to be nonindigenous describing the probability that a species native at donor port i is nonindigenous in recipient port j. P ij (nonindigenous) is a sigmoidal function of geographic distance d ij between the ports, with β and γ being constants, and can be interpreted as the proportion of NIS inoculated in the ballast water of a ship and transported over a certain distance.
Second, the probability of introduction describes the likelihood that a species is introduced from port i to port j on ship route r: It increases with the amount of released ballast water B r that originates from port i on ship route r and decreases with mortality rate μ and travel time Δt r between i and j. Ship routes were established from nearly 3 Mio. port calls (arrival and departure dates at ports) of 32,511 ships during 2007-2008. The arrival and departure dates as well as ship-specific information were reported by the automatic identification system (AIS) and provided by Lloyd's Register Fairplay (www.ihs.com). For each port call of a ship, B r was calculated depending on the ship type, ship size, the mean ballast water tank volume, and the past route of the ship. For a certain ship type and ship size, a mean volume of discharged ballast water was calculated from 717,250 ballast water release protocols provided by the National Ballast Information Clearinghouse for the USA (NBIC 2012). Although these data are restricted to the USA, they represent by far the most comprehensive collection of ballast water release protocols currently available. To estimate the amount of discharged ballast water originating from port i, we require the mean ballast water tank volumes for different size classes and ship types, which were obtained from the American Bureau of Shipping (ABS 2011). While assuming a constant release of ballast water at each port of call, we were then able to calculate for each port call of a ship the mean ballast water volume B r originating from port i and discharged at port j. Travel times Δt r were extracted from these ship routes.
Third, the probability of establishment describes the likelihood that a species native at port i is able to establish a population in the recipient port j:

is a Gaussian function of differences in water temperatures
T and salinities S normalized by standard deviations σ T and σ S , and α being a constant.
The product of the three probabilities gives the probability of invasion P ij (Inv). To obtain invasion probabilities between regions, the invasion probabilities from all ports a in the donor region A to all ports b in the recipient region B were aggregated according to P A,B (Inv)  (Table S2).
Geographic origins of NIS differed between the two regions   (Tables S1 and S2). Pacific were similar to expected numbers from these regions (p > .5; Figure 4, Table 2). However, expected numbers of NIS from the northwest Atlantic and Great Lakes-St. Lawrence River region to the North and Baltic Seas were two and seven times higher, respectively, than observed numbers from these regions (p < .05; Figure 4, Table 2).

| Expected numbers of NIS and their comparison with observed numbers of NIS
The observed number of NIS from the Ponto-Caspian region in the North and Baltic Seas was 14 times higher than expected (p < .05; Figure 4, Table 2). In the case of the Great Lakes-St.Lawrence River, observed numbers of NIS from the North and Baltic Seas were three times lower than expected numbers (p < .05; Figure 4, Table 2). The numbers of observed NIS in the Great Lakes-St.Lawrence River from all other donor regions were higher than expected ones, with Ponto-Caspian species being more than 300 times higher (p < .05; Figure 4, Table 2).

| DISCUSSION
Several studies pointed out that species evolved in the Ponto-Caspian region may be evolutionary predisposed to become NIS (Bij de Vaate Leppäkoski et al., 2002Leppäkoski et al., , 2010Reid & Orlova, 2002;Ricciardi & MacIsaac, 2000); however, the origin of NIS established in T A B L E 2 Estimated average species richness for major donor regions (per 880-km resolution equal-area grids), probabilities of invasion [P(Inv)] for species likely to be transported by ballast water from these regions and established in the North and Baltic Seas or the Great Lakes-St. Lawrence River region, estimated expected number of nonindigenous species (NIS) from major donor regions in the recipient regions, observed number of NIS from major donor regions in the recipient regions, and statistical comparisons of expected and observed numbers of NIS (i.e., chi-square and p-values) are shown. Significant p-values are presented in bold intensively studied in concern to NIS, highly invaded by Ponto-Caspian species, and with different salinity patterns (marine vs. freshwater).
We compared established (observed) to expected numbers of NIS from donor regions and confirmed that there are many more Ponto- to the Great Lakes is a secondary introduction from northern Europe (Leppäkoski et al., 2002;Ricciardi & MacIsaac, 2000), though some species such as the quagga mussel Dreissena rostriformis bugensis likely came to the Great Lakes directly from the Black, Azov, or Caspian Sea (Spidle, Marsden, & May, 1994). As half of the Ponto-Caspian species in the Great Lakes are not established in the North and Baltic Seas, the potential stepping stone dynamics via this region do not provide a parsimonious explanation. Northern European rivers however were not included in our study, and therefore, we cannot confidently disregard the stepping stone hypothesis. Species characteristics and their environmental tolerance, such as the production of dormant stages, r-reproductive strategy (Briski, Ghabooli, Bailey, & MacIsaac, 2011;Briski, Ghabooli, & MacIsaac, 2012), or high phenotypic plasticity, may also explain colonization success of Ponto-Caspian taxa under a lower propagule pressure scenario (i.e., low introduction effort).
Nevertheless, we show here that the Ponto-Caspian region was However, almost all of these species were intentionally introduced fishes (Ricciardi, 2006 Previous studies stated that the transfer of species between different salinity habitats is asymmetrical, with a colonization of freshwater habitats by marine and brackish species becoming increasingly common in recent years, but not vice versa (Grigorovich, Pashkova, Gromova, & van Overdijk, 1998;Lee & Bell, 1999;Sylvester, Cataldo, Notaro, & Boltovskoy, 2013). Ponto-Caspian species originating from brackish areas, with a salinity gradient from freshwater in the east to more saline in the west, accompanied by strong salinity fluctuations (Reid & Orlova, 2002)  The main difference in taxonomic composition of established NIS in the two regions was in the number of established plants.
Tracheophyta was the most represented phylum of NIS established in the Great Lakes-St. Lawrence River (i.e., 30%, 56 species), most likely transported as seeds with solid ballast that was used prior to ballast water (e.g., sand, rocks and mud; Mills, Leach, Carlton, & Secor, 1993;Ricciardi, 2006). However, the phylum was negligible in the North and Baltic Sea region. Lambdon et al. (2008), taking into account the entire area of Europe, also stated that marine habitats are much less invaded by plants than inland waters. Beside the fact that the biodiversity of marine plants is much lower than that of freshwater plants (Les & Cleland, 1997), seeds of the latter are also highly resistant to harsh environments and drying conditions compared to those of the former (Cook, Gut, Rix, & Schneller, 1974;Larkum, Orth, & Duarte, 2007;Leck, 1989;Orth et al., 2000). Environmental tolerance, often dormancy of freshwater seeds, and dates of species discoveries support further the assumptions of solid ballast being the main vector for introduction of these taxa to aquatic habitats (Ricciardi, 2006). After replacement of solid ballast with ballast water at the beginning of the 20th Century, fewer introductions of plants were recorded in the Great Lakes (Ricciardi, 2006).
Taking into account numerous Ponto-Caspian species established in the Great Lakes and Northern Europe, the areas which are greatly connected by shipping to Eastern Asia and coastal North America (Kaluza et al., 2010;Seebens et al., 2013), one would expect Ponto-Caspian species spreading practically all around the world.
However, as the Ponto-Caspian region was not the main donor for the Mediterranean Sea (CIESM, 2015), nor did Ponto-Caspian species establish in high salinities of the North Sea (this study; Paavola et al., 2005), we doubt that Ponto-Caspian taxa may colonize highly saline marine habitats. We suspect that Ponto-Caspian species would colonize big river mouths and estuaries, such as Chesapeake Bay, San Francisco Bay, Yangtze River, and Rio de la Plata. In addition to a salinity match among those regions and the Ponto-Caspian region, very large shipping ports are also located in those areas. Comparative assessment of NIS in multiple regions around the world, including freshwater, brackish, and marine habitats, in connection with transport vectors (i.e., as a proxy for propagule pressure or introduction effort) and species characteristics would elucidate further if Ponto-Caspian species are better colonizers than species evolved in other regions.
However, this comparison would require a huge amount of work and sampling effort to establish reliable and complete lists of NIS, which are lacking for many areas around the world.

ACKNOWLEDGMENTS
Great thanks to Dr. Farrah Chan for producing the salinity map of the