The effects of the diel cycle and the density of an invasive predator on predation risk and prey response

Animal Behaviour 117 (2016) 87e95

Contents lists available at ScienceDirect

Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav

The effects of the diel cycle and the density of an invasive predator on predation risk and prey response L. K. Lopez*, M. Y. L. Wong, A. R. Davis School of Biological Sciences, University of Wollongong, NSW, Australia

a r t i c l e i n f o Article history: Received 12 November 2015 Initial acceptance 25 January 2016 Final acceptance 25 April 2016 MS number 15-00962 Keywords: antipredator density diel cycle invasive predation risk

All prey face a fundamental trade-off between avoiding predation and pursuing activities, such as foraging and mating, that enhance fitness. Therefore, the effects of predation can be both consumptive and nonconsumptive and prey need to assess and respond appropriately to predation risk which in turn varies with environmental and social contexts. We tested the effects of predator density and diel cycle on the consumption, interspecific interactions and behavioural responses of a prey species, the native Australian glass shrimp, Paratya australiensis, exposed to a predator, the invasive eastern mosquito fish, Gambusia holbrooki. In the laboratory, P. australiensis were exposed to low or high densities of conspecifics or predators and observed during the day and at night. While P. australiensis experienced more interspecific approaches and nips when exposed to a high density of G. holbrooki and during the day, neither predator density nor diel cycle influenced the actual number of P. australiensis consumed. Similarly, while P. australiensis engaged in significantly more shelter use and swam less, there was no difference in these behaviours in relation to predator density and diel cycle. Foraging by P. australiensis was not related to species composition, but instead depended on the overall number of animals present with more P. australiensis foraging when exposed to a high density of conspecifics and G. holbrooki. These results indicate that the mechanisms by which G. holbrooki exerts negative effects on P. australiensis can be multiple and wide ranging, from direct predation to a reduction in activity and competition for resources. However, as neither predator density nor diurnal variation altered predation rate, P. australiensis did behave in an adaptive manner, by only adjusting its behavioural responses in proportion to the direct risk of predation. © 2016 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

To reduce the risk of being consumed by predators, prey often exhibit substantial behavioural changes including increased vigilance, refuge use, dispersal and changes in activity patterns (Eccard, € nen, 2008; Lima, 1998; Lima, Valone, Pusenius, Sundell, Halle, & Ylo & Caraco, 1985; Sih, 1986; Sih et al., 2010). While frequently successful in reducing the rate of direct consumption (defined as consumptive effects, CEs), such behavioural changes may also negatively impact upon foraging and reproduction (defined as nonconsumptive effects, NCEs), thus leading to a trade-off between minimizing mortality from predation and maximizing fitness (Lima et al., 1985; Paterson et al., 2013; Sih, 1986). Therefore, as posited by Helfman's (1989) threat-sensitive predator avoidance hypothesis, prey need to maximize their overall fitness by exhibiting antipredator responses that are proportional to the level of predation

* Correspondence: L.K. Lopez, School of Biological Sciences, University of Wollongong, North Fields Avenue, Wollongong, NSW 2522, Australia. E-mail address: [email protected] (L. K. Lopez).

risk to which they are exposed. If the avoidance behaviours exhibited by prey are in excess of the predation risk, then the strength of NCEs on prey fitness will increase, yet if antipredator behaviours are insufficient, the strength of CEs on prey should become greater (Anholt, Werner, & Skelly, 2000; Ferrari, Sih & Chivers, 2009; Helfman, 1989; Lima et al., 1985; Sih, 1986). One of the key factors influencing levels of perceived predation risk is predator density (Ferrari et al., 2009; Foam, Mirza, Chivers, & Brown, 2005; Vucetich, Peterson, & Schaefer, 2002). Density itself can be modulated if predators form aggregations in the environment, in turn generating a predation risk that is spatially heterogeneous (Butler, 1989; Mella, Banks, & MacArthur, 2014; Navarrete & Menge, 1996). Currently, there is mixed support as to whether predator density and CEs are positively correlated (Sih, 1986; Vance-Chalcraft, Soluk, & Ozburn, 2004). Typically, higher predator densities have been related to an increased number of encounters between predators and prey, leading to a higher per capita kill rate, as observed in wolves, Canis lupus, and moose, Alces alces (Stier, Geange, & Bolker, 2013; Vucetich et al., 2002). Even so,

http://dx.doi.org/10.1016/j.anbehav.2016.05.007 0003-3472/© 2016 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved.

88

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

negative correlations between predator density and prey capture and kill rates have been reported, due to an increase in the frequency of competitive interactions among foraging predators, known as mutual interference (Abrams, 1993; Mistri, 2003; Sih, 1979). With regard to NCEs, the relationship with predator density is also often complex. Flathead minnows, Pimephales promelas, show stronger antipredator responses in the form of shoaling and flee behaviours when exposed to the odour cues of 12 pike compared to just two (Ferrari, Messier, & Chivers, 2006). Similarly, exposure to a high density of predatory mites led to greater dispersal by spider mites, Tetranychus kanzawai (Bowler, Yano, & Amano, 2013). However, a recent review by Paterson et al. (2013) reported that crustacean prey from a number of taxa exhibited similar changes in activity levels and refuge use regardless of cue intensity or exposure time to fish predators. At this point it is unknown whether it may be advantageous for prey to display similar avoidance behaviours which may be an effective response to multiple predator densities, or whether the intensity or combination of cue types used overpowered subtle indications of predator density (Paterson et al., 2013). Levels of predation risk can also vary with diurnal changes in predator foraging behaviour (Griffin, Griffin, Waroquiers, & Mills, 2005; Helfman, 1989), effectively providing prey with an indication of predation risk (Clark, Ruiz & Hines, 2003). Typically, prey are more active in periods of ambient light levels in which predators are less active (Benfield & Minello, 1996). For example, buffalo, Syncerus caffer, warthog, Phacochoerus africanus, and kudu, Tragelaphus spp., are predominantly active during the day in areas where nocturnal predators coexist; however, in their absence these prey species are also active at night (Tambling et al., 2015). There is also some evidence to suggest that the diel cycle can interact with other environmental variables such as water depth to alter predation risk and subsequently prey refuge use (Bollens & Frost, 1989; Bollens & Stearns, 1992; Clark et al., 2003). This has been observed in grass shrimp, Palaemonetes spp., in which mortality from predation by fish was found to be depth dependent during the day but not at night (Clark et al., 2003). Less well understood, however, is whether the diel cycle interacts with predator density to influence the magnitude of diurnal shifts in shelter use, swimming, foraging and other behaviours that have been observed in prey. In this study, we investigated the effects of predator density and diel cycle on the consumptive and nonconsumptive effects of the invasive predator, the eastern mosquito fish, Gambusia holbrooki, on the native Australian glass shrimp, Paratya australiensis. This shrimp is widespread in coastal eastern Australia (Cook et al., 2006) and plays a key role in nutrient cycling in freshwater and estuarine ecosystems as well as being a food source for native species (March, Pringle, Townsend, & Wilson, 2002; Richardson, Growns, & Cook, 2004; Walsh & Mitchell, 1995). The ubiquitous G. holbrooki has spread to eight of the 11 main drainage basins on the Australian continent since its introduction to Sydney in 1925 (Pyke, 2008). It is an opportunistic omnivore which feeds on small decapod crustaceans, including P. australiensis (Arthington & Marshall, 1999; Bool, Witcomb, Kydd, & Brown, 2011). Gambusia spp. are able to consume relatively large prey by nipping the body, tail and gills which results in immobilization and death (Komak & Crossland, 2000; Segev, Mangel, & Blaustein, 2009; Shulse & Semlitsch, 2014). For this reason, Gambusia spp. are not considered to be limited in their prey selectivity by their gape size (Baber & Babbitt, 2003; Drake, Anderson, Smith, Lohraff, & Semlitsch, 2014; Smith & Smith, 2015). The density of G. holbrooki is known to vary seasonally, with peak abundance in early autumn after the breeding season and the lowest abundance in spring (Barney & Anson, 1921;

Morton, Beumer, & Pollock, 1988; Pyke, 2008; Zulian, Bisazza, & Marin, 1993). In addition, as it is a visual predator, the foraging behaviour of G. holbrooki is likely to be greater during the day than at night (Bool et al., 2011). Although it has been associated with a decline in populations of native fairy shrimp, Linderiella occidentalis, in California (Leyse, Lawler, & Strange, 2004), surprisingly few studies have quantified the behavioural interactions of G. holbrooki with native biota with the specific purpose of identifying the exact mechanisms behind its negative impacts on native species. Furthermore, native prey may be especially vulnerable as they do not share an evolutionary history with the predator, possibly rendering them less adept at detecting risk and responding accordingly (Bourdeau, Pangle, Reed, & Peacor, 2013; Heavener, Carthey, & Banks, 2014). Specifically, we determined whether predator density and diel cycle affected the number of P. australiensis consumed by G. holbrooki. Direct behavioural interactions, in the form of approaches and nips by G. holbrooki to P. australiensis, were also recorded. Additionally, we assessed whether the behavioural responses of P. australiensis, namely shelter use, swimming and foraging behaviours, covaried with predator density and diel cycle. We hypothesized that (1) the number of predation events and direct interactions between G. holbrooki and P. australiensis would be greatest at the high predator density and during the day and (2) that in response to the greater predation risk, P. australiensis would exhibit greater behavioural changes in the presence of a high density of G. holbrooki and during the day. METHODS Animal Collection and Aquaria Set-up Gambusia holbrooki (mean ± SE mass ¼ 0.19 ± 0.14 g; mean ± SE standard length ¼ 19.46 ± 0.46 mm; mean ± SE total length ¼ 24.21 ± 0.53 mm) were collected from freshwater ponds located on the University of Wollongong campus (34 2401900 S, 150 520 4200 E) using a baited hand-held landing net. Only adult females were collected so as to avoid the mating behaviours displayed by males which may have interfered with the predatory behaviours exhibited by females relevant to this study. Paratya australiensis (mean ± SE mass ¼ 0.07 ± 0.09 g; mean ± SE carapace length ¼ 5.77 ± 0.1 mm) were acquired from the national supplier LiveFish.com. Berried P. australiensis were included in the study and their condition was noted. To conduct the experiment, six recirculating aquarium systems were used at the University of Wollongong, each system containing eight aquaria (37  22 cm and 27 cm high) that were interconnected and subjected to water conditions of 23  C and 5 ppt salinity. Each aquarium was lined with 2 cm of natural river gravel and contained three black plastic tubes (7  2 cm) positioned on the substratum to provide shelter for P. australiensis. The exterior sides of each tank were covered with black plastic to exclude visual cues from individuals in adjacent tanks. To acclimatize G. holbrooki and P. australiensis to laboratory conditions, G. holbrooki (N ¼ 72 total fish) were placed into 12 aquaria spread equally across the six systems (N ¼ 12 fish per system). Paratya australiensis (N ¼ 288 total individuals) were placed into 36 aquaria spread equally across the six systems (N ¼ 48 individuals per system) and separate from those housing G. holbrooki. All G. holbrooki and P. australiensis were maintained under these conditions for 7 days to ensure adequate acclimation to laboratory conditions. During this time, G. holbrooki were fed a commercial fish flake (New Life Spectrum Thera formula) and P. australiensis were fed a shrimp granule (Fluval). Water changes (<20%) were made once a fortnight and new water was supplemented with Fluval bacterial and shrimp mineral additives. Owing

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

to the sensitivities of P. australiensis to traces of copper and other metals in the available tap water, deionized water was also treated with CupriSorb (Seachem) to remove any traces of copper.

Experimental Design To quantify the consumptive effects (CEs) and nonconsumptive effects (NCEs) of G. holbrooki on P. australiensis, we established two conditions: (1) treatment condition, in which aquaria contained both G. holbrooki and P. australiensis (N ¼ 24 aquaria) and (2) control condition, in which aquaria contained only P. australiensis (N ¼ 24 aquaria). Within the treatment condition aquaria (both species present), two density levels of G. holbrooki were established: (1) low density (N ¼ 1 individual/aquarium; N ¼ 12 aquaria) and (2) high density (N ¼ 5 individuals/aquarium; N ¼ 12 aquaria), both with a constant density of P. australiensis (N ¼ 4 individuals/ aquarium). Within the control condition (P. australiensis only), the total number of P. australiensis was equal to the total number of P. australiensis and G. holbrooki in the equivalent low and high density treatment condition (N ¼ 5 individuals in the low density; N ¼ 12 aquaria, and N ¼ 9 individuals in the high density; N ¼ 12 aquaria). Treatment densities were based on estimates of seasonal low and high G. holbrooki densities in University of Wollongong campus ponds relative to the aquaria size used in this study (L. K. Lopez, personal observation). An alternative ratio design (see Inouye, 2001) was considered inappropriate as by changing the ratio of P. australiensis to G. holbrooki it would be impossible to determine whether any differences in the response variables were due to changes in the overall density of animals or species composition.

Behavioural Observations and Data Collection Following the acclimatization period, each P. australiensis was measured using hand-held callipers (carapace length (CL) ± 0.1 mm) and weighed with an electronic balance (±0.1 g). To enable the identification of individuals and control for any effect of body size, four P. australiensis (‘focal’ individuals) from each aquarium were tagged using a coloured fluorescent polymer elastomer (red, pink, orange or green; Northwest Technologies Inc., Estacada, OR, U.S.A.) injected into the musculature of the telson. Four P. australiensis were introduced into a test aquarium for 24 h prior to the addition of either one G. holbrooki (treatment) or P. australiensis (control) in the low density condition, or five G. holbrooki (treatment) or P. australiensis (control) in the high density condition. This order of residency was intended to simulate the natural temporal occurrence of species in the event of an invasion by G. holbrooki. Observations commenced after a 60 min acclimation period followed by a 5 min adjustment to the presence of the observer. Only one person (L.K.L.) conducted observations to avoid potential issues with observer bias. It was not possible to observe the animals blind as the species composition and number differed between treatments and controls. Each tagged P. australiensis was observed twice, once during the day (0900e1300 hours under natural daylight and under red light) and once at night (1700e2100 hours under red light). Each observation period lasted for 10 min during which the time spent in shelter (s), time spent engaged in specific activities (swimming and foraging, s) and direct behavioural interactions with G. holbrooki (approaches and nips) were recorded. The two species were therefore exposed to each other for a maximum of 12 h after which they were separated and placed back into separate aquaria containing only conspecifics.

89

Data Analysis Paratya australiensis was predated upon by G. holbrooki in the high density (N ¼ 9 individuals) and low density (N ¼ 6 individuals) treatments, and there was nonpredation mortality in the low density control condition (N ¼ 1 individual). Therefore, behavioural data collected during the experiment from these predated subjects were excluded from subsequent analyses in order to use an approach incorporating time as a repeated measures factor. The effects of predator density (low or high) and time of day (day or night) on the number of predation events by G. holbrooki on P. australiensis was analysed using a two-way analysis of variance (ANOVA) with density (fixed), time (fixed) and tank ID (random) factors in the model (using SPSS Statistics 21, IBM, Armonk, NY, U.S.A.). To assess whether the size difference between P. australiensis and G. holbrooki influenced whether a predation attempt (nipping by G. holbrooki) resulted in consumption of P. australiensis (successful outcome) or not (unsuccessful outcome), prey and predator size were scaled. This scaling was achieved by dividing the average carapace length of P. australiensis in each low or high density treatment replicate tank by the average standard length of G. holbrooki in the same tank (following Bence & Murdoch, 1986). Standard length of fish was used as it is linearly related to gape size in G. holbrooki (Bence & Murdoch, 1986). Scaled prey size was analysed using an ANOVA with density (fixed), outcome (fixed) and tank ID (random) as factors. To assess the effects of predator density on the number of interspecific interactions (the sum of approaches and nips), a generalized linear mixed model (GLMM) with backward stepwise elimination was used, incorporating a negative binomial distribution and a log link function which is appropriate for analysing zeroinflated count data. The model consisted of density (fixed), time of day (repeated measure), carapace length (covariate) and tank ID (random) as factors (using SPSS Statistics 21). To assess the NCEs of G. holbrooki density and whether there was an effect of diel cycle, the amount of time in which focal P. australiensis engaged in shelter use, swimming and foraging (s) was first converted into a percentage of total activity for each day and night observation period. For shelter use, data were categorized into whether individuals spent 50% or less (0) or more than 50% (1) of the observed time in shelters. Owing to the small amount of time P. australiensis were observed swimming and foraging, the most suitable model was also found to use a binomial distribution, categorizing the response variable depending on whether swimming and foraging were observed for an individual shrimp (1 ¼ yes, 0 ¼ no). To investigate the effects of predator density and diel cycle on shelter use, swimming and foraging in P. australiensis, a GLMM with backward stepwise elimination, incorporating a binomial distribution and logit function was used, with density (fixed), condition (fixed), time of day (repeated measure), carapace length (covariate) and tank ID (random factor) as factors in the model. Ethical Note As advised by Brennan, Leber and Blackburn (2007), P. australiensis were not anaesthetized during the tagging procedure so as to avoid the risk of mortality as well as to reduce handling and recovery times. As a result, we did not observe mortality due to tagging. Upon completion of the experiment, G. holbrooki were euthanized using clove oil as it is illegal to release an invasive species under NSW legislation. Paratya australiensis were maintained in aquaria until their natural deaths. To minimize stress on both G. holbrooki and P. australiensis, G. holbrooki were not starved prior to exposure to P. australiensis. Shelters were provided

90

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

to allow P. australiensis to hide from G. holbrooki when needed and tank sides were covered to prevent animals from being affected by the possible predation activities occurring in adjacent tanks. We limited behavioural trials to a maximum of 12 h and separated P. australiensis and G. holbrooki immediately after this period. The methods used for animal capture, housing and tagging were approved by the Animal Ethics Committee of the University of Wollongong (Animal Ethics Protocol No. 13/09) and adhered to the Scientific Collection guidelines (permit No. P13/0011-1.3) of the NSW Department of Primary Industries.

Table 1 Scaled P. australiensis size (P. australiensis carapace length/G. holbrooki standard length, mm) in successful and unsuccessful predation attempts in low and high predator density treatments Predation attempt outcome

Low density treatment High density treatment

Successful

Unsuccessful

0.21±0.04 0.29±0.05

0.25±0.01 0.25±0.03

N ¼ 12.

RESULTS Consumptive Effects In this study G. holbrooki were observed to predate upon P. australiensis by nipping the telson and body until the prey was immobilized and died, upon which consumption would occur. Gambusia holbrooki were not observed to consume the eggs of berried P. australiensis. In total, predation events of P. australiensis by G. holbrooki were observed in the low (N ¼ 6 events) and high (N ¼ 9 events) density treatments during the course of this experiment (Fig. 1). There were no significant effects of predator density (ANOVA: F1,44 ¼ 0.37, P ¼ 0.54), time of day (F1,44 ¼ 3.31, P ¼ 0.076) or an interaction between predator density and time of day (F1,44 ¼ 0.04, P ¼ 0.84) on the number of P. australiensis predated upon. There was no significance difference in scaled P. australiensis size between successful and unsuccessful predation attempts at either low or high density treatments (ANOVA: density: F1,14 ¼ 1.08, P ¼ 0.31; outcome: F1,14 ¼ 0.06, P ¼ 0.82), and no significant interaction between density and outcome (F1,14 ¼ 0.78, P ¼ 0.39; Table 1). Nonconsumptive Effects Direct interspecific interactions Paratya australiensis exposed to a high predator density received significantly more nips and approaches than P. australiensis in the low density treatment (GLMM Density, F1,156 ¼ 23.84, P ¼ <0.0001) (Fig. 2). In addition, P. australiensis received significantly more nips and approaches during the day than at night regardless of treatment (F1,156 ¼ 22.22, P < 0.0001; Fig. 2). There was no significant interaction between density and time of day (F1,155 ¼ 0.31, P ¼ 0.58). Overall, smaller P. australiensis received significantly

more nips and approaches (F1,156 ¼ 5.38, P ¼ 0.02) although there was a significant interaction between density and shrimp size (F1,156 ¼ 6.80, P ¼ 0.01). Therefore, smaller P. australiensis received significantly more nips and approaches in the high density than low density treatments (Fig. 3a). Shelter usage Paratya australiensis spent significantly more time in shelters during the day than at night in both treatment and control conditions (GLMM: F1,347 ¼ 43.90, P < 0.0001; Fig. 4a). Shelter use for P. australiensis was significantly greater in the treatments (both low and high density) than in the controls (both low and high density) during the day and at night (F1,347 ¼ 17.61, P < 0.0001). However, there was no effect of density on shelter usage (F1,346 ¼ 2.73, P ¼ 0.10) nor an interaction between condition and density (F1,341 ¼ 0.01, P ¼ 0.91; Fig. 4a). Although animal density alone did not have an effect on the time spent in shelters, there was a significant interaction between body size and density, with larger P. australiensis spending more time in shelters under high than low density control and treatment conditions (F2,347 ¼ 3.56, P ¼ 0.03; Fig. 3b). Swimming Paratya australiensis exposed to both low and high densities of G. holbrooki spent significantly less time swimming than those where G. holbrooki was absent (GLMM: F1,350 ¼ 10.12, P < 0.01; Fig. 4b). There was no significant effect of time, density or size on swimming behaviour (time: F1,345 ¼ 1.24, P ¼ 0.27; density: F1,343 ¼ 0.08, P ¼ 0.77; carapace length: F1,343 ¼ 0.73, P ¼ 0.39). Foraging Paratya australiensis foraged for significantly longer periods of time at night than during the day in both treatment and control

Average number of predation events

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Low density treatment

High density treatment

Figure 1. Mean ± SE number of predation events by G. holbrooki on P. australiensis in low and high density treatments during daytime (open bars) and night-time (closed bars) observation periods. N ¼ 12.

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

91

Average number of interspecifc interactions

4 3.5 3 2.5 2 1.5 1 0.5 0

Low density treatment

High density treatment

Figure 2. Mean ± SE number of interspecific interactions (the sum of approaches and nips by G. holbrooki to P. australiensis) in low and high density treatments during daytime (open bars) and night-time (closed bars) observation periods. N ¼ 12.

30 Mean interspecific interactions

(a) 25 20 15 10 5 0 0

2

4

6

8

10

12

2

4

6

8

10

12

(b) 100

Time (%)

80 60 40 20 0 0

Carapace length (mm) Figure 3. The relationship between (a) carapace length of P. australiensis and number of interspecific interactions (the sum of approaches and nips by G. holbrooki) in low and high density controls (grey diamonds, dashed line) and treatments (black squares, solid line) and (b) carapace length and shelter use of P. australiensis exposed to low density (grey diamonds, dashed line) and high density (black squares, solid line) treatments and controls.

92

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

80

(a)

70 60 50 40 30 20 10 0

6

Low density control

Low density treatment

High density control

High density treatment

Low density treatment

High density control

High density treatment

Low density treatment

High density control

High density treatment

(b)

Time (%)

5 4 3 2 1 0

35

Low density control

(c)

30 25 20 15 10 5 0

Low density control

Figure 4. Mean ± SE time spent by P. australiensis engaging in (a) shelter use, (b) swimming and (c) foraging in the low and high density treatments and controls during daytime (open bars) and night-time (closed bars) observation periods. N ¼ 12.

conditions (GLMM: F1,349 ¼ 33.67, P < 0.0001; Fig. 4c). While there was no significant effect of G. holbrooki presence or absence (F1,348 ¼ 0.07, P ¼ 0.79), there was an effect of density: P. australiensis in the high density control and treatment conditions spent more time foraging than P. australiensis in the low density control and treatment conditions (F1,349 ¼ 6.37, P ¼ 0.01; Fig. 4c). There was no significant effect of size on foraging behaviour (F1,343 ¼ 0.38, P ¼ 0.54). DISCUSSION The threat sensitivity predator avoidance hypothesis (Helfman, 1989) states that the magnitude of antipredator behaviours exhibited by prey should be directly proportional to the level of predation risk to which they are exposed. In this study, we tested the effects of predator density and diel cycle on the consumptive and nonconsumptive effects (CE and NCEs) of predation by the

invasive eastern mosquito fish, G. holbrooki, on the native Australian glass shrimp, P. australiensis. As predicted, the frequency of direct behavioural interactions between predator and prey significantly increased with predator density and during the day when G. holbrooki is active; however, these factors did not significantly influence the number of P. australiensis actually consumed. Paratya australiensis displayed antipredator behaviours in response to the presence of G. holbrooki, specifically an increase in shelter usage and a reduction in swimming. Yet the magnitude of these antipredator behaviours did not depend on predator density or the time of day. In contrast, whether P. australiensis undertook foraging was primarily dependent on the total density of animals present and not species composition, with the number of P. australiensis foraging being greatest at a high density of both fish and shrimp. In addition, no interactions between predator density and diel cycle predicted any of these behaviours. Therefore, P. australiensis, by not altering its behaviours in relation to density or diel cycle, responded

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

appropriately to the level of predation risk to which it was actually exposed. Predator density has previously been either positively or negatively correlated with the frequency of direct interactions between predator and prey and the number of prey killed (Chang & Snyder, n, 2011; Sih 1979; Vucetich et al., 2002). In 2004; Gregory & Quijo the present study, the frequency of approaches and nips that P. australiensis received from G. holbrooki was significantly greater at a higher predator density. Additionally, significantly more nips and approaches were recorded during the day than at night, regardless of G. holbrooki density. This concurs with previous studies which have reported a nocturnal decrease in interactions between G. holbrooki and P. australiensis most probably due to a decrease in the predator's activity levels (Bool et al., 2011). However, despite the fact that predator density is commonly used as a proxy for predation risk (Ferrari et al., 2009; Paterson et al., 2013), our results demonstrate that predator density and diel cycle did not alter the number of prey consumed, and therefore the level of predation risk. These findings are consistent with the occurrence of mutual interference, whereby aggressive and competitive behavioural interactions between predators can reduce foraging efficiency (Abrams, 1993; Mistri, 2003; Nilsson & Ruxton, 2004; Sih 1979). This possibility would be an interesting avenue of future research. Notably, smaller P. australiensis, particularly in the high density treatment, received significantly more nips and approaches from G. holbrooki, particularly when the predator's density was high. However, since scaled prey size did not differ between successful or unsuccessful predation attempts in either low or high density treatments, G. holbrooki were apparently not limited to consuming smaller P. australiensis. This finding differs from that of Capps et al. (2008), who reported selective predation by the western mosquito paeʻula fish, Gambusia affinis, upon smaller individuals of theʻo shrimp, Halocaridina rubra. However, G. affinis has also been found to select medium-sized and large prey that would provide a higher benefit according to the optimal diet theory (Bence & Murdoch, 1986; Metzke & Pederson, 2006; Pulliam, 1974). For example, Metzke and Pederson (2006) reported that despite the gape size of G. affinis being smaller than the total length of large Daphnia spp., G. affinis did not reject large prey more than medium-sized prey. It is also important to note that smaller P. australiensis in the current study spent less time in shelters when the density of G. holbrooki was high, increasing the likelihood of an encounter with G. holbrooki. One possibility for this result was that smaller P. australiensis were aggressively excluded from shelters by larger P. australiensis, as has previously been observed in crayfish and teleost fish (Figler, Cheverton, & Blank, 1999; Martin & Moore, 2007; Mikheev, Metcalfe, Huntingford, & Thorpe, 1994). Increased refuge use and changes in activity are common behavioural responses observed in crustacean prey in response to n, Lehtiniemi, & Viitasalo, predatory fish (Capps et al., 2008; Linde 2003) and are likely to be a means of reducing visibility to a visually foraging predator (Carey et al., 2011; Kunz, Ford, & Pung, 2006; Lammers, Warbuton, & Cribb, 2009; Maeger, Williamson, Loneragain, & Vance, 2005). In the present study, P. australiensis reacted to the presence of G. holbrooki by significantly increasing shelter use and reducing swimming. However, there was no difference in the magnitude of change in these behaviours observed in P. australiensis exposed to low and high densities of G. holbrooki. Additionally, and in contrast to other studies (Capps et al., 2008; Carey et al., 2011; Clark et al., 2003), there was no diurnal shift in shelter usage and swimming in P. australiensis, in response to G. holbrooki at either low or high density. While P. australiensis spent more time in shelters during the day than at night, this was consistent across all treatments and densities. However, studies

93

that have noted shifts in diurnal shelter usage and activities have predominantly examined field populations of shrimp and fish predators and therefore such behavioural changes may occur over a timescale not possible in our study (Weissburg, Smee, & Ferner, 2014). While swimming and shelter use were predominantly affected by the presence of G. holbrooki, the frequency of foraging by P. australiensis was determined by the overall density of animals present. Specifically, P. australiensis in the high density control and treatment conditions foraged more than shrimp in the low density control and treatment conditions. In contrast to Bool et al. (2011), there was no shift from diurnal to nocturnal foraging upon exposure to G. holbrooki. While P. australiensis exposed to a high density of G. holbrooki did forage less during the day and more at night than P. australiensis in the high density control condition (without G. holbrooki), the interaction between density and diel variation was not significant. As with shelter use and swimming, it is also possible that this is simply due to the length of time for which P. australiensis was exposed to G. holbrooki in the present study. However, as a generalist species, it is also possible that G. holbrooki competes with P. australiensis, as has previously been observed between larval killifish, Fundulus heteroclitus, and postlarval grass shrimp, Palaemontes pugio, whereby shrimp experienced reduced growth in the presence of fish (Cross & Stiven, 1997; Pyke, 2008). The number of P. australiensis predated upon did not differ significantly between low and high density treatments. Therefore, P. australiensis was responding appropriately to the level of predation risk to which it was exposed, supporting the threat-sensitive predator avoidance hypothesis (Helfman, 1989). Our results also suggest that P. australiensis was reacting to some form of alarm cue released by predated individuals as has been observed in some taxa (Buowma & Hazlett, 2001; Hazlett & McLay, 2005; Paterson et al., 2013), although this would require further testing. Proportional antipredator responses have previously been reported for wood frog tadpoles, Rana spp. (Schoeppner & Relyea, 2008; Van Buskirk & Arioli, 2002). For example, Schoeppner and Relyea (2008) manipulated predation risk by altering the number of predators that consumed a constant amount of prey and the number of prey consumed by a constant number of predators. Hiding and activity levels in Rana sylvatica tadpoles increased with the number of conspecifics consumed, but not with predator density (Schoeppner & Relyea, 2008). While a number of other studies have found similar responses regardless of predator density and cue intensity in other crustaceans (Paterson et al., 2013), there is also evidence in a number of crustacean taxa that some species are able to detect and respond appropriately to different levels of predation risk (Arundell, Wedell, & Dunn, 2014; Hill & Weissburg, 2013). For example, mud crabs, Panopeus herbstii, decreased movement, foraging and distribution patterns in response to exposure to high biomass predator treatments of multiple small and single large blue crabs, Callinectes sapidus, both of which exert different forms of predation pressure. However, mud crabs did not respond to single small blue crabs (Hill & Weissburg, 2013). Therefore, there currently remains evidence both in support of and in opposition to the threat sensitivity predator avoidance hypothesis (Helfman, 1989) in regard to crustacean species. In conclusion, this study demonstrates that the native crustacean P. australiensis is able to assess and respond appropriately to the level of predation risk posed by the invasive G. holbrooki in a relatively short period of time. This finding corresponds to other studies which have reported the innate or learned abilities of invertebrates to respond to the cues of both familiar and unfamiliar predatory fish (Paterson et al., 2013). However, we also provide the first evidence that even at low densities G. holbrooki has multiple simultaneous negative impacts on P. australiensis via direct

94

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95

predation, significant behavioural shifts and potentially competition for resources. It would be valuable to further extend this study to reflect the natural temporal variation in both P. australiensis and G. holbrooki densities as has been observed in the field (Pyke, 2008; Walsh & Mitchell, 1998). As such, the existence of multiple interactions may make coexistence between these species more challenging and suggests that traditional management techniques for invasive species, including reducing densities, may prove to be ineffective (Kornis, Carlson, Lehrer-Brey, & Vander Zanden, 2014; Mills, Rader, & Belk, 2004). Acknowledgments We thank Matthew Chard for his technical assistance. This project was supported by funding from the Institute of Conservation Biology & Environmental Management and the School of Biological Sciences, University of Wollongong. References Abrams, P. A. (1993). Why predation rate should not be proportional to predator density. Ecology, 74(3), 726e733. Anholt, B. R., Werner, E., & Skelly, D. K. (2000). Effect of food and predators on the activity of four larval ranid frogs. Ecology, 81(12), 3509e3521. Arthington, A. H., & Marshall, C. J. (1999). Diet of the exotic mosquitofish, Gambusia holbrooki, in an Australian lake and potential for competition with indigenous fish species. Asian Fisheries Science, 12, 1e16. Arundell, K. L., Wedell, N., & Dunn, A. M. (2014). The impact of predation risk and parasitic infection on parental care in brooding crustaceans. Animal Behaviour, 96, 97e105. Baber, M. J., & Babbitt, K. J. (2003). The relative impacts of native and introduced predatory fish on a temporary wetland tadpole assemblage. Oecologia, 136, 289e295. Barney, R. L., & Anson, B. J. (1921). Seasonal abundance of the mosquito destroying top-minnow, Gambusia affinis, especially in relation to male frequency. Ecology, 2, 53e69. http://dx.doi.org/10.2307/1929529. Bence, J. R., & Murdoch, W. W. (1986). Prey size selection by the mosquitofish: relation to optimal diet theory. Ecology, 67(2), 324e336. Benfield, M. C., & Minello, T. J. (1996). Relative effects of turbidity and light intensity on reactive distance and feeding of an estuarine fish. Environmental Biology of Fishes, 46, 211e216. Bollens, S. M., & Frost, B. W. (1989). Predator-induced diel vertical migration in a planktonic copepod. Journal of Plankton Research, 11, 1047e1065. Bollens, S. M., & Stearns, D. E. (1992). Predator-induced changes in the diel feeding cycle of a planktonic copepod. Journal of Experimental Marine Biology Ecology, 156, 179e186. Bool, J. D., Witcomb, K., Kydd, E., & Brown, C. (2011). Learned recognition and avoidance of invasive mosquitofish by the shrimp, Paratya australis. Marine and Freshwater Research, 62, 1230e1236. Bourdeau, P. E., Pangle, K. L., Reed, E. M., & Peacor, S. D. (2013). Finely tuned response of native prey to an invasive predator in a freshwater system. Ecology, 94(7), 1449e1455. Bowler, D. E., Yano, S., & Amano, H. (2013). The non-consumptive effects of a predator on spider mites depend on predator density. Journal of Zoology, 289, 52e59. Brennan, N. P., Leber, K. M., & Blackburn, B. R. (2007). Use of coded-wire and visible implant elastomer tags for marine stock enhancement of juvenile red snapper Lutjanus campechanus. Fisheries Research, 83, 90e97. Buowma, P., & Hazlett, B. A. (2001). Integration of multiple predator cues by the crayfish Orconectes propinquus. Animal Behaviour, 61, 771e776. Butler, M. T. (1989). Community responses to variable predation: field studies with sunfish and freshwater macroinvertebrates. Ecological Monographs, 59(3), 311e328. Capps, K. A., Turner, C. B., Booth, M. T., Lombardozzi, D. L., McArt, S. H., Chai, D., et al. (2008). Behavioral responses of the endemic shrimp Halocaridina rubra (Malacostraca: Atyidae) to an introduced fish, Gambusia affinis (Actinopterygii: Poeciliidae) and implications for the trophic structure of Hawaiian anchialine ponds. Pacific Science, 63(1), 27e37. Carey, C. C., Ching, M. P., Collins, S. M., Early, A. M., Fetzer, W. W., Chai, D., et al. (2011). Predator dependent diel migration by Halocaridina rubra shrimp (Malacostraca: Atyidae) in Hawaiian anchialine pools. Aquatic Ecology, 45, 35e41. Chang, G. C., & Snyder, W. E. (2004). The relationship between predator density, community composition, and field predation of Colorado potato beetle eggs. Biological Control, 31, 453e461. Clark, K. L., Ruiz, G. M., & Hines, A. H. (2003). Diel variation in predator abundance, predation risk and prey distribution in shallow-water estuarine habitats. Journal of Experimental Marine Biology and Ecology, 287, 37e55. Cook, B. D., Baker, A. M., Page, T. J., Grant, S. C., Fawcett, J. H., Hurwood, D. A., et al. (2006). Biogeographic history of an Australian freshwater shrimp, Paratya

australiensis (Atyidae): the role life history transition in phylogeographic diversification. Molecular Ecology, 15(4), 1083e1093. Cross, R. E., & Stiven, A. E. (1997). Experimental studies of competition in salt marsh juvenile fish (Fundulus heteroclitus Linneaus) and shrimp (Palaemonetes pugio Holthuis) in laboratory microcosms. Journal of Experimental Marine Biology and Ecology, 212, 133e146. Drake, D. L., Anderson, T. L., Smith, L. M., Lohraff, K. M., & Semlitsch, R. D. (2014). Predation of eggs and recently hatched larvae of endemic ringed salamanders (Ambystoma annulatum) by native and introduced aquatic predators. Herpetologica, 70(4), 378e387. € nen, H. (2008). Foraging patterns Eccard, J. A., Pusenius, J., Sundell, J., Halle, S., & Ylo of voles at heterogeneous avian and uniform mustelid predation risk. Oecologia, 157, 725e734. Ferrari, M. C. O., Messier, F., & Chivers, D. P. (2006). The nose knows: minnows determine predator proximity and density through detection of predator odours. Animal Behaviour, 72, 927e932. Ferrari, M. C. O., Sih, A., & Chivers, D. P. (2009). The paradox of risk allocation: a review and prospectus. Animal Behaviour, 78, 579e585. Figler, M. H., Cheverton, H. M., & Blank, G. S. (1999). Shelter competition in juvenile red swamp crayfish (Procambarus clarkii): the influence of sex differences, relative size and prior residence. Aquaculture, 178, 63e75. Foam, P. E., Mirza, R. S., Chivers, D. P., & Brown, G. E. (2005). Juvenile convict cichlids (Archocentrus nigrofasciatus) allocate foraging and antipredator behaviour in response to temporal variation in predation risk. Behaviour, 142, 129e144. n, P. A. (2011). The impact of a coastal invasive predator on Gregory, G. J., & Quijo infaunal communities: assessing the role of density and a native counterpart. Journal of Sea Research, 66, 181e186. Griffin, P. C., Griffin, S. C., Waroquiers, C., & Mills, L. S. (2005). Mortality by the moonlight: predation risk and snowshoe hare. Behavioural Ecology, 16(5), 938e944. Hazlett, B. A., & McLay, C. (2005). Responses to predation risk: alternative strategies in the crab Heterozius rotundifrons. Animal Behaviour, 69, 967e972.  between a Heavener, S. J., Carthey, A. J. R., & Banks, P. B. (2014). Competitive naïvete highly successful invader and a functionally similar native species. Oecologia, 175(1), 73e84. Helfman, G. S. (1989). Threat-sensitive predator avoidance in damselfishtrumpetfish interactions. Behavioural Ecology and Sociobiology, 24, 47e58. Hill, J. M., & Weissburg, M. J. (2013). Predator biomass determines the magnitude of non-consumptive effects (NCEs) in both laboratory and field environments. Oecologia, 172, 79e91. Inouye, B. D. (2001). Response surface experimental designs for investigating interspecific competition. Ecology, 8(10), 2696e2706. Komak, S., & Crossland, M. R. (2000). An assessment of the introduced mosquitofish (Gambusia affinis holbrooki) as a predator of eggs, hatchlings and tadpoles of native and non-native anurans. Wildlife Research, 27, 185e189. Kornis, M. S., Carlson, J., Lehrer-Brey, G., & Vander Zanden, M. J. (2014). Experimental evidence that ecological effects of an invasive fish are reduced at high densities. Oecologia, 175, 325e334. Kunz, A. K., Ford, M., & Pung, O. J. (2006). Behavior of the grass shrimp Palaemonetes pugio and its response to the presence of the predatory fish Fundulus heteroclitus. American Midland Naturalist, 155(2), 286e294. Lammers, J. H., Warbuton, K., & Cribb, B. W. (2009). Anti-predator strategies in relation to diurnal refuge use and exploration in the Australian freshwater prawn, Macrobrachium australiense. Journal of Crustacean Biology, 29(2), 175e182. Leyse, K. E., Lawler, S. P., & Strange, T. (2004). Effects of an alien fish, Gambusia affinis, on an endemic California fairy shrimp, Linderiella occidentalis: implications for conservation of diversity in fishless waters. Biological Conservation, 118, 57e65. Lima, S. L. (1998). Nonlethal effects in the ecology of predator-prey interactions. Bioscience, 48(1), 25e34. Lima, S. L., Valone, T. J., & Caraco, T. (1985). Foraging-efficiency-predation-risk tradeoff in the grey squirrel. Animal Behaviour, 33, 155e165. n, E., Lehtiniemi, M., & Viitasalo, M. (2003). Predator avoidance and behaviour Linde of Baltic littoral mysids Neomysis integer and Pranus flexuosus. Marine Biology, 143, 845e850. Maeger, J. J., Williamson, I., Loneragain, N. R., & Vance, D. J. (2005). Habitat selection of juvenile banana prawns, Penaeus merguiensis de Man: testing the roles of habitat structure, predators, light phase and prawn size. Journal of Experimental Marine Biology and Ecology, 324, 89e98. March, J. G., Pringle, C. M., Townsend, M. J., & Wilson, A. I. (2002). Effects of freshwater shrimp assemblages on benthic communities along an altitudinal gradient of a tropical island stream. Freshwater Biology, 47, 377e390. Martin, A. L., III, & Moore, P. A. (2007). Field observations of agonism in the crayfish Orconectes rusticus: shelter use in a natural environment. Ethology, 113, 1192e1201. Mella, V. S. A., Banks, P. B., & MacArthur, C. (2014). Negotiating multiple cues of predation risk in a landscape of fear: what scares free-ranging brushtail possums? Journal of Zoology, 294, 22e30. Metzke, B. A., & Pederson, C. L. (2006). Cladoceran community dynamics reflect temperature gradients in a cooling water reservoir. American Midland Naturalist, 167(2), 344e355. Mikheev, V. N., Metcalfe, N. B., Huntingford, F. A., & Thorpe, J. E. (1994). Size-related differences in behaviour and spatial distribution of juvenile Atlantic salmon in a novel environment. Journal of Fish Biology, 45(3), 379e386.

L. K. Lopez et al. / Animal Behaviour 117 (2016) 87e95 Mills, M. D., Rader, R. B., & Belk, M. C. (2004). Complex interactions between native and invasive fish: the simultaneous effects of multiple negative interactions. Oecologia, 141, 713e721. Mistri, M. (2003). Foraging behaviour and mutual interference in the Mediterranean shore crab, Carcinus aestuarii, preying upon the immigrant mussel Musculista senhousia. Estuarine, Coastal and Shelf Science, 56, 155e159. Morton, R. M., Beumer, J. P., & Pollock, B. R. (1988). Fishes of a subtropical Australian saltmarsh and their predation upon mosquitoes. Environmental Biology of Fishes, 21, 185e194. Navarrete, S. A., & Menge, B. A. (1996). Keystone predation and interaction strength: interactive effects of predators on their main prey. Ecological Monographs, 66, 409e429. Nilsson, P. A., & Ruxton, G. D. (2004). Temporally fluctuating prey and interfering predators a positive feedback. Animal Behaviour, 68, 159e165. Paterson, R. A., Pritchard, D. W., Dick, J. T. A., Alexander, M. E., Hatcher, M. J., & Dunn, A. M. (2013). Predator cue studies reveal strong trait-mediated effects in communities despite variation in experimental designs. Animal Behaviour, 86, 1301e1313. Pulliam, H. R. (1974). On the theory of optimal diets. American Naturalist, 108, 59e75. Pyke, G. H. (2008). Plague minnow or mosquito fish? A review of the biology and impacts of introduced Gambusia species. Annual Review of Ecology and Systematics, 39, 171e191. Richardson, A. J., Growns, J. E., & Cook, R. A. (2004). Distribution and life history of caridean shrimps in regulated lowland rivers in southern Australia. Marine and Freshwater Research, 55, 295e308. Schoeppner, N. M., & Relyea, R. A. (2008). Detecting small environmental differences: risk-response curves for predator-induced behavior and morphology. Oecologia, 154, 743e754. Segev, O., Mangel, M., & Blaustein, L. (2009). Deleterious effects of mosquitofish (Gambusia affinis) on fire salamander (Salamandra infraimmaculata) larvae. Animal Conservation, 12, 29e37. Shulse, C. D., & Semlitsch, R. D. (2014). Western mosquitofish (Gambusia affinis) bolster the prevalence and severity of tadpole tail injuries in experimental wetlands. Hydrobiologia, 723, 131e144. http://dx.doi.org/10.1016/ j.ejphar.2013.10.064.

95

Sih, A. (1979). Stability and prey behavioural responses to predator density. Journal of Animal Ecology, 48(1), 79e89. Sih, A. (1986). Antipredator responses and the perception of danger by mosquito larvae. Ecology, 67(2), 434e441. Sih, A., Bolnick, D. I., Luttbeg, B., Orrock, J. L., Peacor, S. D., Pintor, L. M., et al. (2010). Predator-prey naivete, antipredator behavior, and the ecology of predator invasions. Oikos, 119, 610e621. Smith, G. R., & Smith, L. E. (2015). Effects of western mosquitofish (Gambusia affinis) on tadpole production of gray treefrogs (Hyla versicolor). Herpetological Conservation and Biology, 10(2), 723e727. Stier, A. C., Geange, S. W., & Bolker, B. M. (2013). Predator density and competition modify the benefits of group formation in a shoaling reef fish. Oikos, 122, 171e178. Tambling, C. J., Minnie, L., Meyer, J., Freeman, E. W., Santymire, R. M., Adendorff, J., et al. (2015). Temporal shifts in activity of prey following large predator reintroductions. Behavioural Ecology and Sociobiology, 69, 1153e1161. Van Buskirk, J., & Arioli, M. (2002). Dosage response of an induced defense: how sensitive are tadpoles to predation risk? Ecology, 83(6), 1580e1585. Vance-Chalcraft, H. D., Soluk, D. A., & Ozburn, N. (2004). Is prey predation risk influenced more by increasing predator density or predator species richness in stream enclosures? Oecologia, 139, 117e122. Vucetich, J. A., Peterson, R. O., & Schaefer, C. L. (2002). The effect of prey and predator densities on wolf predation. Ecology, 83(11), 3003e3013. Walsh, C. J., & Mitchell, B. D. (1995). The freshwater shrimp Paratya australiensis (Kemp, 1917) (Decapoda: Atyidae) in estuaries of south-western Victoria, Australia. Marine and Freshwater Research, 46, 959e965. Walsh, C. J., & Mitchell, B. D. (1998). Factors associated with variations in abundance of epifaunal caridean shrimps between and within estuarine seagrass meadows. Marine and Freshwater Research, 49, 769e777. Weissburg, M., Smee, D. L., & Ferner, M. C. (2014). The sensory ecology of nonconsumptive predator effects. American Naturalist, 184(2), 141e157. Zulian, E., Bisazza, A., & Marin, G. (1993). Determinants of size in male eastern mosquito fish (Gambusia holbrooki): inheritance and plasticity of a sexual selected character. Bollotino di zoologia, 60, 317e322.