Sea freshening may drive the ecological impacts of emerging and existing invasive non‐native species

The spread of invasive non‐native species (INNS) will pose major threats to global biodiversity over the coming decades. However, predicting how key effects of climate change will influence the abilities of INNS to establish and exert ecological impact is a major challenge. One overlooked aspect of global change is the expected freshening of certain marine systems, which may interact with INNS and lead to drastic effects on community structure and stability.

ingly connected transport networks facilitating invasions worldwide (Hulme, 2009;Seebens et al., 2019). While many INNS fail to establish (Williamson & Fitter, 1996), many spread and exert severe impacts, affecting biodiversity, ecosystem function, human, animal and plant health, and global food security Mazza et al., 2014;Paini et al., 2016). However, with the number of species introductions unabating (Seebens et al., 2018), predicting how such impacts are affected by other major threats to global biodiversity, particularly climate change (Thomas et al., 2004), and the vast associated suite of biotic and abiotic consequences (Brook et al., 2008), is a vital, albeit difficult task (Urban, 2015).
While warming, ocean acidification and changing weather patterns are widely studied and documented (Harley et al., 2006), the ecological effects associated with salinity shifts of sea water remain understudied (Illing et al., 2016). This has been described as a "rich get richer" mechanism (Chou et al., 2009), whereby highly saline marine regions are getting saltier, and relatively fresh regions are getting fresher (Durack et al., 2012). Such events could have severe consequences globally, with the greatest impact of freshening likely to occur in coastal and partially enclosed fjordic systems (Convey & Peck, 2019). Indeed, freshening has triggered mass mortality events for amphipods in Arctic waters (Eiane & Daase, 2002) and shifts from a krill-dominated system to a salp-dominated system off the West Antarctic Peninsula (Ballerini et al., 2014;Deppeler & Davidson, 2017). Further, periods of low salinity have in the past been shown to reduce the abundances and spatial distributions of a number of species in the Baltic Sea (e.g. Ojaveer & Kalejs, 2005).
How physiological stress resulting from salinity changes will differentially affect INNS and native species is currently unknown and requires urgent assessment and prediction to forecast the identities of likely future INNS, their potential impacts and any effective mitigation strategies.
The Ponto-Caspian region is a donor hotspot for INNS , with many euryhaline species deemed pre-adapted to invade and establish in new environments (Casties et al., 2016;Paiva et al., 2018;Pauli et al., 2018). Indeed, many Ponto-Caspian INNS can thrive in areas of anthropogenic alteration, including pollution and large salinity ranges, at the expense of natives (Den Hartog et al., 1992). Amphipod crustaceans are one group of particularly successful INNS, with many Ponto-Caspian species undergoing drastic range expansions over the last two decades (Clinton et al., 2018;Cuthbert et al., 2020;Grabowski et al., 2006). Amphipods are major drivers of disturbance through predation, herbivory, competition for substrate and modification of sediment (Conlan, 1994), and their invasions have lead to major changes in the faunal make-up of the systems in which they establish (Dick & Platvoet, 2000;Jazdzewski et al., 2004;Kelly et al., 2006). Predicting the likely identities of future INNS and recipient areas at risk are major goals of invasion ecology (Gallardo et al., 2016;Lucy et al., 2020;Peyton et al., 2019;Roy et al., 2014) and one Ponto-Caspian amphipod expected to spread through Europe in the near future is Pontogammarus maeoticus (Baltazar-Soares et al., 2017). Endemic to the Caspian, Black and Azov Seas (Stock et al., 1998), this species has a limited documented invasion history, with only some reports from Turkey and Ukraine in recent decades (Ahmet et al., 2003;Alexandrov et al., 2007).
However, an INNS that has already arrived and established in Europe is G. tigrinus from North America, and it has been cited as a cause of reduced native species abundances (Grabowski et al., 2006).
Relative to trophically analogous natives, G. tigrinus tends to have greater salinity tolerance (0-25 PSU: Grabowski et al., 2007), more generations per year and lower susceptibility to human impacts such as pollution and habitat degradation (Grabowski et al., 2007).
One representative system potentially at risk from P. maeoticus, and where G. tigrinus has already established, is the Baltic Sea in Northern Europe, which has been deemed especially sensitive to salinity changes (Meier & Kauker, 2003). The Baltic Sea has shorelines on nine countries and is subject to high volumes of shipping traffic, and approximately one hundred INNS have been recorded there (Casties et al., 2016;Leppäkoski et al., 2002). This large, semi-enclosed brackish-water sea area has a salinity range between 2 and 24ppt (Leppäkoski et al., 2002) due to a large freshwater supply, a narrow and shallow connection with the North Sea, and the mixing of outflowing brackish water with salty inflowing water (Rodhe & Winsor, 2002). Since the late 1970s, there has been a prolonged period of freshening (Ojaveer & Kalejs, 2005) and this is expected to continue, with salinities of 10ppt likely to become increasingly common (Vuorinen et al., 2015), and Kiel Fjord, to the west of the Baltic, predicted to see a ~ 2ppt decrease of salinity to less than 13ppt by the end of the century (Gräwe et al., 2013).
Here, we assess the effect of decreasing salinity on the predatory impacts of three focal amphipod species: the potential Ponto-Caspian INNS, P. maeoticus, and two of the most common amphipods in the northern Baltic, the established North American INNS, G. tigrinus, and the Baltic native and trophically similar G. salinus (Kotta et al., 2011). We use the comparative functional response method (CFR: Cuthbert et al., 2019;Dick et al., 2014;, which asseses ecological impact by quantification and comparison of the effect of prey density K E Y W O R D S functional response, invasive non-native species, life history traits, Pontogammarus maeoticus, relative impact potential, sea freshening on prey consumption rates (see Holling, 1959;Solomon, 1949), while allowing the incorporation of a wide range of biotic and abiotic contexts (e.g. oxygen: habitat complexity: Cuthbert et al., 2019;temperature: Wasserman et al., 2018; parasites . We then use the Relative Impact Potential metric (RIP: Dickey et al., 2020), which improves the predictive power of the CFR method by combining FR parameters with proxies of the consumer numerical response (NR), such as consumer abundance, density or certain life history trait-based measures (e.g. see Dickey et al., 2018), to establish present and future relative impacts of the three predator species under sea freshening.

| Specimen collection and maintenance
Specimens of P. maeoticus were collected in October 2014 in Jafrud, Iran (37°37' N 49°07' E), transported to Kiel, Germany, and kept in laboratory at 18°C and 10ppt. Gammarus tigrinus and G. salinus were collected in August 2017 in Travemünde, Germany (53°83' N 10°64' E) and Kiel, Germany (54°40' N 10°20' E), and kept at 16˚C, and 10ppt and 16ppt, respectively. Salinities and temperatures were determined based on conditions of the collection sites. All three species were held in constantly aerated 56 L glass aquaria, filled with 5-μm filtered Kiel fjord water, with salinity being adjusted by adding artificial seawater (System Instant Ocean®) or potable tap water.
Sand and artificial structure, such as ceramic tubes, were added to the tanks to simulate natural habitats. The animals were fed ad libitum with a mixture of commercial crustacean food (Tetra Mix, Tetra Crusta, and Dr. Shrimp Healthy), while the light/dark cycle was 12:12 hr.

| Functional response experiments
Experiments were conducted between 27 February and 6 March 2018, with the three amphipod species examined concurrently and with Artemia franciscana as prey. Two weeks prior to the experiments, 20 size-matched individuals of each of the three amphipod species were selected from their holding aquaria based on head to pleon length (mean ± SE: P. maeoticus, 13.602 ± 0.294 mm; G. tigrinus, 13.116 ± 0.331 mm; G. salinus, 13.273 ± 0.296 mm), and acclimated to a laboratory temperature of 17(±1)°C, which is currently common in shallow areas of the Baltic Sea and projected to become widespread by the end of the century (Holopainen et al., 2016).
Two experimental salinities were chosen, that is 16ppt and 10ppt, to reflect a common current salinity on the western Baltic shoreline and an expected future freshened level, repectively (Vuorinen et al., 2015). Amphipod species were housed at densities of 10 individuals per 2 L plastic aquarium (i.e. two aquaria per species) with aerated water, with ceramic tubes for habitat. Each species was fed twice per day with food pellets (see above), and given half water changes daily.
For P. maeoticus and G. tigrinus, individuals were adapted to 16ppt from initial holding salinities of 10ppt, and for G. salinus, individuals were adapted from 16ppt to 10ppt. In each case, salinities were changed by 2ppt per day towards the target salinities for three days. For individuals with an experimental salinity equal to their holding tank salinity, that is P. maeoticus and G. tigrinus at 10ppt, and G. salinus at 16ppt, water of the same salinity was added to standardise physical disturbance across species and salinities. The water used for experiments was a combination of water from Kiel Fjord and potable tap water, both filtered through a 5-µm filter and mixed to obtain allotted salinities.
The prey, Artemia franciscana (5-7 mm), was obtained commercially from Fischfutter Etzbach, Gemünd and maintained in the same laboratory as the predators (see before). Prey was collectively acclimated to the same conditions as predators from an initial salinity of 40ppt (i.e. supplier level) in two stages. First, all individuals were adapted to 30ppt on the day of purchase, 25ppt on the second day and 20ppt on the third day. Second, prey was then segregated, with half the supply acclimated to 16ppt and the other half to 10ppt via one further reduction each on the fourth day. All prey individuals were then given at least two days to acclimate to their experimental salinities. This prey was chosen as a commercially available species tolerant of a wide range of salinities, which had high survival and exhibited normal behaviour throughout experimentation and represented a general, readily consumed prey item (MacNeil et al., 1997).
Feeding experiments were conducted in 1L plastic jars filled with 700 ml of either 16ppt or 10ppt water that had been aerated for 24 hr prior and ordered at random. Five densities of prey were supplied, that is 2, 4, 8, 16 and 32 (n = 6 per prey density, per experimental group) and allowed to settle for 30 min, with trials commencing upon the addition of a single predator. Trials lasted for six hours and the number of live prey was recorded at the end of this time to enumerate numbers consumed, following removal of predators. Controls for each prey density at both salinities (n = 3 per prey density, per experimental group) were used to quantify any background mortality levels in the absence of amphipod predators. Remaining live prey after the experiment were further classified as "free-swimming" or "wounded" (i.e. at the bottom of the experimental arenas, moving thoracopods but unable to enter water column). Dissolved oxygen levels, measured using ProfiLine Oxi 3205 probe (WTW, Germany), did not fall below 85% saturation over the six-hour period. If any of the predators moulted during the experiments, the given treatment replicate was repeated with another inter-moult animal.
Due to the limited number of individuals of each amphipod species, size-matched individuals were re-used a maximum of three times (see Alexander et al. 2014). No individuals at either salinity were exposed to the same density of prey more than once, and a 48 hr recovery period was allowed between trials. To ensure no individuals were given the same prey density, all amphipods were held individually within the larger holding aquaria in 50ml test tubes with a mesh top (to facilitate dissolved oxygen diffusion), containing a ceramic tube for habitat, to track identity.

| Statistical analysis
All analyses were carried out in R v.3.2.2 (R Development Core Team 2015). Functional responses (FRs) were modelled using the "frair" package (Pritchard et al., 2017), and the type of curve (Type I, II or III) was derived through logistic regression of the proportion of prey consumed as a function of prey density. A significantly negative firstorder term indicates a Type II FR, whereas a significantly positive first-order term, followed by a significantly negative second-order term, is indicative of a Type III response (Juliano, 2001). Functional responses of each species at each salinity were modelled using maximum likelihood estimation (MLE; Bolker et al., 2009) and the random predator equation (Rogers, 1972), due to prey not being replaced as they were consumed: where N e is the number of prey consumed, N 0 is the initial density of prey, a is the attack rate, h is the handling time and T is the total time available (i.e. six hours). A second FR was calculated whereby N e represented the sum of the number of prey consumed and the number of prey wounded (described above), as such prey individuals are unlikely to survive and reproduce, that is are effectively removed from the prey population. Both models were fit to the data using the Lambert W function owing to the recursive nature of the random predator equation (Bolker, 2008). The initial a and h estimates were non-parametrically bootstrapped (n = 2000) to construct 95% confidence intervals around the functional response curve for each treatment.
The potential ecological impact of an INNS under context-dependencies can be predicted using the Impact Potential (IP) metric Dickey et al., 2020), calculated by taking the product of the predator FR and a proxy of the predator numerical response (NR): Here, we used the FR estimate of "maximum feeding rate" (curve asymptote), calculated as the inverse of handling time (1/h: Dick et al., 2014), as derived in the above experiment at salinities of 16ppt and 10ppt. This FR measure was combined with two life history trait proxies of the NR (see Table 1), deemed highly predictive of successful invasive gammarids (Grabowski et al., 2007). Firstly, IP was derived using the Partial Fecundity Index (PFI; Table 1): whereby PFI is calculated as follows:  & Marques, 2003;Neuparth et al., 2002;Xue et al., 2013), and thus these fecundity values were kept constant across the two salintities of the present study. RIP biplots Dickey et al., 2020; were created with "maximum feeding rate" on the x-axes and the above NR proxies (from Equations 4 and 6) on the y-axes for comparison among amphipod species at each salinity level, whereby ecological impact increases from the bottom left to top right.

| RE SULTS
Prey survival in all controls was 100%, and thus experimental consumption did not require adjustment for background prey mortality. For both prey consumption alone and for prey consumption plus prey wounding, Type II FRs were exhibited by all three amphipod species under both salinity treatments, as determined by significantly negative first-order terms (Table 2, Figure 1). The functional response curves of P. maeoticus and G. tigrinus heightened under reduced salinities, while that of G. salinus lowered (Figure 1). This was driven by both INNS exhibiting lower handling times h (and hence higher maximum feeding rates, 1/h) with decreased salinity, whereas the native exhibited lower attack rates and higher handling times (and hence lower maximum feeding rates) with decreased salinity (Table 2; Figure 1; Figure 2). Of the three study species, the maximum feeding rate of G. tigrinus was most greatly heightened by accounting for prey wounding (Table 2; Figure 1; Figure 2).

| D ISCUSS I ON
Understanding how the myriad consequences of climate change are likely to affect the ecological impacts exerted by invasive alien species (INNS) is a pressing concern for biodiversity conservation globally (Hellmann et al., 2008;Mainka & Howard, 2010

TA B L E 2
First-order terms derived from logistic regression of a) the proportion of prey consumed, and b) the proportion of prey consumed and wounded, as a function of prey density, with parameter estimates from Rogers' random predator equation  From the CFR aspect of the experiment, we quantified both "prey consumed" and "prey consumed and wounded" across a range of prey densities. To date, FR experiments have tended to only quantify the number of prey killed (Dick et al., 2014), and this may underestimate the impact a predator exerts on prey populations as unaccounted wounded prey are unlikely to survive or reproduce in the long term. Here, using both FR measures, we found that both P. maeoticus and G. tigrinus had higher maximum feeding rates at the lower salinity, whereas the maximum feeding rate of G. salinus decreased with freshening. These results strongly suggest heightened predatory impacts of the potential and established INNS under future reduced salinity conditions, as differential laboratory FRs are strongly linked to differential ecological impacts in the field (Dick et al., 2013;. While the above CFR method is highly effective at highlighting the role played by abiotic conditions on predatory impact, the need to incorporate proxies of the consumer numerical response (NR) into impact quantification has been highlighted Dickey et al., 2020), as their addition offers greater predictive power when assessing overall INNS impacts. That is, the total impact of a species is the product of the per capita effect of individuals and some measure of the number of individuals in the consumer population having those individual effects Dickey et al., 2020). For that reason, we compared the Impact Potentials of the three species, defining impact as the product of per capita effect (specifically the maximum feeding rates derived from the CFR experiments) and relevant life history traits related to the NR. While NR proxies such as abundance and density have been the default in the past , this practice is all but impossible when potential INNS with limited invasion history are being assessed (see Dickey et al., 2018Dickey et al., , 2020 (Grabowski et al., 2007), and here our measure of impact combined both (i.e. life history traits as the NR proxy, tolerance for different salinities incorporated within the maximum feeding rates) to assess potential impact. We used the partial fecundity index (PFI) of Grabowski TA B L E 3 Impact Potential (IP) calculations accounting for a) prey consumed, and b) prey consumed and wounded, whereby IP PFI = maximum feeding rate × Partial Fecundity Index, and IP APFI = maximum feeding rate × Annual Partial Fecundity Index and O. crassus as the most invasive Ponto-Caspian amphipods in the Baltic, and the fact P. maeoticus outnumbers the latter in the Caspian Sea (Mirzajani, 2003) may strengthen our assessment.
Gammarus tigrinus is regularly highlighted as a damaging INNS, with evidence suggesting it has been expanding its range rapidly over recent decades in the Baltic Sea (Herkül & Kotta, 2007) and beyond (Platvoet et al., 2009). Here, the ability of G. tigrinus to combine high feeding rates with high reproductive output means it had the highest IP PFI and IP APFI at both salinities, indicative of high impact in the Baltic Sea currently, and with increasing impact expected with freshening. While certain aspects of its success were not assessed in this study (e.g. aggressive and predatory behaviour towards native amphipods: Dick, 1996;Kotta et al., 2010), the high ecological impact displayed in our IP biplots corroborates its notoriety as a highly damaging INNS. Particular concern should surround its propensity to "wound" prey, and by accounting for this in our "consumed and wounded" biplots, we highlight the potential impact to prey species populations beyond direct consumption. Using both "consumed" and "consumed and wounded" measures, we see heightened predation rates at the lower salinity, indicating potential for enhanced impact in a freshening system for a species already linked to declining abundances of native amphipods, such as G. salinus (Orav-Kotta et al., 2009). Paterson et al. (2015), using the measure of "partial consumption", also found that different prey species can elicit different consumptive behaviours from amphipod predators, which they explained as interference caused by high densities of active prey interrupting predator feeding. However, that study did not attempt to assess a predator species effect. Here our "consumed and wounded" biplots demonstrate an enhanced disparity between G. tigrinus and the other study species, and highlight different species-specific foraging behaviours, possibly indicative of G. tigrinus selectively targeting the most nutritious parts of prey (Paterson et al., 2015).
Amphipod INNS are expected to continue to spread around the coastal areas of the Baltic Sea (Holopainen et al., 2016), and beyond (Grabowski et al., 2007;Son et al., 2020), with changing temperature and salinity conditions likely to further enhance ecological impacts.
Intraguild predation is a common feature of coexisting amphipod species (Dick et al., 1999;MacNeil et al., 2004); however, the influence of climate change on intraguild predation has received little focus to date (Brambilla et al., 2019). While some studies have found differential effects of water conductivity on the degree of intraguild predation between amphipod species (Dick & Platvoet, 1996;Kestrup et al., 2011) (Aronson et al., 2015). The adaptability of certain INNS (Stern & Lee, 2020) combined with changing salinities in aquatic systems globally means that potential future INNS, as well as established non-natives and native species, need to be subject to relative impact assessments across further abiotic contexts, as per this study.
Quantification of how INNS impacts is mediated by less conspicuous regime shifts associated with global climatic change requires urgent consideration by scientists and practitioners. We propose that our new metrics, such as the impact potential metric, offer user-friendly and informative means of assessing and, crucially, predicting said impacts, regardless of the invasion history of the species.

ACK N OWLED G M ENTS
JWED, JTAD and EB conceived the study, with JWED conducting the experiments, with assistance from GTS. JWED and RNC conducted statistical analyses, with JWED, RNC and JTAD preparing the initial manuscript. All authors provided valuable input to the development of the final manuscript and have given approval for publication. JWED was supported by Inland Fisheries Ireland (IFI), RNC by the Alexander von Humboldt Foundation and EB by the Alexander von Humboldt Sofja Kovalevskaja Award.

PEER R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1111/ddi.13178.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available