Predicting the ecological impacts of a new freshwater invader

pulex , a trait that became more evident as the invader increased in size. Differences in functional responses and prey selectivity were prey species speciﬁc, with higher to lower predicted impacts in the order A. aquaticus , D. magna and Chironomus sp. This is in accord with the impact of this invasive species on macroinvertebrates in the ﬁeld. 4. We thus provide understanding of the known ecological impact of D. villosus and discuss the utility of the phenomenological use of comparative functional responses and resource use as a tool through which the potential ecological impacts of invasive species may be identiﬁed.

The ability of a novel species to become established within a community may depend on its relative foraging capabilities to pioneer previously un-utilised resources and/or its ability to use resources more efficiently and perhaps compete with resident species for available resources (MacArthur & Levins, 1967;Tilman, 1980). As resource use by invasive species may be much greater as compared to co-evolved resource use relationships of native species (e.g. predator/prey dynamics), the impact of invaders may be assessed by examining such differences (Dick et al., 2013b). Therefore, a potential approach for assessing the ecological impacts of an invading species on a community is the comparison of its rate of resource uptake with that of a trophically analogous native species (Bollache et al., 2008;Dick, Alexander & MacNeil, 2012;Dick et al., 2013a,b). Such predatory capacity, and more generally the use of resources by consumers (see Dick et al., 2013a), can be quantified by measuring the 'functional response' (Solomon, 1949;Abrams, 1990), the relationship between resource consumption rate (e.g. predation rate) and resource density (e.g. prey availability; Holling, 1966;Juliano, 2001). A Type I functional response describes a consumption rate that increases linearly with prey density and is associated with animals utilising a filter feeding mechanism (Jeschke, Kopp & Tollrian, 2004). A Type II functional response describes a consumption rate that increases with prey density and then gradually decelerates to an asymptote as handling time becomes a limiting factor (Holling, 1966). With Type III responses, prey experience a refuge from predation through, for example, the consumer switching to more abundant prey (Holling, 1966), or through refugia provided as a result of habitat complexity . Measurements of functional responses have been suggested as a mechanism through which species with the potential to become damaging invaders could be identified (e.g. Bollache et al., 2008;Dick et al., 2010Dick et al., , 2013a, as the type of functional response (Type II or III) has potential implications for resource (such as prey) population stability, as Type III functional responses are likely to be more stabilising towards prey populations, whereas Type II responses can be destabilising and lead to local extinctions of prey over certain ranges of density (Murdoch & Oaten, 1975;Juliano, 2001).
The form of functional response is not always fixed for a particular pair of interacting species, and many factors such as sediment type (Grant, 1984), light levels (Koski & Johnson, 2002) and habitat complexity  can affect forging success and prey vulnerability and, hence, alter the response type. It is therefore important to establish the form of functional response under variations in key environmental variables. In addition to this, as predators may respond differently to different prey types, empirical measurements of functional responses should be investigated over a range of prey types that encapsulate variations in prey morphology and behaviour. Further, differences in prey selection by predators are known to alter the composition of a community (e.g. Hambright & Hall, 1992;Alto et al., 2009), and differential predation by invasive species in particular has been shown to have dramatic effects on ecosystem function (Vander Zanden et al., 1999). The role of prey selection in the process of invasion may therefore also be particularly important when assessing the potential impacts of an invasive species.
Invasive species impact throughout terrestrial and aquatic environments, but the enhanced innate dispersal capabilities associated with aquatic organisms makes freshwater environments particularly susceptible (Dudgeon et al., 2006;V€ or€ osmarty et al., 2010) and, notably, crustaceans are a particularly successful group at expanding their freshwater ranges (Gherardi, 2007). Dikerogammarus villosus, a freshwater amphipod native to the Ponto-Caspian region of Eastern Europe, has undergone a dramatic range expansion across Western Europe in the last 20 years (Pockl, 2009). Extensive alterations to the structure of communities invaded by D. villosus (e.g. Dick & Platvoet, 2000;Dick, Platvoet & Kelly, 2002;MacNeil et al., 2013) have resulted in the inclusion of this species among the 100 worst invasive species in Europe (www.europe-aliens.org) and it is likely that D. villosus will invade the North American Great Lakes (Ricciardi & Rasmussen, 1998;Bollache et al., 2008). In September 2010, D. villosus was reported in the U.K. (MacNeil et al., 2010), being discovered in a reservoir in the south east of England, in two locations in Wales (South West, U.K.; Madgwick & Aldridge, 2011), and more recently from Barton Broad in East England (Dirk Platvoet, pers comm.). Predicting the likely ecological impacts of this species is thus a high priority to inform management actions.
In this study, we utilised recent advances in the demonstrated predictive power of comparative functional responses (see Dick et al., 2013a,b), to forecast the likely impacts of D. villosus on native freshwater species, and supplemented this with prey selection experiments. Specifically, we investigated the relative predatory capacity of the invader, D. villosus, and an analogous native species, Gammarus pulex, by examining the functional responses and prey selectivity towards three common and representative prey types found in freshwater systems: an isopod, Asellus aquaticus, a cladoceran, Daphnia magna and a dipteran larva, Chironomus sp. Our aims were to establish whether (i) functional responses differ between the invasive and native amphipods (comparing the larger invader with the native; as well, considering body size-matched individuals of the two species); (ii) functional responses are of Type II or Type III, and if these are influenced by environmental heterogeneity in the form of the presence or absence substrate, and; (iii) differences in prey selection exist between the native and invader.

Experimental organisms
Between September and November 2011, in Cambridgeshire (U.K.), the invasive amphipod Dikerogammarus villosus was collected from Grafham Water (Lat: 52 o 18′ 36 N; Long: 0 o 19′06 W) and the native amphipod Gammarus pulex from Duloe Brook (Lat: 52 o 13′60 N; Long: 0 o 22′36W). Juveniles of the isopod Asellus aquaticus were collected from nearby Pitsford Water (Lat: 52 o 19′10 N Long: 0 o 53′35 W), and the cladoceran Daphnia magna and the chironomid Chironomus sp. were bought from a commercial supplier (Livefishfood, Surrey, U.K.). All animals were kept in aquaria with water, substrate and plant material from source locations at 14°C in a 10:14 h light/dark regime for 4 days prior to use in experiments, after which the amphipod predators were killed in 80% ethanol. The length (rostrum to urosome) and constant dry weight of all amphipods were then measured using a microscope and callipers.
Only amphipods free of obvious parasites were used in experiments (e.g. see Dick et al., 2010). Amphipods were divided into three groups based on a visual assessment of body size; large D. villosus, intermediate D. villosus and large G. pulex. As intermediate D. villosus and large G. pulex are of comparable body length, this division allowed size matching of these two groups, thus removing body size as a confounding variable and allowing assessment of inherent species differences in predation rates. However, because D. villosus were generally heavier than G. pulex at equal body lengths (see Fig. 1), we used slightly shorter, heavier D. villosus and slightly longer, lighter G. pulex. One-factor ANOVA thus revealed significant differences in both length (F 2,693 = 1267.6, P < 0.001) and weight (F 2,693 =1037.4, P < 0.001) among the three species/size groups of amphipods, with 'large' D. villosus significantly longer (meanAESE, 19.4 mm AE 0.1) and heavier (28.7 mg AE 0.9) than 'intermediate' D. villosus (14.2 mm AE 0.1; 12.1 mg AE /0.2) and 'large' G. pulex (14.9 mm AE 0.1; 11.2 mg AE /0.1; Fig. 1, all P < 0.001). For intermediate-sized D. villosus and large G. pulex, further analysis by ANCOVA revealed a significant 'species 9 length' interaction effect (F 3,460 =230.1, P < 0.001; Fig. 1), in line with our observation above. We thus used principal components analysis to reduce amphipod length and weight to an index of amphipod body size. The first principal component explained 84% of the variation in amphipod length and weight, providing a very good index of body size. A one-factor ANOVA of the extracted PC1 scores with respect to species revealed no significant difference in the body size of intermediate-sized D. villosus and large G. pulex used in the experiments (F 1,462 =1.29, P = 0.26).

Single-prey experimentspredator functional responses
We presented individual male amphipods with a single-prey species of either A. aquaticus (mean AE SE 3.2 mm AE 0.1), Daphnia magna (3.3 mm AE 0.1) or Chironomus sp. (11.5 mm AE 0.2). A. aquaticus and Chironomus sp. were presented at nine densities (2,4,6,8,10,16,20,30,40 individuals; n = 4 per density), and D. magna at 11 densities (two additional densities of 70 and 140 individuals; n = 4), with or without substrate, in plastic experimental arenas (7.5 cm diameter) with 250 ml of water at 1 : 1 ratio from Grafham Water to Duloe Brook (amphipod source waters). 'With substrate' comprised a 10 mm length of plastic pond weed, anchored in 10 mm of sand (mean particle size 1 mm), on which lay one large stone (mean grain size = 40 mm) and two small stones (mean grain size = 20 mm). Replicates were initiated at 17:00 h with the addition of an individual amphipod (starved for 24 h to standardise hunger) to the arena. Prey were already present in arenas, having been acclimatised for 3 h prior to the start of the trial. Replicates were terminated after 16 h (at 09:00) with the removal of the amphipod, which was then monitored for 24 h to assess survivorship and moulting. Amphipods that died or moulted before, during or within 24 h of the experiment were removed from analyses and the replicate rerun. We counted deaths due to predation as those prey either wholly or partially consumed or bitten to death (see Dick et al., 2002). Control arenas were prey at each density with and without substrate (n = 4 for each combination) without amphipods present. Controls were run in parallel with predation groups.

Mixed prey experimentspredator selectivity
Individual male amphipods (starved for 24 h) were presented with equal proportions of A. aquaticus, D. magna and Chironomus sp. (prey sizes as before) at ten densities (1 of each prey type, 2 of each prey type, up to 10 of each prey type; n = 4 per density). Experimental arenas were as above, with and without substrate and containing 250 ml of mixed amphipod source water. As in the previous experiment, prey were added 3 h prior to the start of the trial at 17.00 h and replicates were terminated after 16 h (at 09:00) with the removal of the amphipod (which was then monitored for 24 h as before). Again, we counted deaths due to predation as those prey either wholly or partially consumed or bitten to death (see Dick et al., 2002). Once again, controls were experimental arenas (with and without substrate) containing prey without predators present.

Statistical analyses
Single-prey experimentspredator functional response. Mean number of prey eaten was examined separately for A. aquaticus, D. magna and Chironomus sp. with respect to three factors [amphipod group (large D. villosus, intermediate D. villosus and large G. pulex), prey density (see above) and substrate type (with/without substrate)] in a general linear model with negative binomial error structure and Tukey post hoc tests.
There are numerous modelling approaches to assess functional responses, and model choice may depend on whether a particular study is mechanistic or phenomenological in approach (Jeschke, Kopp & Tollrian, 2002). Thus, the mechanistic application of parameters such as attack rate and handling time must be approached with caution or be supported with empirical measurements of parameter estimates (Caldow & Furness, 2001;Jeschke et al., 2002;Jeschke & Hohberg, 2008). Phenomenological use of these parameters does, however, provide a tool to examine differences in functional response types and parameter estimates in comparative or factorial experiments and this is the approach taken here (see also Alexander et al., 2012;Dick et al., 2013a). Thus, to determine whether the current predators displayed Type II as opposed to Type III functional responses, we used logistic regression to test for, in the case of Type II responses, a significant negative linear coefficient in the relationship between the proportion of prey eaten and prey density, and in the case of Type III responses, a significant positive first-order term followed by a significant negative second-order term (Trexler, McCulloch & Travis, 1988;Juliano, 2001). As we did not replace prey during the experiments, and consequently prey density declined as prey were consumed, for a Type II functional response, the 'random predator equation' (Rogers 1972) is appropriate (Juliano, 2001): where N e is the number of prey eaten, N 0 is the initial prey density, a is the attack constant, h is the handling time and T is the total time available. Estimated maximum feeding rate was estimated as 1/hT. The Type II functional response was modelled using maximum likelihood estimation (Bolker, 2010). We did not find any Type III functional responses in the present study, but see Alexander et al. (2012) for their modelling. Following the model fitting, bootstrapping was used to generate multiple estimates (n = 30) of the response parameter of maximum feeding rate (1/hT), which was then compared for each prey type separately with respect to amphipod group and substrate conditions (two-factor ANOVA and Tukey post hoc tests). When data were non-normal (Shapiro-Wilks test, P < 0.05) and heteroscedastic (Bartlett's test, P < 0.05), parameter estimates were (x'=log 10 (x + 1)) transformed.
Mixed prey experimentspredator selectivity. The proportion of each prey type eaten relative to the total number of prey items provided was calculated and then reduced to an index of prey selectivity using principal components analysis (PCA). We tested prey selection differences for all amphipod groups based on the first two extracted principal component scores with respect to amphipod group, prey density and substrate conditions (three-factor ANOVA and Tukey HSD post hoc tests).
The origin of the two PC axes represents the point in PC space where there is no prey selection (i.e. all prey types are eaten in equal proportion), thus increasing distance from the origin is representative of a move from an indiscriminate feeding strategy (i.e. no prey selection) to a selective feeding strategy (i.e. selection for specific prey type). Feeding strategy was measured as the distance from the origin to each point in PC space (i.e. the PC1, PC2 co-ordinate). We tested feeding strategy with respect to amphipod group, prey density and substrate conditions (three-factor ANOVA and Tukey HSD post hoc tests).
All statistical analyses were performed in R, version 2.13.1 (R Development Core Team, 2010). . The majority of deaths in experimental arenas were thus the result of amphipod predation. This was further evidenced through observations of direct predation by both predator species as well as the presence of partly consumed prey in experimental arenas following the removal of the amphipod.

Single-prey experimentspredator functional responses
Prey: Asellus aquaticus. The minimum model revealed a significant two-way interaction between amphipod species and A. aquaticus density (Table 1; Fig. 2a). This reflected the increased disparity between amphipod species in mean prey consumed at higher densities (Fig. 2a).  Fig. 2b). Logistic regression revealed significant negative estimates of the linear coefficient for all predator/A. aquaticus prey groups (except for G. pulex with substrate present; see Table 2); therefore, most amphipod groups exhibited Type II functional responses (Fig. 3a). Large D. villosus had a significantly greater maximum feeding rate compared with both intermediate D. villosus and G. pulex (between which there was no difference, Table 3, Fig. 4a). The significant 'amphipod group x substrate' interaction (Table 3, Fig. 4a) indicated higher feeding rates of D. villosus compared with G. pulex where no substrate was present, but the opposite when substrate was present (Fig. 4a).
Prey: Daphnia magna. The minimum model revealed a significant two-way interaction between amphipod species and D. magna density (Table 1; Fig. 2c). This reflected the increased disparity between amphipod species in mean prey consumed at higher densities (Fig. 2c). Significant differences in mean prey consumed between the native and invasive species as a result of body size differences became apparent at prey density of 30 (Z = 3.804, P = 0.043) and continued up to a prey density of 140 (40, Z = 3.844, P = 0.037; 70, Z = 5.097, P < 0.001; 140, Z = 4.868, P < 0.001). The presence of substrate did not significantly influence the mean number of D. magna consumed (Table 1).
Logistic regression revealed significant negative estimates of the linear coefficient (Table 2); therefore, all predator prey groups exhibited Type II functional responses (Fig. 3b). Both size groups of D. villosus had significantly greater maximum feeding rates than G. pulex and large D. villosus significantly greater than intermediate D. villosus (Table 3; Fig. 4b). There were no differences in maximum feeding rate in the presence or absence of substrate, and there was no significant 'amphipod group x substrate' interaction (Table 3).
Prey: Chironomus sp. The minimum model revealed no significant interactions between amphipod, prey density and substrate type. There was a strong trend for a significant difference in the mean number of Chironomus sp. consumed among amphipod groups (  Fig. 2f).
Logistic regression revealed significant negative estimates of the linear coefficient (Table 2); therefore, all predator/Chironomus sp. groups exhibited Type II functional responses (Fig. 3c). Maximum feeding rate of large D. villosus and G. pulex was significantly higher than intermediate D. villosus, and there was no significant difference between the maximum feeding rate of large D. villosus and G. pulex (Table 3; Fig. 4c). There were no differences in maximum feeding rate in the presence or absence of substrate, and there was no significant 'amphipod group x substrate' interaction (Table 3). The first and second scores from the PCA of prey selectivity explained 59 and 39% of the variation, respectively, accounting for a total of 97% of the total variation in prey selectivity. PC1 was positively loaded (+0.739) for D. magna selection and negatively loaded for Chironomus sp. (À0.605) and A. aquaticus (À0.296); thus, a large positive PC1 score was indicative of predatory selection for D. magna, while a small PC1 score was indicative of  Fig. 5). Overall, G. pulex exhibited a significantly more positive selection for pelagic prey types compared with intermediate D. villosus (Fig. 5). There was, however, no significant difference in benthic/pelagic prey selection between G. pulex and large D. villosus (Table 4), and there were no differences between the two sizes of D. villosus (Fig. 5). There was a significant effect of substrate presence on benthic/ pelagic prey selection, with a significant change from a benthic prey selection in the absence of substrate to a pelagic prey selection in the presence of substrate  ( Fig. 5). Increasing prey density moved predatory selection towards benthic prey types (Fig. 5).

Mixed prey trials
The three-way ANOVA of PC2 revealed significant differences in mean prey selectivity for the two benthic prey (A. aquaticus and Chironomus sp.) among amphipod groups (Fig. 5). G. pulex exhibited significantly greater selection for Chironomus sp. compared with both D. villosus groups (Fig. 5). There was no difference in selection for benthic prey between the two D. villosus groups. A significant 'amphipod group x prey density' interaction reflected the greater increasing selection for Chironomus sp. at higher prey density by G. pulex compared with the two D. villosus groups, which exhibited a greater selection for A. aquaticus at high prey density. The presence/absence of substrate had no statistically significant effect on differences in benthic prey selectivity.
Feeding strategy (measured as the distance from the origin of PC1 and PC2) differed significantly among amphipod groups, under different substrate conditions and with prey density (Table 4). G. pulex was significantly more selective compared with both groups of D. villosus (Fig. 6) that were significantly more indiscriminate in their feeding strategy, and large D. villosus was more indiscriminate than intermediate D. villosus (Fig. 6a). Amphipods showed a significantly more selective feeding strategy in the presence of substrate (Table 4, Fig. 6b) and with increasing prey density (Fig. 6c). There was a significant 'amphipod group x substrate' interaction (Table 4) reflecting the greater Table 4 ANOVA models for prey selection (principal component scores) and predatory generalism (see text for details) with amphipod group, supplied prey density and substrate conditions (amphipod = amphipod group (three levels), density = prey density (nine levels for Asellus aquaticus and Chironomus sp., and 11 levels for Daphnia magna), substrate = substrate conditions (two levels) and '*'denotes an interaction). Non-significant terms are detailed in greyed italics disparity between large D. villosus and G. pulex compared with intermediate D. villosus, which was less likely to change feeding strategy in the presence of substrate. There was a significant 'prey density x substrate' interaction (Table 4) indicating a greater difference in selective feeding at higher prey densities in the absence of substrate compared with that of in the presence of substrate.

Discussion
The development of tools that can forecast the ecological impacts of invasive species on recipient communities is a major objective of invasion ecology research that has seen limited success (e.g. see Ricciardi, 2003;Lockwood, Hoopes & Marchetti, 2007;Davis, 2009;Dick et al., 2013a,b; but see Nentwig, Kuhnel & Bacher, 2009). In this study, we make use of a comparative functional response methodology to assess relative use of resources by invasive and native species, as well as examining prey selectivity, to predict the likely ecological impacts on native prey of the 'killer shrimp', Dikerogammarus villosus, newly invasive in the U.K. (MacNeil et al., 2010) and likely to invade outside of Europe, such as the North American Great Lakes (Ricciardi & Rasmussen, 1998;Bollache et al., 2008). Under our experimental conditions, both the invader D. villosus and the native G. pulex exhibited Type II functional responses towards three prey species, Asellus aquaticus, Daphnia magna and Chironomus sp. Furthermore, this form of response was conserved with the addition of substrate, counter to a number of studies that report a shift to Type III Predator generalism (measured as distance from the origin, increasing distance from the origin equates to a move from indiscriminate feeding to selective feeding) differences (a) among amphipod groups; (b) in the absence or presence of substrate and; (c) with increasing supplied prey density.
functional responses in gammarids and other predators under such conditions of habitat heterogeneity (see Alexander et al., 2012). This may be important for prey at the population level, as Type III functional responses tend to be stabilising, whereas Type II responses may destabilise prey populations over certain ranges of density (Murdoch & Oaten, 1975;Juliano, 2001). Such functional response results are congruous with the known ecological impacts of both species, which show negative abundance relationships with many macroinvertebrates that have led to local and regional extinctions (Dick & Platvoet, 2000;Kelly et al., 2006;MacNeil et al., 2013). The invader, D. villosus, had significantly higher functional responses, with greater maximum feeding rates, towards both A. aquaticus and D. magna. There was, however, less of a difference between the invasive and native amphipod functional responses towards Chironomus sp. We thus predict greater ecological impacts for the former two prey species than for the latter. Indeed, there is some field evidence that the presence of Asellidae is more affected than other taxa, including Chironomidae, when D. villosus invades (MacNeil et al., 2013). The overall higher predatory rate of D. villosus may be attributed to a number of factors, including relatively larger antennae and mouthparts (Platvoet, 2007;Mayer et al., 2009;Stoffels et al., 2011), differences in physiology (Maazouzi et al., 2011) and resource assimilation rates (Gergs & Rothhaup, 2008). In addition to this, D. villosus attains a greater maximum size in comparison with G. pulex (the former can be approximately 20% longer and twice as heavy); thus, species specific differences are amplified by the greater maximum body size of the invader. D. villosus also demonstrates a tendency for partial predation of prey, a phenomena that has been observed and photographed in this species (see Dick et al., 2002). The greater maximum feeding rate of G. pulex compared with D. villosus towards A. aquaticus in the presence of substrate may be indicative of different hunting strategies, such as active searching versus sit-and-wait, but further research is required to unravel this.
Our prey selectivity experiment further highlighted the differences in predatory behaviour between the invasive and native amphipods. Compared with the native amphipod G. pulex, D. villosus was more selective of the benthic prey and was specifically more selective of A. aquaticus, a feeding strategy that was amplified at higher prey densities. This differential prey selectivity for A. aquaticus, coupled with the greater maximum feeding rate of the invader on this prey type, indicates that any impact on A. aquaticus populations would, under natural conditions, likely be amplified to a greater degree when compared with Chironomus sp. populations following invasion. In fact, Chironomus sp. populations might not be especially impacted following the invasion of D. villosus, given the similarity in maximum feeding rate between the native and invasive species and the lack of positive selection for Chironomus sp. by the invader. J. Dodd (personal observation) found that in two adjacent, uninvaded reservoir systems, A. aquaticus represented 34-64% and Chironomus sp. represented 1-11% of the biomass of the macroinvertebrate community. This was in stark comparison with a complete lack of detection of both A. aquaticus and Chironomus sp. in Grafham water, the invaded reservoir, where 97% of the biomass of the macroinvertebrate community was represented by D. villosus. This pattern of field observation supports the results reported in this study and could tentatively indicate potential effects on ecosystem function. The increased predatory impact on A. aquaticus by D. villosus in invaded lake systems could result in changes to energy transfer in food webs (MacNeil et al., 2011). A. aquaticus has been described to use a wide resource base (Moog, 2002), but are generally described as detritivores (Adcock, 1979). Their role in the food web is the facilitation of energy transfer between trophic levels through the processing of allochthonous material (Adcock, 1979). The mechanisms through which some types of this material are processed have been shown to differ between A. aquaticus, G. pulex and D, villosus, with the latter showing a much lesser processing efficiency than the two former species (MacNeil et al., 2011). A. aquaticus, D. magna and Chironomus sp. are also prey for lake dwelling fish species, and both Cladocerans and Chironomidae form a large component of fish diet in Grafham Water (Lindsey & Lowe, 2001). The routes of energy transfer within food webs form the basis of how an ecosystem functions (Hooper et al., 2005); thus, changes in the energy transfer route may have serious consequences on the stability (Hooper et al., 2005) and resilience (Richmond, Breitburgh & Rose, 2005) of a system invaded by D. villosus. Indeed, both the change in the availability of A. aquaticus, D. magna and Chironomus sp. and the increased availability of D. villosus as an alternative food resource, has the potential to drive evolutionary change within some species of lake dwelling fish, for example Arctic charr, Salvalinus alpinus, which have been shown to be particularly susceptible to resource use-driven speciation (Adams & Huntingford, 2004;Knudsen et al., 2011).
Predatory differences between the invader and native amphipods on the pelagic D. magna are complex, with the invader showing a greater maximum feeding rate compared with the native species when presented with a single-prey type; however, when presented with multiple prey types, the invasive D. villosus exhibited lower selection for this prey species when compared with G. pulex. It is therefore likely that potential amphipod impacts on D. magna under natural conditions are likely to show greater variation depending on the availability of other food resources.
In addition to differences in specific prey types, D. villosus also exhibited differences in feeding strategy when compared with G. pulex. D. villosus was significantly more indiscriminate in prey selection, a tendency that became stronger as the invader increased in size. The indiscriminate use of available resources has been highlighted as another trait that may confer an advantage to invasive species (Romanuk et al., 2009;H€ anfling, Edwards & Gherardi, 2011;Keller et al., 2011), and the reasons surrounding such an increase in generalist feeding ability of D. villosus are likely to be similar to those physical advantages conferring a greater maximum feeding rate on this species, such as larger, more powerful mouth parts and larger antennae, as detailed above.
The combination of functional response studies and prey selection experiments has the potential to not only indicate those native species most at risk of impact following invasion by a novel species, but also the degree to which such groups are likely to be affected. There is growing support that functional response analysis in particular provides reliable predictions of such invader impact (see Dick et al., 2013a,b) and has indicated that invasive species may in general have higher functional responses compared with native species (e.g. for parasitoids; Greenberg, Legaspi & Jones, 2001;Jones et al., 2003). While direct comparisons of invasive and native species functional responses are rare, Haddaway et al. (2012) showed that an invasive crayfish has a higher functional response than a native, although this was not directly related to field impacts on prey. However, most recently, Dick et al. (2013a) show that the invasive 'bloody red shrimp', Hemimysis anomala, has a higher functional response than analogous native species and that the greatest differentials in functional responses were associated with the prey that suffered the greatest field impacts. Further, Dick et al. (2013a) show that this difference in functional responses is consistent across the geographical range of the invader. Functional response techniques can offer some advantages over trait-based predictions (e.g. see Sakai et al., 2001), by providing predictions of the potential consequences for specific prey. However, it is likely that the best information will be generated using these techniques in concert; trait-based information can be gathered simply (e.g. Kolar & Lodge, 2001) and inform which species need further investigation through functional response models.
Risk assessments for invasive species require some element of likely ecological impact, but without an invasion (and hence impact) history, this is difficult to derive (see Leung et al., 2012). Comparative functional responses have been utilised in the field of biocontrol to assess the efficacy of native and introduced biocontrol agents (Fernandez- Arhex & Corley, 2003;Madadi et al., 2011). We suggest that comparative functional responses provide a powerful route to investigate the impact of existing, emerging and potential invasive species. Furthermore, functional responses and resource selection can be derived for consumers other than predators (e.g. Hobbs et al., 2003;Sarnelle & Wilson, 2008), their derivation can be in the laboratory or field, and the method is widely applicable across taxonomic and trophic groups (see also Dick et al., 2013a,b). Ultimately, further exploration of these ideas could move invasion ecology from a descriptive to a more predictive science.