Invasion success depends on invader body size in a size-structured mixed predation competition community

Transcription

1 Journal of Animal Ecology 2009, 78, doi: /j x Invasion success depends on invader body size in a size-structured mixed predation competition community Arne Schro der*, Karin A. Nilsson, Lennart Persson, Tobias van Kooten and Birte Reichstein Department of Ecology and Environmental Science, Umea University, Umea SE-90187, Sweden Summary 1. The size of an individual is an important determinant of its trophic position and the type of interactions it engages in with other heterospecific and conspecific individuals. Consequently an individual s ecological role in a community changes with its body size over ontogeny, leading to that trophic interactions between individuals are a size-dependent and ontogenetically variable mixture of competition and predation. 2. Because differently sized individuals thus experience different biotic environments, invasion success may be determined by the body size of the invaders. Invasion outcome may also depend on the productivity of the system as productivity influences the biotic environment. 3. In a laboratory experiment with two poeciliid fishes the body size of the invading individuals and the daily amount of food supplied were manipulated. 4. Large invaders established persistent populations and drove the resident population to extinction in 10 out of 12 cases, while small invaders failed in 10 out of 12 trials. Stable coexistence was virtually absent. Invasion outcome was independent of productivity. 5. Further analyses suggest that small invaders experienced a competitive recruitment bottleneck imposed on them by the resident population. In contrast, large invaders preyed on the juveniles of the resident population. This predation allowed the large invaders to establish successfully by decreasing the resident population densities and thus breaking the bottleneck. 6. The results strongly suggest that the size distribution of invaders affects their ability to invade, an implication so far neglected in life-history omnivory systems. The findings are further in agreement with predictions of life-history omnivory theory, that size-structured interactions demote coexistence along a productivity gradient. Key-words: coexistence, competitive recruitment bottleneck, life-history omnivory, ontogenetic niche shift, productivity gradient Introduction The majority of animal populations are characterized by size-structure, where individuals differ in body size and grow throughout their life cycle (Peters 1983; Calder 1984; Werner 1988). As ecologically important traits such as attack rate, metabolic costs and vulnerability to predation scale with body size (Peters 1983; Werner 1994; Persson et al. 1998; Kooijman 2000), growing individuals often experience different predation risks (Paine 1976; Chase 2003) or feed on different resources (Polis 1984; Olson 1996; Cisneros & Rosenheim 1997) over the course of their life (Werner & Gilliam 1984; *Correspondence author. arne.schroder@emg.umu.se Present address: Wageningen IMARES, PO Box 68, 1970 AB Ijmuiden, The Netherlands Ebenman & Persson 1988). Such ontogenetic niche shifts lead to that individuals occupy different trophic positions at different phases of their life cycle and engage in ontogenetically variable mixed interactions (Neill 1975, 1988; Polis 1984; Werner & Gilliam 1984, Wilbur 1988; Persson & Greenberg 1990; Olson, Mittelbach & Osenberg 1995; Bystro m, Persson & Wahlstro m 1998; Cameron et al. 2007). In case the interactions change from competition to predation (life-history omnivory; Pimm & Rice 1987), competitive recruitment bottlenecks can be imposed on juvenile predators by their competitively superior, later prey (Werner & Gilliam 1984; Neill 1988; Persson & Greenberg 1990; Olson et al. 1995; Byström et al. 1998; Walters & Kitchell 2001). On the other hand, large predators, when present, diminish prey densities and may thus release their own offspring from competition (Neill 1988; Wilbur 1988; Walters & Kitchell 2001). Ó 2009 The Authors. Journal compilation Ó 2009 British Ecological Society

2 Mixed interactions and size-dependent invasion success 1153 The dynamics of trophic configurations with mixed predation competition interactions has mainly been investigated within the theoretical framework of intraguild predation (IGP; Polis, Myers & Holt 1989; Holt & Polis 1997). This research has led to a number of important insights regarding the dynamics of communities with omnivory, including (i) the necessary condition of competitive superiority of the IGprey for its persistence, (ii) the exclusion of the IG-prey at high productivity when competition effects are mitigated and predation effects are relatively strong, (iii) the possibility of alternative stable states contrasting in number and identity of species, (iv) the overall smaller scope for coexistence of both species compared with tri-trophic food chains and (v) the pattern of community shifts along gradients of productivity (Holt & Polis 1997; Morin 1999; Diehl & Feißel 2000, 2001; Mylius et al. 2001). The implications of size-structure, recruitment bottlenecks and ontogenetic niche shifts for the dynamics of growing predator and prey have been discussed already in early literature on mixed predation competition systems (Werner & Gilliam 1984; Ebenman & Persson 1988; Polis et al. 1989). In addition, numerous short-term studies demonstrate sizedependent mixed interactions (Rice, Crowder & Rose 1993; Crumrine 2005; Griffin & Byers 2006; Rudolf & Armstrong 2008). Still, these features have only rarely been incorporated into IGP theory (Pimm & Rice 1987; Mylius et al. 2001; Van de Wolfshaar, de Roos & Persson 2006). While these studies have corroborated several of the findings of unstructured IGP theory, they have also led to new insights. Size-structure, particularly in co-occurrence with the ubiquitous feature of food-dependent individual development, demotes coexistence of IG-prey and IG-predator and expands the parameter region of mutual exclusion along a productivity gradient compared with the unstructured IGP systems (Mylius et al. 2001; Van de Wolfshaar et al. 2006). Notwithstanding this progress, the understanding of ontogenetically variable trophic interactions in terms of their impacts on community dynamics remains sketchy. One aspect that has been neglected so far is that the body size of invaders may matter for their invasion success. When trophic interactions and ecological roles change with body size over ontogeny, different sized invaders may face different biotic environments, either beneficial or detrimental to their ecological performance and thus for the invasion outcome. As the biotic environment is also influenced by the productivity of the system, invasion outcome may also depend on productivity. Here it was tested how invader body size and productivity affect the invasion outcome in a life-history omnivory system using two poeciliid fish species, Heterandria formosa [Agassiz 1855] or Least Killifish, and Poecilia reticulata [Peters 1859] or Common Guppy. H. formosa is the smaller species, so it was assumed that it would function as the IG-prey and P. reticulata as the IG-predator (Fig. 1). The experiment involved the invasion of small numbers of either large or small P. reticulata into resident populations of H. formosa at different productivities. First, it was predicted that large P. reticulata Hf large would establish persistent populations with the likelihood of invasion success increasing with productivity as they would benefit from the increasing density of the IG-prey. Second, it was predicted that small invaders would fail irrespective of productivity as they would always face competition by the resident IG-prey without being able to benefit from predation. Materials and methods STUDY SPECIES Hf small H. formosa and P. reticulata are small viviparous, sex-dimorphic poeciliid fishes native to the coastal plains in southeastern USA or the Lesser Antilles and the coastal regions of northern South America, respectively. They live in freshwater ponds and streams under a wide range of abiotic conditions (Rosen & Bailey 1963). H. formosa juveniles are born at a size of 5 8 mm. Size at maturation is 12 and 14 mm for males and females, respectively. Males reach a maximum size of 18 mm while females can attain a size of 36 mm (Fraser & Renton 1940). Females simultaneously carry embryo litters of different developmental stages and they give birth to litters of 1 7 young in intervals of 7 21 days (Cheong et al. 1984). The generation time is about 7 weeks (Travis & Henrich 1986). H. formosa founder individuals for the stock in Umeå were wild caught in June 2003 in Lake Jackson, Florida. P. reticulata juveniles are born at a size of 6 9 mm. Males mature at a size of 14 mm and females reach maturity at a size of 16 mm. Maximum size is 19 and 41 mm for males and females, respectively. Litters consist of young; the brood interval is 25 days. The generation time is c weeks (Reznick & Miles 1989). P. reticulata founder individuals were caught wild in 2002 in the Turure River, Trinidad. Individuals of both species can easily be distinguished from each other at any age and sex. H. formosa individuals are in general smaller and always show dark, vertical stripes on body and tail. Older individuals have partially orange-coloured back fins. P. reticulata individuals are of lighter colour; mature males have horizontal orange ornamental stripes. EXPERIMENTAL SET-UP Resource Pr small Pr large Fig. 1. Schematic presentation of realized (solid arrows) trophic interaction links in the Heterandria formosa Poecilia reticulata life-history omnivory system. The presence or absence of predation and cannibalism links are based on Fig. 6. Individual body growth is indicated by dashed arrows. The experiment was performed between May and December The set-up was a semi-closed circulation system of 32 aquaria each

3 1154 A. Schro der et al. with a volume of 80 L plus a reservoir with a volume of 600 L in which the water was aerated, warmed and mixed. Water was exchanged by a constant inflow of 20 L h )1 tab water into the reservoir. Water turnover in each aquarium was four times each day. The water was untreated well water with a hardness of 3Æ8 dh and a ph of 8Æ4. To prevent infections with ectoparasites salt was added daily to achieve a salinity of 0Æ5& and an ultraviolet (UV) water sterilizer (Wiegandt HW 4000; Wiegandt, Krefeld, Germany) was installed. Three bio-filters (Eheim 2080; Eheim, Deizisau, Germany) and one activated carbon filter were included to facilitate bacterial denitrification, detritus breakdown and waste product removal. Water temperature was held at 25 C. The aquaria were lit from above by 15 W neon lights with a 14 h light 10 h dark regime. No refuge was offered. The fish were fed pelletized food (Sera Microgran). Pellet feeding was accomplished by custom-made high-precision microfeeders (9Æ45 mg ± 0Æ41 SD per feeding event) centrally controlled from a computer. TREATMENTS The experimental design was full-factorial with productivity (low, medium, high) and invader body size (large, small) as factors. Replication was fourfold. Productivity was manipulated by varying the semi-chemostatic food supply rate. The daily amounts of food supplied were: 37Æ6 mgday )1 (low), 75Æ3 mg day )1 (medium) and 150Æ6 mg day )1 (high). Invader size treatment levels were either four P. reticulata adults (large; two females and two males, mean length ±1 SE 28Æ3 mm±1æ1 and 17Æ5 mm±1æ4, respectively) or five P. reticulata juveniles (small; mean length ±1 SE 6Æ8 mm ± 0Æ4). These low numbers were used to relate the study to: (i) real-world scenarios where invasive species usually enter new habitats in low densities and (ii) theory where the positive (or negative) growth of an invading population at low densities is used to infer about the invasion success. Two individuals of each sex were used to create balanced sex ratios and to account for potential demographical stochasticity in potential mortality. Five juveniles were used because the sex ratio at maturation is slightly biased towards males and to account for this demographical stochasticity. Only females which were not highly pregnant as judged by the form of their belly and the visibility of embryos through the transparent skin were chosen for invasions. Treatment combinations were semi-randomly assigned to aquaria as the feedercontrolling units made it necessary to group four aquaria of the same productivity regime together. Low-, medium- and high-productivity treatments were stocked with 56 (12 $$, 12##, 32 juveniles), 112 (26 $$, 26##, 60 juveniles) and 224 (52 $$, 52##, 120 juveniles) H. formosa individuals, respectively. All individuals were chosen and assigned randomly to a replicate. The initial size-structure and density of the resident populations chosen were based on the long-term average over 1½ years of eight control populations in a previous experiment in the same aquarium system, with food amount corresponding to the intermediate productivity level used here (Schro der, Persson & de Roos 2009). P. reticulata individuals were added the following day. Invading juveniles had been allowed to get acquainted to the presence of H. formosa for the prior 24 h in small breeding cages inside the stocked aquaria. A separate experiment was performed to investigate whether (i) P. reticulata could invade and persist in the system alone and (ii) assess the background mortality and maturation time of P. reticulata. The same numbers of P. reticulata individuals as before were added to four empty aquaria without H. formosa for each invader size class. These eight aquaria received 37Æ6 mg of food per day. Duration of the experiments was determined a priori. The end point was set by the maturation of the offspring of the small P. reticulata invaders. Thus, the experiments covered more than a complete life cycle of P. reticulata. One juvenile and one male P. reticulata died during the first six days of the experiment and were replaced with new, similar sized individuals the same day as such early mortality may reflect handling stress. If during the experiment only P. reticulata males were left of the initial invader individuals, half of them (or all when it was a single male) were replaced with similar sized females from stocking tanks to account for stochastic effects in maturation outcome. The experiment and the usage of vertebrates were approved of and followed the guidelines by the Swedish Animal Welfare Agency (DSM), applications Dnr A and Dnr A SIZE-DEPENDENT CAPTURE RATES Different sized juveniles (victims; range 6 12 mm) of each species were exposed to different sized females (consumer; range mm) of the other species to quantify size-dependent predatory capture rates. For each size combination trial, ten victims of the same size were placed in small plastic aquaria without refuge. Two similar sized consumers were placed in transparent plastic tubes inside the aquarium. Individuals had 24 h to get visually acquainted with the presence of consumers before the consumers were released. After another 24 h the remaining living victims were counted. Species-specific cannibalistic capture rates were obtained in similar ways. SAMPLING, MEASUREMENTS AND MAINTENANCE Sampling intensity was initially higher because it was expected that especially invasion failures would manifest already at the beginning of the experiment. The initial sampling intervals differed slightly between the large and the small invader size treatments. The entire population of an aquarium was captured by repeated hand netting until no individual could be seen during visual inspection of the aquarium. The individuals were then sorted into species and into females, males and juveniles. These subpopulations were placed in a transparent plastic bowl filled with just enough water for the individuals to stay upright and were then photographed. The individuals were counted and measured on a computer screen using an image analysis software (Optimas; Media Cybernetics, Silver Spring, MD, USA). The individuals were measured from the tip of the snout to where the cone-shaped, pigmented body merges into the straight and transparent tail fin (a close approximation to standard length). Relative lengths were transformed into metric units by using reference objects as internal standards. During the samplings, the aquaria were cleaned by netting out debris and the aquaria walls were scraped free of algae. The feeders were cleaned approximately every fourth week, checked for proper functioning and refilled. CALCULATIONS AND STATISTICAL ANALYSES A necessary condition for invasion success was the production of recruits by invaders, that is, small invaders had to mature and produce offspring, while the offspring of large invaders had to mature and produce recruits. Invasion success, failure and coexistence were evaluated at the termination of the experiment by noting the presence of individuals of the two species. The invasion was counted as a success when at least one female or juvenile of P. reticulata was present. If only one or more males were left the invasion was judged as a failure. Similarly, populations of H. formosa with at least one female or juvenile present were evaluated as persistent whereas populations with only one or more males left were viewed as extinct. Coexistence

4 Mixed interactions and size-dependent invasion success 1155 was defined as the presence of at least one female or juvenile of each species. In cases where males had been replaced by females, invasion outcome was also evaluated at the termination of the experiment. To test for differences in invasion success between treatments, a Generalized Linear Model was fitted to these binary success failure data with invader size and productivity as explanatory factors, specifying a binomial error distribution and logit link function. The significance of the factors and their interaction was examined in an analysis of deviance by using conditional F-statistics to correct for overdispersion (Crawley 2002). To quantify the invasion success of P. reticulata, a model of the form y=aæ exp () r max Æ time) with a as the initial density and r max as the maximum population growth rate at low density was fitted to the time series of the species total abundance. For the large invader treatment and the replicates in the absence of H. formosa, the model fitting was restricted to the first 100 days of the experiment. This allowed ignoring the precise form of the population growth over ranges of densities. As the densities of H. formosa also, in the majority of the small invader replicates, unexpectedly declined near-monotonically throughout the experiment the same procedure was used to estimate the decline rates of total H. formosa abundances and of abundances of juvenile and adult H. formosa. Similar subdivision of P. reticulata founder populations into juveniles and adults was not meaningful because populations were, at least initially, homogenous in their life-history stage and size structure. The absolute and log-transformed (to cure non-normality and heteroscedasticity to meet test assumptions) parameter estimates for r max were then used as the dependent variable in anova. A first analysis tested for differences in r max of total abundances in relation to species, invader size and productivity levels. A second analysis tested for differences in r max between H. formosa juveniles and adults, depending on the invader size and productivity. A third analysis tested for differences in r max of the total P. reticulata and total or stage-specific H. formosa abundance between productivities in the small invader treatments. Standard diagnostic plots were checked for non-normality, heteroscedasticity and the presence of highly influential points. The residuals were reasonably well behaved. In each analysis, two highly influential data points were removed; these outliers mainly concerned data points from aquaria where population dynamics were against the treatment trend. To yield information about the potential cannibalistic and predatory pressure the average size of vulnerable juveniles ( 12 mm) present was plotted against the maximum size of predatory adults ( 25 mm) present for each replicate and sampling during the first 100 days of the experiment when the population change was most pronounced. The size limits are based on the capture rates. This was done for each species separately to quantify potential cannibalism and for large P. reticulata vs. small H. formosa and vice versa to quantify potential interspecific predation. The maximum piscivore size was used to not lose information on extreme size combinations that may have disproportional large effects. Note that this makes the conclusion on the absence of predation cannibalism conservative. Predation cannibalism between size-structured consumer and victim is generally constrained by a lower victim : consumer size ratio below which the predator cannot detect the prey and an upper victim : consumer size ratio above which the consumer cannot handle the prey anymore (Rice, Crowder & Marschall 1997; Claessen, de Roos & Persson 2000). Because the juveniles of both species used here are born at a size far above the lower boundary, size combination data were compared with only the upper boundary of predation or cannibalism size windows. These boundaries were obtained from sizedependent capture rate data by regressing to the size of the largest victim successfully attacked by a consumer of a certain size over the consumer size. To obtain information about the intensity of potential competition that small invading P. reticulata individuals were experiencing, their body growth rate at different productivities and in the controls were compared using a mixed-effect ancova with the mean growth rate over all present individuals of a given replicate of the small invader treatments and sampling as the dependent variable. As explanatory variables, a single fixed treatment factor consisting of the three productivity levels and the controls was specified and time (the first 50 days of the experiment) was used as the covariate. The replicate was added as a random factor (Crawley 2002). Hypothesis testing was always two-sided and the a-level used was always 0Æ05. All computations and statistical analyses were performed in R2Æ7Æ0 (R Development Core Team 2008). Results In the absence of H. formosa, P. reticulata individuals invaded all empty aquaria and established persistent populations. This was independent of the invader size. The populations were dominated by juveniles (48Æ3 ± 4Æ9 juveniles vs. 14Æ3 ±1Æ8 adults, mean of eight replicates over the last four samplings ± 1 SE). Invading adults reproduced already during the first 18 days. The invading juveniles experienced no background mortality in these aquaria at least until their maturation (Fig. 2). r max values were higher for the small invaders than for the large ones (Fig. 5b; Wilcoxon s W =0, P =0Æ03). Large P. reticulata successfully invaded and established persisting populations in 10 out of 12 of their trials, while invading small P. reticulata mostly died and were successful in only 2 out of 12 of their trials (Figs 3 and 4). Productivity had no significant influence on invasion outcome (two-way analysis of deviance, invader size: F 1,18 =15Æ63, P <0Æ001; productivity: F 2,18 =1Æ55, P =0Æ22; invader size productivity: F 2,18 =0Æ07, P =1Æ0). In three out of the four cases when single males were replaced by females, these secondary invasions also failed. When invaders were successful, the resident populations usually went extinct over the course of the experiment. H. formosa populations also declined when the invaders were unsuccessful. Several of these populations showed, however ambiguous, signs of stabilization or even recovery towards the end of the experiment (Figs 3a,f,i and 4c,f,g,h). Only in three replicates, the two species coexisted at the end of the experiment. In one of these replicates, H. formosa densities were still declining while P. reticulata densities still increased (Fig. 3k), whereas in another replicate, only one P. reticulata female remained (Fig. 4a). Large invaders usually died before maturation of their offspring that later successfully reproduced themselves. Small invaders usually went extinct during the first days of the experiment before they matured. Total P. reticulata invader populations had positive initial population growth rates when consisting of large individuals but negative r max when consisting of small individuals. Total H. formosa population growth rates were always negative but total H. formosa populations declined faster in the large

5 1156 A. Schro der et al. Total abundance 1000 (a) (b) prey on P. reticulata juveniles or to cannibalize their own offspring in both invader size treatments (Fig. 6). Small invading P. reticulata grew more slowly in the presence of H. formosa than when invading the empty aquaria (mean body growth rates in mm day 1 ±CI 95% : empty aquaria = 0Æ28 ± 0Æ04; high productivity = 0Æ10 ± 0Æ06; medium productivity = 0Æ19 ± 0Æ06; low productivity = 0Æ18 ± 0Æ06; mixed-effect ancova, time: F 1,38 =395, P < 0Æ001; treatment: F 3,12 =8Æ8, P =0Æ002; time treatment: F 3,38 =11Æ7, P <0Æ001). Treatment contrasts showed that the treatment effect was because of differences in growth rate between the aquaria with and without the presence of H. formosa (P <0Æ05) and not because of differences between the productivity levels (P > 0Æ05). Discussion Day of experiment Fig. 2. The total abundance of Poecilia reticulata in the absence of Heterandria formosa over time for each replicate. (a) Large invader size; (b) small invader size. Note the log-scale of the ordinate. than in the small invader treatments. Productivity had no effect on total population growth rates (Fig. 5a, three-wayanova, species: F 1,34 =8Æ8, P <0Æ005; species invader size: F 1,34 =9Æ2, P =0Æ005; all other sources: F 1 or 2,34 3Æ1, P > 0Æ05). Juvenile as well as adult H. formosa subpopulations declined faster when large P. reticulata invaded than when invaded by small P. reticulata. Juvenile H. formosa declined consistently faster than adults. Productivity had no effect on these decline rates (Fig. 5a, three-way anova, stage: F 1,34 =155Æ9, P <0Æ001; invader size: F 1,34 =47Æ1, P < 0Æ001; all other sources: F 1 or 2,34 2Æ5, P >0Æ05). In case of the small invader treatment, P. reticulata declined faster than either the total H. formosa populations or H. formosa juveniles or adults (Fig. 5a, two-way anova, species stage: F 3,34 =26Æ9, P <0Æ001; productivity: F 2,34 =4Æ9, P = 0Æ013; species stage productivity: F 6,34 =2Æ3, P =0Æ06), apart from the low productivity treatment where this difference showed the same trend but was not statistically significant as revealed by treatment contrasts (P > 0Æ05). The potential for both predation and cannibalism during the first 100 days of the experiment was higher in the large than in the small invader treatments (Fig. 6). This was mainly because of the presence or absence of large P. reticulata individuals. H. formosa individuals were rarely large enough to THE COMMUNITY The trophic configuration in the experiment was close to the simplest case of a size-structured, ontogenetically variable mixed predation competition community (Fig. 1), similar to the life-history omnivory systems discussed by Werner & Gilliam (1984), Wilbur (1988) and Walters & Kitchell (2001) and mathematically analysed by Van de Wolfshaar et al. (2006). Large P. reticulata could prey on small H. formosa, while the reciprocal predation link with large H. formosa feeding on small P. reticulata was negligible as the corresponding size combinations were virtually absent. For similar reasons, the potential cannibalistic link within H. formosa was not realized whereas cannibalism in P. reticulata was present in the large invader treatment. Cannibalism may promote coexistence between IG-prey and IG-predator (Van de Wolfshaar 2006; Rudolf 2007) when the cannibalistic rate is stronger than the IGP rate (Rudolf 2007). However, here the reverse seems to hold, as is suggested by the (although statistically not significantly) lower intercept and slope of the upper predation and cannibalism boundaries (Fig. 6) and the higher capture rates by large P. reticulata of small H. formosa compared with small P. reticulata (data not shown). Thus, interaction links that could have increased the trophic complexity and influence the invasion outcome (cf. HilleRisLambers & Dieckmann 2003; Rudolf 2007) were at most weak. These findings led us to conclude that P. reticulata and H. formosa functioned as IG-predator and IG-prey, respectively. The higher density of P. reticulata alone at low productivity results from reporting abundances instead of biomass. P. reticulata populations were dominated by juveniles while initial H. formosa populations had a more balanced size-structure with a high fraction of adults. INVASION OUTCOME AND SIZE-DEPENDENT MECHANISMS Body size determined the invasion success of IG-predators as large invaders had positive initial population growth rates and successfully established persistent populations whereas

6 Mixed interactions and size-dependent invasion success 1157 (a) (e) (i) Total abundance (b) (f) (j) (c) (g) (k) (d) (h) (l) Day of experiment Fig. 3. The total abundance of Heterandria formosa (filled symbols) and Poecilia reticulata (open symbols) over time for each replicate of the large invader body size treatment at low (a d), medium (e h) and high productivity (i l). In (f), the arrow indicates the replacement of two of four P. reticulata males with two females. In (e) and (j), the single H. formosa individuals are males. Note the log-scale of the ordinate. small invaders had negative initial population growth rates and failed in their invasions. In correspondence with this pattern, the resident populations declined more rapidly when large invaders entered the aquaria than when small invaders came in (Fig. 5a). Furthermore, a strong potential predation pressure on resident IG-prey juveniles was observed when large IG-predators invaded. The capture rates and size combinations indicate that this predation was exclusively exerted by the large invader individuals and not by large residents (Fig. 6a). These results imply that IGP on the resident IGprey was actually taking place. In contrast, small invaders experienced a faster decline in abundance than the resident IG-prey (Fig. 5a). In addition, small invaders had positive initial population growth rates when invading empty aquaria (Fig. 5b). Together, this strongly suggests a detrimental and, more importantly, an asymmetric effect of resident IG-prey

7 1158 A. Schro der et al. (a) (e) (i) Total abundance (b) (f) (j) (c) (g) (k) (d) (h) (l) Day of experiment Fig. 4. The total abundance of Heterandria formosa (filled symbols) and Poecilia reticulata (open symbols) over time for each replicate of the small invader body size treatment at low (a d), intermediate (e h) and high productivity (i l). In (a), (f) and (k), each of the arrows indicate the replacement of one P. reticulata male with one female. In (a), the single P. reticulata individual is one of these females. In (k), the single P. reticulata is a male. Note the log-scale of the ordinate. individuals on small IG-predators. A density effect where small invaders went extinct first (with the residents following) simply because of their low initial numbers is not consistent with this observed asymmetry. The detrimental effect is also indicated by the lower body growth rates of small IG-predators in the presence of IG-prey. Predation by resident IG-prey on small invaders can be ruled out as the cause behind the asymmetry in population growth rates between species in the small invader treatments (Fig. 6b). In conclusion, we suggest that large invaders entered the residential populations in their ecological role as predators whereas small IG-predators invaded in their ecological role as inferior competitors. Small invaders suffered from a competitive recruitment bottleneck imposed by the IG-prey and

8 Mixed interactions and size-dependent invasion success 1159 Fig. 5. (a) The initial population growth rates (r max )ofpoecilia reticulata and Heterandria formosa for all treatment combinations. (b) r max of P. reticulata in the absence of H. formosa. Column heights represent the arithmetic mean of four replicates and error bars represent 1 SE. (a) r max Large low Large medium Large high Small low Small medium P.r. total H.f. total H.f. adults H.f. juveniles Small high (b) Large Small consequently were out-competed before maturation and before reaching sizes that allowed them to feed on the IGprey. In contrast, predation allowed the large invaders to successfully establish persistent populations by driving the IGprey extinct and thereby diminishing its negative effect on the survival and growth of the IG-predator s offspring. Overall, this scenario is consistent with the presence of a juvenile competitive recruitment bottleneck and an ontogenetic niche shift in the IG-predator (Neill 1988; Wilbur 1988; Walters & Kitchell 2001; Van de Wolfshaar et al. 2006). The mode of interspecific competition that limited the recruitment of small IG-predator is less clear. Individuals showed pronounced aggressive behaviour towards each other and the slower decline of small invaders in the low productivity replicates with their lower density of residents and thus lower frequency of interspecific encounters may speak for interference competition. On the other hand, the otherwise similar population growth rates and the similarity of body growth rates of small invaders across all productivity treatments indicate that food availability limited food intake, and hence points towards exploitation competition. PRODUCTIVITY, COEXISTENCE AND ALTERNATIVE COMMUNITIES Productivity had no influence on the outcome of size-dependent invasion and stable coexistence between IG-predator and IG-prey was virtually absent across the tested range of productivities. This observed pattern of exclusion is in agreement with the results of theoretical analyses of life-history omnivory systems (Mylius et al. 2001; Van de Wolfshaar et al. 2006). In comparison with unstructured IGP systems, coexistence is demoted and the parameter space for mutual exclusion of IG-predator and IG-prey along the productivity gradient is expanded in communities with life-history omnivory. One mechanism behind this pattern is that in lifehistory omnivory systems, individuals only hold one trophic position at a time and change their ecological role by fooddependent transitions between life-history stages (Werner & Gilliam 1984; Pimm & Rice 1987; Mylius et al. 2001; Van de Wolfshaar et al. 2006). Van de Wolfshaar et al. (2006) showed that this separation can establish a positive feedback between size-dependent IGP, relaxed interspecific competition and faster body growth of IG-predators, thereby demoting coexistence. This process of increasing predation pressure as a result of food-dependent body growth eventually leads to the extinction of the IG-prey. In our case, the mechanism for IG-prey extinction is likely to be different as the large size of the invading IG-predator adults immediately exposed IGprey to a high predation pressure. This led to IG-prey exclusion already within a generation and a direct, concomitant improvement of the food conditions of IG-predator offspring. Although we did not find any evidence for productivity dependence, at sufficiently low productivity levels invasion of large IG-predators is expected to fail. In contrast to the invasion of large IG-predators, invasion failure of small IG-predators irrespective of productivity was expected. This can be explained by that interspecific competition with the resident IG-prey always limits their recruitment. Thus, small IG-predators in their ecological role as inferior competitors will never benefit from the increase in IG-prey with productivity. For this, food-dependent survival and or body growth are essential processes because only these processes can establish the competitive recruitment bottleneck, unless mortality is caused by a food-independent process such as, for example, predation. This latter factor was, however, not likely in our study as is indicated by the absence of mutual IGP. Despite the wide productivity gradient used in our experiment with a fourfold difference between the lowest and the highest levels, productivity had no influence on what community established. Other studies have demonstrated major

9 1160 A. Schro der et al. Average prey size (mm) (a) (b) Whether the contrasting single-species states found here represent alternative stable communities predicted by life-history omnivory theory for intermediate productivity (Mylius et al. 2001; Van de Wolfshaar et al. 2006) is unclear. While there clearly is a priority effect of H. formosa over small P. reticulata and contingency on initial invader body size in what species establishes, the non-invasibility of the P. reticulataonly state has not been tested. However, we see it as unlikely that H. formosa could invade a resident P. reticulata population as predation by large residents would prevent any recruitment. Further, because of the density decline of the resident IG-prey populations also in most of the small invader replicates we cannot unambiguously argue for the stability of the H. formosa-only state. Other studies (Schro der et al. 2009; Schro der, Persson & Reichstein, unpublished data) in the same set-up demonstrated that H. formosa populations fluctuate in densities, rebounding after days. These conclusions are based on populations with refuge offered but a field study shows that H. formosa can also persist when vegetation cover is low (Richardson, Gunzburger & Travis 2006). We are therefore confident that H. formosa populations in the absence of P. reticulata, despite their decline, viably persist Max. predator size (mm) Pr -> Hf Hf -> Pr Pr -> Hf Hf -> Pr Fig. 6. The size combinations of the average prey and largest predator for large (a) and small (b) invader treatments. Symbols are for each replicate and sampling during the first 100 days of the experiment. Filled circles show the size combinations of average Heterandria formosa prey and largest Poecilia reticulata predators; open circles indicate the size combinations of average P. reticulata prey and largest H. formosa predators; the black lines depict the upper limit of the predation size windows for P. reticulata (solid lines) and H. formosa (dashed lines). Thus, circles below the lines indicate high predation vulnerability. Potential cannibalism pressure for each species was assessed in a similar way (data not shown; intercept and slope for upper P. reticulata cannibalism limit is )0Æ21 and + 0Æ28, respectively, compared with + 0Æ31 and + 0Æ32 for P. reticulata predation on H. formosa). shifts in community structure along similar wide gradients (Morin 1999; Diehl & Feißel 2000, 2001). This difference in results can be related to that the latter studies used protozoan communities where population size-structure is much less pronounced. In contrast, a similar independence of invasion success of IG-predators on productivity along an even wider productivity gradient has recently been observed in a sizestructured mite community (Montserrat et al. 2008). The lack of coexistence of IG-predator and IG-prey in a life-history omnivory community has also recently been demonstrated in a whole-lake invasion experiment (Persson, de Roos & Bystro m 2007). GENERAL IMPLICATIONS The majority of animal taxa are size-structured (Peters 1983; Calder 1984; Werner 1988) and ontogenetic niche shifts or recruitment bottlenecks are frequently observed (e.g. Woodward & Hildrew 2002). Size-dependent IGP and life-history omnivory may therefore be a common phenomenon (e.g. Dick, Montgomery & Elwood 1999; Turner, Turner & Ray 2007) and hence also the dependence of invasion outcome on invader body size. Size-dependent invasion success has not been considered in earlier theoretical or experimental studies but may explain unexpected results of previous IGP experiments. For example, Montserrat et al. (2008) demonstrated consistent exclusion of IG-prey after IG-predator invasion along a productivity gradient. They did not report the size-structure of the invader populations but our results suggest that these might have been dominated by adults. Size-dependent invasion success can also explain the nonrecovery of overexploited predatory fish stocks (Hutchings 2000) when after the collapse increased forage fish populations reduce the shared resource and limit predator recruitment (Walters & Kitchell 2001). Such a scenario and the associated ecosystem-wide regime shifts have also recently been proposed for the Baltic Sea cod population (Casini et al. 2009). Overall, our results show that ontogenetically variable mixed interactions can demote coexistence and imply that the size-structure of an invading population can affect its invasion success. Further empirical and theoretical research is needed to investigate the necessary and sufficient conditions and the generality of this phenomenon. In particular, the interaction between the size distribution of the invader population and system productivity deserves further study.

10 Mixed interactions and size-dependent invasion success 1161 For example, how more intermediate-sized invaders than used here which would need to grow in body size to undergo the ontogenetic niche shift but might be less vulnerable to competition would fare at different productivities remains to be investigated. Acknowledgements The authors thank Blair Daniel for help with the experiment. Emma Göthe measured the fish and helped maintaining the set-up. Joseph Travis and David Reznick sent the fish from Florida and California, respectively. Lars Lundmark, William Larsson and Lars-Ola Westlund provided technical support for the feeders. André M. de Roos, Zlatko Petrin and three anonymous reviewers gave valuable comments on earlier drafts of the manuscript. The research was supported by grants from the Memorial of J.C. Kempe Foundation to A. Schro der and L. Persson and the Swedish Research Council to L. Persson. References Byström, P., Persson, L. & Wahlström, E. (1998) Competing predators and prey: juvenile bottlenecks in whole-lake experiments. Ecology, 79, Calder, W.A. III (1984) Size, Function and Life-History. Harvard University Press, Cambridge, MA. Cameron, T.C., Wearing, H.J., Rohani, P. & Sait, S.T. (2007) Two-species asymmetric competition: effects of age structure on intra- and interspecific interactions. Journal of Animal Ecology, 76, Casini, M., Hjelm, J., Molinero, J.-C., Lo vgren, J., Cardinale, M., Bartolino, V., Belgrano, A. & Kornilovs, G. (2009) Trophic cascades promote threshold-like shifts in pelagic marine ecosystems. Proceedings of the National Academy of Sciences of the USA, 106, Chase, J.M. (2003) Experimental evidence for alternative stable equilibria in a benthic pond food web. Ecology Letters, 6, Cheong, R.T., Henrich, S., Farr, J.A. & Travis, J. (1984) Variation in fecundity and its relationship to body size in a population of the Least Killifish, Heterandria formosa (Pisces: Poeciliidae). Copeia, 3, Cisneros, J.J. & Rosenheim, J.A. (1997) Ontogenetic change of prey preference in the generalist predator Zelus renardii and its influence on predator predator interactions. Ecological Entomology, 22, Claessen, D., de Roos, A.M. & Persson, L. (2000) Dwarfs and giants: cannibalism and competition in size-structured populations. The American Naturalist, 155, Crawley, M.J. (2002) Statistical Computing. An Introduction to Data Analysis using S-Plus. John Wiley & Sons, Chichester, UK. Crumrine, P.W. (2005) Size structure and substitutability in an odonate intraguild predation system. Oecologia, 145, Dick, J.T.A., Montgomery, W.I. & Elwood, R.W. (1999) Intraguild predation may explain an amphipod replacement: evidence from laboratory populations. Journal of Zoology, 249, Diehl, S. & Feißel, M. (2000) Effects of enrichment on three-level food chains with omnivory. The American Naturalist, 155, Diehl, S. & Feißel, M. (2001) Intraguild prey suffer from enrichment of their resources: a microcosm experiment with ciliates. Ecology, 82, Ebenman, B. & Persson, L. (eds) (1988) Size-Structured Populations: Ecology and Evolution. Springer-Verlag, Berlin, Germany. Fraser, E.A. & Renton, R.M. (1940) Observations on the breeding and development of the viviparous fish, Heterandria formosa. Quarterly Journal of Microscopical Science, 81, Griffin, B.D. & Byers, J.E. (2006) Intraguild predation reduces redundancy of predator species in multiple predator assemblages. Journal of Animal Ecology, 75, HilleRisLambers, R. & Dieckmann, U. (2003) Competition and predation in simple food webs: intermediately strong trade-offs maximize coexistence. Proceedings of the Royal Society of London Series B, 270, Holt, R.D. & Polis, G.A. (1997) A theoretical framework for intraguild predation. The American Naturalist, 149, Hutchings, J.A. (2000) Collapse and recovery of marine fishes. Nature, 406, Kooijman, S.A.L.M. (2000) Dynamic Energy and Mass Budgets in Biological Systems. Cambridge University Press, Cambridge, UK. Montserrat, M., Magalhães, S., Sabelis, M.W., de Roos, A.M. & Janssen, A. (2008) Patterns of exclusion in an intraguild predator prey system depend on initial conditions. Journal of Animal Ecology, 77, Morin, P. (1999) Productivity, intraguild predation, and population dynamics in experimental food webs. Ecology, 80, Mylius, S.D., Klumpers, K., de Roos, A.M. & Persson, L. (2001) Impact of intraguild predation and stage structure on simple communities along a productivity gradient. The American Naturalist, 158, Neill, W.E. (1975) Experimental studies of microcrustacean competition, community composition and efficiency of resource utilization. Ecology, 56, Neill, W.E. (1988) Community response to experimental nutrient perturbations in oligotrophic lakes: the importance of bottlenecks in size-structured populations. Size-Structured Populations: Ecology and Evolution (eds B. Ebenman & L. Persson), pp Springer-Verlag, Berlin, Germany. Olson, M.H. (1996) Ontogenetic niche shifts in Largemouth Bass: variability and consequences for first-year growth. Ecology, 77, Olson, M.H., Mittelbach, G.G. & Osenberg, C.W. (1995) Competition between predator and prey resource-based mechanisms and implications for stage-structured dynamics. Ecology, 76, Paine, R.T. (1976) Size-limited predation: an observational and experimental approach with the Mytilus Pisaster interaction. Ecology, 57, Persson, L. & Greenberg, L.A. (1990) Juvenile competitive bottlenecks, the perch (Perca fluviatilis) roach (Rutilus Rutilus) interaction. Ecology, 71, Persson, L., Leonardsson, K., de Roos, A.M., Gyllenberg, M. & Christensen, B. (1998) Ontogenetic scaling of foraging rates and the dynamics of a sizestructured consumer-resource model. Theoretical Population Biology, 54, Persson, L., de Roos, A.M. & Byström, P. (2007) State-dependent invasion window for prey in size-structured predator prey systems: whole lake experiments. Journal of Animal Ecology, 76, Peters, R.H. (1983) The Ecological Implications of Body Size. Cambridge University Press, Cambridge, MA. Pimm, S.L. & Rice, J.C. (1987) The dynamics of multi-species, multi-life stage models of aquatic food webs. Theoretical Population Biology, 32, Polis, G.A. (1984) Age structure component of niche width and intraspecific resource partitioning: can age groups function as ecological species? The American Naturalist, 123, Polis, G.A., Myers, C.M. & Holt, R.D. (1989) The ecology and evolution of intraguild predation: potential competitors that eat each other. Annual Reviews of Ecology and Systematics, 20, R Development Core Team (2008) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN ; URL: Reznick, D. & Miles, D.B. (1989) Review of life history patterns in poeciliid fishes. Ecology and Evolution of Live-Bearing Fishes (Poeciliidae) (eds G.K. Meffe & F.F. Snelson), pp Prentice Hall, Englewood Cliffs. Rice, J.A., Crowder, L.B. & Rose, K.A. (1993) Interactions between size-structured predator and prey populations: experimental test and model comparison. Transactions of the American Fisheries Society, 122, Rice, J.A., Crowder, L.B. & Marschall, E.A. (1997) Predation on juvenile fishes: dynamic interactions between size-structured predators and prey. Early Life History and Recruitment in Fish Populations (eds R.C. Chambers & E.A. Trippel), pp Chapman & Hall, London, UK. Richardson, J.M.L., Gunzburger, M.S. & Travis, J. (2006) Variation in predation pressure as a mechanism underlying differences in numerical abundance between populations of the poeciliid fish Heterandria formosa. Oecologia, 147, Rosen, D.E. & Bailey, R.M. (1963) The poeciliid fishes (Cyprinodontiformes): their structure, zoogeography, and systematics. Bulletin of the American Museum of Natural History, 126, Rudolf, V.H.W. (2007) The interaction of cannibalism and omnivory: consequences for community dynamics. Ecology, 88, Rudolf, V.H.W. & Armstrong, J. (2008) Emergent properties of cannibalism and size-refuge in prey on intraguild predation systems. Oecologia, 157, Schro der, A., Persson, L. & de Roos, A.M. (2009) Culling experiments demonstrate size-class specific biomass increases with mortality. Proceedings of the National Academy of Sciences of the USA, 106, Travis, J. & Henrich, S. (1986) Some problems in estimating the intensity of selection through fertility differences in natural and experimental populations. Evolution, 40,