Two s company, three s a crowd: Food and shelter limitation outweigh the benefits of group living in a shoaling fish

Transcription

1 Ecology, 94(5), 2013, pp Ó 2013 by the Ecological Society of America Two s company, three s a crowd: Food and shelter limitation outweigh the benefits of group living in a shoaling fish JOHN R. FORD 1 AND STEPHEN E. SWEARER Department of Zoology, Melbourne University, Parkville, Victoria 3010 Australia Abstract. Identifying how density and number-dependent processes regulate populations is important for predicting population response to environmental change. Species that live in groups, such as shoaling fish, can experience both direct density-dependent mortality through resource limitation and inverse number-dependent mortality via increased feeding rates and predator evasion in larger groups. To investigate the role of these processes in a temperate reef fish population, we manipulated the density and group size of the shoaling species Trachinops caudimaculatus on artificial patch reefs at two locations with different predator fields in Port Phillip Bay, Australia. We compared mortality over four weeks to estimates of predator abundance and per capita availability of refuge and food to identify mechanisms for density or number dependence. Mortality was strongly directly density dependent throughout the experiment, regardless of the dominant predator group; however, the limiting resource driving this effect changed over time. In the first two weeks when densities were highest, densitydependent mortality was best explained by refuge competition and the abundance of benthic predators. During the second two weeks, food competition best explained the pattern of mortality. We detected no effect of group size at either location, even where pelagic-predator abundance was high. Overall, direct density effects were much stronger than those of group size, suggesting little survival advantage to shoaling on isolated patch reefs where resource competition is high. This study is the first to observe a temporal shift in density-dependent mechanisms in reef fish, and the first to observe food limitation on short temporal scales. Food competition may therefore be an important regulator of postsettlement reef fish cohorts after the initial intense effects of refuge limitation and predation. Key words: density-dependent mortality; group size effect; population regulation; Port Phillip Bay, Australia; resource limitation; safety in numbers; southern hulafish; Trachinops caudimaculatus. INTRODUCTION Density-dependent processes resulting from resource limitation are fundamental to population regulation in natural systems (Murdoch 1994, Turchin 1999), and can influence growth (Peters and Barbosa 1977, Sedinger et al. 1998), reproductive output (Dhondt et al. 1992, Ostfeld et al. 1993), and ultimately survival (Skogland 1985, He and Duncan 2000). These demographic rates will appear to be density independent until a critical resource is saturated (Osenberg et al. 2002), beyond which the per capita availability of this resource declines, resulting in direct density dependence. Although populations may exist for long periods at densities below the point at which direct density dependence occurs (Sale and Tolimieri 2000), it is widely accepted that, while density-independent processes cause short-term variability in demographic rates, population abundances are ultimately regulated by density-dependent processes (Hixon et al. 2002). Manuscript received 30 October 2012; revised 19 December 2012, accepted 2 January Corresponding Editor: L. Ciannelli. 1 jford@unimelb.edu.au 1069 In marine reef fish, refuge from predators is widely identified as the most important limiting resource driving direct density-dependent mortality (Holbrook and Schmitt 2002, Steele and Forrester 2005, Johnson 2006a), and the most widely identified immediate agent of mortality is predation (Webster 2002, Almany and Webster 2006, White and Warner 2007b). The strength of competition for predator refuges will often depend upon the ratio of numbers of prey to the number of available refuges (Samhouri et al. 2009), where an increase in habitat complexity can offset densitydependent patterns entirely (Shima and Osenberg 2003, Johnson 2007a). Evidence in reef fish for direct densitydependent mortality through food limitation is not strong, and if present, will likely result in a delayed effect on survival as competition for food gradually reduces condition and hence increases mortality risk (Webster 2004, Hixon and Jones 2005, Johnson 2007b). For gregarious species, such as many reef fish, the competitive costs of increased density can be balanced by the benefits derived from group living strategies (Shima 2001). Shoaling behavior is one such strategy with observed benefits to predator detection and avoidance (Ioannou and Krause 2008), foraging efficiency (Pitcher et al. 1982), and survival (White et al.

2 1070 JOHN R. FORD AND STEPHEN E. SWEARER Ecology, Vol. 94, No ). Such aggregative behavior is often termed the safety in numbers effect (Beauchamp 2008), where individuals can devote less time to being vigilant and more time to feeding while the group maintains a similar ability to detect predators (Lima 1995) and individuals reduce predation risk through encounter dilution and confusion effects (Roberts 1996). Survival benefits from larger shoal sizes, or inverse number-dependent mortality, have been observed in a number of reef fish species at the scale of individual aggregations (Booth 1995, Sandin and Pacala 2005, White and Warner 2007b, Stier et al. 2012); however, aggregation of predators to larger prey patches (Anderson 2001, White 2007) and refuge limitation at high densities can also facilitate the reverse effect of direct density-dependent mortality. The two processes are not mutually exclusive; however, in many experimental observations it is very difficult to quantify the role of both processes simultaneously (Shima 2002). In a review of postsettlement mortality in shoaling reef fish, White et al. (2010) highlighted this disparity between studies demonstrating direct density-dependent and inverse number-dependent mortality. Direct density-dependent mortality is more likely where predator foraging is on a similar scale to prey aggregation (e.g., resident predators foraging on shoaling wrasse; Overholtzer-McLeod 2006), and in non-aggregating species that compete for limited refuge space (e.g., non-shoaling gobies; Forrester and Steele 2004). The more intense competition observed on patch reefs results in stronger density-dependent mortality compared to continuous reefs (White et al. 2010). Mortality in aggregating species, on the other hand, would be more likely to be inversely number dependent at the level of an individual aggregation on continuous reefs (e.g., pelagic predators foraging on shoaling damselfish; Sandin and Pacala 2005), but density dependent across a larger area of a predator s foraging domain (White and Warner 2007b). Aggregations on small, widely spaced patch reefs are less likely to receive survival benefits from larger shoal sizes, as pelagic predators target larger aggregations for a more profitable return (White et al. 2010). This framework explains the observations of most previous investigative studies of postsettlement mortality in reef fishes; however, there may be more complex effects of multiple predators with different behaviors and interactions with prey that require further investigation. Shoaling reef fish face predation from two broad functional groups: pelagic and benthic predators (Hixon and Carr 1997), which differ in their hunting mode, sizes of their foraging domains, and frequency of interaction with prey. Pelagic predators have large foraging domains, are transient, and are characterized by a chase and capture foraging mode. Benthic predators have foraging domains often of similar scale to prey, with a higher chance of encounter with prey and a predominantly ambush hunting mode. Shoaling is beneficial to prey survival in the presence of a single predator group, as prey can shoal above the benthic predator foraging domain and reduce attacks, and can utilize benthic predator-free reef refuge when threatened by only pelagic predators (Ford and Swearer 2012). When both predator groups are present, however, shoaling can result in risk enhancement for prey (Ford and Swearer 2012) and facilitate density-dependent mortality (Hixon and Carr 1997). Given the complex interaction among predators, prey, and habitat determining the strength and direction of density-dependent processes, there is broad scope to test the predictions of White et al. (2010) in other systems. In this study we investigated the roles of density- and number-dependent mortality in a shoaling temperate reef fish, Trachinops caudimaculatus, a social aggregator that utilizes the reef matrix as refuge. We manipulated both density and shoal size on artificial patch reefs in locations with either a benthic-dominant or pelagicdominant predator regime, and observed prey mortality after two and four weeks. We tested the predictions for four hypotheses that may explain density-dependent mortality in this species: (1) refuge limitation alone (Holbrook and Schmitt 2002), (2) food limitation alone (Johnson 2007b), (3) refuge and food limitation combined (Hixon and Jones 2005), and (4) a predator aggregation hypothesis (Anderson 2001). Under the framework of White et al. (2010), we expected the single aggregations of T. caudimaculatus on small patch reefs to show characteristics of direct density-dependent, but not inverse, number-dependent mortality. More specifically we predicted that: (1) Direct density-dependent mortality through refuge limitation (Hypothesis 1) will be detected in both locations in the first two weeks of the experiment due to high densities of prey and the use of small patch reefs which intensifies scramble competition for refuge, and (2) The safety in numbers effect will not be detected, even at the location where pelagic predators are dominant. Due to the isolation of each patch reef, aggregations were identified as discrete patches by pelagic predators with large foraging domains (White and Warner 2007b). Aggregations may therefore not be targeted equally, and benefits of increased group size are offset by targeting of larger shoals by pelagic predators (Hypothesis 4; see Methods: Comparing models of density-dependent mechanisms). (3) In the second two weeks of the experiments, mortality would be either weakly directly density dependent through continued refuge limitation (Hypothesis 1) or switch to density independence and be related only to predator abundance. METHODS Study species The southern hulafish, Trachinops caudimaculatus,isa small-bodied (,100 mm), short-lived (1 5 yr), and highly abundant fish on rocky reefs in southeast Australia. As a shoaling zooplanktivore, they are highly conspicuous and can form shoals of thousands of fish at high densities (.100 fish/m 2 ) in sheltered embayments

3 May 2013 TWO S COMPANY, THREE S A CROWD 1071 FIG. 1. Experimental design for the detection of density and shoal size effects on mortality in the shoaling reef fish Trachinops caudimaculatus, southern hulafish. Fish were placed on artificial reefs of either one or two units at four densities: very low (LL), with 12 fish/unit; low (L), with 25 fish/unit; high (H), with 50 fish/unit; and very high (HH), with 100 fish/unit. Because fish on a reef shoaled together regardless of the number of reef units, we could compare two shoal sizes within each density treatment. At all densities, the shoal size doubled from the small to large treatment, and hence, the shoal size treatment represents a doubling of shoal size independent of density. (Appendix A: Fig. 1A; Hunt et al. 2011). Fish are highly site attached after settlement and no movement has been recorded across expanses of open water greater than 20 m (Ford and Swearer 2012). Artificial reefs In November 2008 we constructed eight artificial reefs separated by.100 m of open sand or muddy bottom, in two locations in Port Phillip Bay, Australia: Altona in the north ( S, E) and Carrum in the east ( S, E). At both locations, four of the reefs consisted of a single 1-m 3 unit and the other four of two 1-m 3 units placed immediately adjacent (,30 cm) to one another. Artificial reefs mimicked natural reef habitat by providing both small refuge spaces, larger cavities and overhangs, and were naturally fouled with encrusting and sessile organisms before the experiment began (for further details see the Appendix). The two locations were chosen through prior knowledge that each represented very different predator fields. The Altona location is dominated by benthic predators that colonize artificial reefs from drifting patches of macroalgae that are highly abundant in the northwest of Port Phillip Bay (Chidgey and Edmunds 1997). The Carrum location is dominated by pelagic predators due to its location closer to the ocean entrance and very low abundances of benthic predators in the surrounding sandy bottom. To confirm this, visual surveys of benthic and pelagic predators of T. caudimaculatus (for predator identification see the Appendix) were carried out at each location during each replicate run of the experiment. Density, reef size, and shoaling In January 2009, juvenile hulafish (,35 mm standard length [SL],,2 months postsettlement based on otolith back calculation) were caught from nearby natural reefs by SCUBA divers herding fish into a large-mouthed (1200-mm square) set net with a 2-m tapered codend. All fish translocated to a reef array were caught on the same day from the same site, and fish were mixed in a large bucket and drawn randomly before seeding reefs. Fish were placed onto reefs at four densities: very low (LL), with 12 fish/unit; low (L), with 25 fish/unit; high (H), with 50 fish/unit; and very high (HH), with 100 fish/unit (Fig. 1). Because the hulafish is a shoaling species, fish aggregated and generally formed single shoals regardless of the size of the reef. This allowed us to compare between differing shoal sizes within each density treatment. Thus, the experiment was a two-factor, fully crossed design (fish density [four levels] and shoal size [two levels]), resulting in a total of eight treatments allocated randomly to patch reefs at each location (Fig. 1). The experiment was run three times at each location resulting in three replicate runs per treatment, with run as a block effect. Fish survival was visually surveyed on days 14 and 28, after which all survivors were removed prior to the start of the next experimental run. Fourteenday time intervals were chosen to represent two distinct postsettlement periods: the initial phase of very high and often direct density-dependent mortality (Holbrook and Schmitt 2002, Almany and Webster 2006), and the following period of often higher survival, where fish had become acclimated to the reef environment (Ford and Swearer 2012). The 28-day period also provided an opportunity for possible food limitation to influence mortality rates through reduced condition. We calculated instantaneous daily per capita mortality rates (M ) between days 0 14 and using the methods of Johnson (2006a, b) and White and Warner (2007b) as follows: lnðinitial numberþ lnðfinal numberþ M ¼ t where t represents the 14 days between each survey period. This equation is similar to that used by Shima (2001), but adds a temporal component by calculating a daily rate. One reef at Altona was detected as an outlier and excluded from analysis due to 100% hulafish mortality at all sampling dates in all replicate runs regardless of density treatment. We initially used two four-way ANOVAs to test whether instantaneous mortality between days 0 14 and

4 1072 JOHN R. FORD AND STEPHEN E. SWEARER Ecology, Vol. 94, No. 5 ation rate has the potential to modify all four mechanisms by affecting detection of predators, and by reducing feeding efficiency and opportunity. We tested support for the four hypotheses against instantaneous mortality as the response variable, for days 0 14 and We selected the best-fit model from all subset models within each hypothesis based on Akaike s Information Criterion, corrected for small sample sizes (AIC c ; Johnson and Omland 2004). We compared the best-fit models to evaluate relative support for each hypothesis and calculated the relative contribution of each term using the Lindeman, Merenda, and Gold (LMG) method described by Gro mping (2007), which uses a sequential sum of squares approach averaged over all orderings of regressors. FIG. 2. Abundance (mean SE) of benthic predators and pelagic predators at the two experimental locations (Altona and Carrum, Port Phillip Bay, Australia) across all replicate runs. An asterisk indicates a significant difference (P, 0.05) between locations from t tests. There was no significant difference in abundance of either predator group among replicate runs differed among the following factors: location (fixed, two levels), density (fixed, four levels), shoal size (fixed, two levels), replicate run (random block, three levels), and included all interaction terms. Fully crossed designs revealed no significant interactions (P. 0.25) with replicate run, so to increase our degrees of freedom and power of detectability, we ran the models again excluding these interaction terms. All statistics were performed using SYSTAT v. 13 (SYSTAT Software 2009). Comparing models of density-dependent mechanisms We tested the four hypotheses that may explain the patterns of prey mortality in our study: (1) refuge limitation, (2) food limitation, (3) both food and refuge limitation acting together, and (4) predator aggregation. To compare these hypotheses, we constructed a series of multiple linear regression models from environmental data that best represented the mechanisms of each hypothesis (Appendix: Table A1; see Johnson and Omland 2004). We used combinations of five environmental variables in our candidate models: per capita refuge availability (number of free holes available), per capita food availability (zooplankton collected in tube traps), sedimentation rates (fine sediment collected in tube traps), and the abundance of benthic and pelagic predators (see the Appendix for details on how these variables were calculated).these variables were calculated separately for each period, using measures from day 0 and 14, respectively. The first three hypotheses represent resource limitation mechanisms, which may also be affected by the abundance and type of predators. The predator hypothesis poses that predators may drive density dependence by targeting larger prey groups. RESULTS Predator abundances at the two study locations Predator abundance differed significantly between locations (for pelagics, t 35 ¼ 2.4, P ¼ 0.02; for benthics, t 46 ¼ 7.9. P, 0.001; Fig. 2) and confirmed our prediction of different predator fields between locations. The abundance of benthic predators was more than four times higher at Altona ( fish/reef, mean 6 SE) than Carrum ( fish/reef ), and pelagic-predator abundance was more than five times higher at Carrum ( fish/reef ) than Altona ( fish/reef ). Visitation rates by pelagic predators were 2.73 higher at Carrum (40% of observations detected pelagic predators) compared to Altona (15% of observations). There was no relationship between predator abundance and the density or number of prey. Density- and number-dependent effects on mortality Per capita mortality was.40% higher in the first 14 days ( , mean 6 SE) compared to days ( ). Clear differences among density treatments were established by day 14 and persisted for the length of the experiment (Fig. 3). Days Stocking density had a significant effect on hulafish mortality rates (Table 1a). Instantaneous mortality in the HH treatment (100 fish/reef) was more than three times higher than the lowest density treatment. A subsequent post hoc Tukey s test showed significantly higher instantaneous mortality in the HH treatment over all other treatments and higher mortality in the H treatment (50 fish/reef) than the two lowest density treatments (Fig. 3). Average mortality was 30% higher at Altona than Carrum; however, the location effect was not significant. There was no significant effect of shoal size, replicate run, or interactions. Subsequent power analyses revealed a detectable difference in average group mortalities (at power (b) ¼ 0.8) of 0.1 for the shoal and location factors, indicating that we had sufficient power to detect a biologically meaningful effect. Days A significant effect of initial stocking density on mortality persisted in the second period

5 May 2013 TWO S COMPANY, THREE S A CROWD 1073 FIG. 3. Instantaneous mortality (mean SE) of hulafish under the four initial density treatments for days 0 14 and Instantaneous daily mortality rates were calculated by dividing the difference between the natural logs of the initial and final abundances by the number of days between surveys (see Methods). Bars identified by different letters were significantly different (P, 0.05) according to post hoc Tukey s tests. Data are pooled across locations and replicate runs. (Table 1b), where mortality in the HH treatment was almost twice that in the L and LL treatments. A subsequent post hoc Tukey s test showed significantly higher instantaneous mortality in the HH treatment over the L and LL treatments (Fig. 3). There was a significant replicate run effect, with a subsequent post hoc Tukey s test showing significantly higher instantaneous mortality in the third run over the first run. This replicate run effect was not correlated with any environmental factor measured such as predator abundance, food, or sediment. There was no significant effect of shoal size or location; however, there was a marginally nonsignificant interaction between location and shoal size. As this effect was of particular interest, we investigated the effect of shoal size separately for each location. At Carrum, the effect of shoal size was again marginally nonsignificant (ANOVA, F 1,22 ¼ 3.9, P ¼ 0.06). The trend, however, was the opposite of that predicted, as mortality was higher in larger shoal sizes. At Altona there was no effect of shoal size (ANOVA, F 1,19 ¼ 1.3, P ¼ 0.30). Comparing mechanistic models of density dependence Days Hypothesis 1 (refuge limitation) received the greatest support in explaining patterns of mortality in the first 14 days. The best-fit model included the terms per capita holes, benthic-predator abundance, and sedimentation rate, and explained 62% of the variation in instantaneous mortality (Table 2a). The per capita refuge term contributed the most to the model (partial r ), followed by benthic-predator abundance (0.17) and sediment load (0.03). There was significant support for all four hypotheses, most notably hypothesis 3 (combined limitation model). However, the hypothesis 3 model is only a more complex form of the best-fit hypothesis 1 model, with the addition of the per capita food term not adding any further explanatory power. Days Much less variation was explained by our hypotheses in the second period. Hypothesis 2 (food limitation) received the greatest support (Table 2b). The best fit model included the terms per capita food and sediment load and explained 29% of the variation in instantaneous mortality between days 14 and 28. The per capita food term contributed the most to the model (partial r ), followed by sediment (0.04). There was significant support for hypothesis 3 (combined limitation); however, similar to the first period, the significant model added a single term to the food limitation model without adding further explanatory power. There was no significant support for hypotheses 1 (refuge limitation) or 4 (predator aggregation). DISCUSSION Direct density dependence can act to regulate populations while simultaneously interacting with inverse density-dependent effects of group size (Shima 2002, White et al. 2010). As a result, some studies have been left with inconclusive or contrasting results when testing for effects of density vs. number effects (Connell 2000, Shima 2001). In this study we tested for the presence of both processes in shoaling reef fish populations subject to different predator regimes, and attempted to identify the mechanisms responsible for these patterns of mortality. Our results confirm much of the theoretical framework of White et al. (2010) in predicting strong direct density- Results of four-factor ANOVA testing the effects of density and shoal size on instantaneous hulafish Trachinops caudimaculatus mortality from (a) 0 14 days and (b) 0 28 days. TABLE 1. Source df MS F P a) Days 0 14 Location Density ,0.001 Shoal size 1, Location 3 density Location 3 shoal size 1, Density 3 shoal size 3, Location 3 density 3 shoal size Replicate run 2, Error b) Days Location Density Shoal size 1, Location 3 density 3, Location 3 shoal size Density 3 shoal size Location 3 density 3 shoal size 3, Replicate run Error Notes: The main factors were location (Altona and Carrum), initial density (four treatments), shoal size (two treatments), and replicate run (blocking factor). A fully crossed model revealed no significant interactions with run (P. 0.25), so these interaction terms were excluded from the final analysis.

6 1074 JOHN R. FORD AND STEPHEN E. SWEARER Ecology, Vol. 94, No. 5 Best-fit multiple regression models from lowest AIC c score, predicting instantaneous hulafish mortality from five environmental factors affecting survival. TABLE 2. Hypothesis and terms Effect R 2 (adj) DAIC c a) Days 0 14 Refuge limitation Per capita refuge Benthic predators Food limitation Per capita food Benthic predators Combined limitation Per capita refuge Per capita food Benthic predators Predator aggregation Benthic predators b) Days Refuge limitation Per capita refuge Pelagic predators Food limitation Per capita food Combined limitation Per capita refuge Per capita food Predation Pelagic predators Notes: Factors are per capita availability of refuge or food, abundance of benthic or pelagic predators, and sedimentation rates. We tested density-dependent mechanisms as model hypotheses described in the Appendix (Table A1) and selected the best-fit (lowest AIC c ) model from each hypothesis for each sampling period. Positive and negative signs in the effect column refer to the direction of that term s effect on mortality rates. The hypothesis highlighted in boldface type was chosen as the model of best support. dependent mortality related to refuge competition on patch reefs. We expand this theory by observing densitydependent mortality under very different predator fields and producing the first evidence for short-term food limitation driving direct density-dependent mortality in shoaling reef fish, an effect that has traditionally been thought to manifest on much longer timescales. We did not observe a benefit to larger shoal sizes, even at the location dominated by pelagic predators, with weak evidence instead for an increase in mortality in larger shoal sizes. Thus, the strength of resource limitation on patch reefs far outweighed the benefits of increased predator evasion and feeding efficiency in larger group sizes. Shifting mechanisms of density-dependent mortality As predicted under the framework of White et al. (2010), we observed strong direct density-dependent mortality related to refuge competition in the first two weeks of the experiment. Competition for shelter is a common regulating mechanism in reef fish communities immediately postsettlement (Osenberg et al. 2002), and persisted in our study on similar time scales as previous studies (i.e., one to two weeks). Direct density-dependent patterns occurred at both locations, despite very different relative abundances of benthic and pelagic predators. The strength of mortality was, however, related to benthic, but not pelagic-predator abundance. This suggests that the presence of both guilds may limit spatial refuge for prey (Ford and Swearer 2012) and facilitate density-dependent mortality (Hixon and Carr 1997), but benthic predators are the main source of prey mortality when refuges are limiting. We identified a short-term shift in the nature of a limiting resource regulating populations through direct density-dependent mortality, from refuge competition to food limitation. T. caudimaculatus populations were subsequently regulated by food accessibility in the second two weeks, with the refuge limitation hypothesis receiving no significant support. The observed food limitation effect was weaker than the strength of refuge competition, although it is likely to persist longer than the effects of shelter limitation (Hixon and Jones 2005). Direct density-dependent mortality via food competition was not predicted in this study as food limitation has not been directly related to density-dependent postsettlement mortality in reef fishes. There is evidence for density-dependent food limitation in adult populations (Forrester 1990), and food availability plays an important role in juvenile growth (Tupper and Boutilier 1995, Johnson 2007b) and size-selective mortality of smaller individuals (Carr and Hixon 1995, Holmes and Mc- Cormick 2006). While Hixon and Jones (2005) observed direct density dependence attributed to food competition on patch reefs, the effect took 17 months to manifest as densities were reduced to the long-term carrying capacity of patch reefs. Instead, we observed limitation on shorter time scales, suggesting food competition may be an important regulator of postsettlement reef fish cohorts after the initial intense effects of refuge limitation and predation. Absence of a group size effect We confirmed our second prediction and did not observe an inverse number-dependent, or group size, mortality effect, even at the location of higher pelagicpredator abundance. While shoaling provides survival benefits against predation from pelagic predators through enhanced predator detection, attack abatement and risk dilution effects (Pitcher and Parrish 1993), White et al. (2010) proposed that this benefit is unlikely to be detected on small isolated patch reefs. Our results confirm this prediction, although we did not detect any targeting of larger shoal sizes by pelagic predators, and there was little support for hypothesis 4 (predator aggregation) throughout the experiment. The absence of an inverse number-

7 May 2013 TWO S COMPANY, THREE S A CROWD 1075 dependent effect corresponds to other manipulative studies on patch reefs, none of which have demonstrated a benefit of larger shoal sizes. Any small effects of shoal size were likely obscured by the very strong population regulation through direct density-dependent competition (Shima 2001). However, there is likely strong selective advantages to shoaling that are not reflected in short-term mortality rates, but allow the behavior to persist in this species. Advantages such as maximizing feeding time, enhanced food detection, and improved reproductive success may play out on much longer time scales, or be more pronounced in continuous reef systems with multiple aggregations. Furthermore, we did not test the benefit of shoaling over not shoaling, and it is possible that survival benefits are unrelated to the size of the shoal, or only evident at very small shoal sizes. Although we did not receive significant support for the predator aggregation hypothesis in explaining direct density-dependent mortality, it is possible that an undetected aggregative functional response by predators was responsible for the absence of a group size effect. This may occur in two separate ways. Firstly, if pelagic predators exhibit an aggregative response on a scale larger than the reef array, visitation rates to patch reefs could be lower than those experienced on larger or natural reefs with greater prey abundance (e.g., Overholtzer-McLeod 2006). This may reduce pelagic predation pressure on patch reefs to the point where the benefits of shoaling are insignificant, although such low levels of pelagic predation are unlikely given the high prey mortality rate in this study and the observed synergistic predation by both predator guilds (Ford and Swearer 2012). Secondly, if instead, pelagic predators are exhibiting an aggregative prey response at the scale of the patch reef array (as predicted by White et al. [2010]), increased visitation and strikes at larger group sizes may negate the benefits of shoaling. We did not observe a relationship between pelagic-predator visitation and shoal size, or pelagic-predator abundance and mortality. However, our use of SCUBA observations in detecting pelagic-predator visitation rates may provide unreliable results through attraction or avoidance of divers by predators. Use of underwater cameras (e.g., Sandin and Pacala 2005, Overholtzer-McLeod 2006) would provide a more reliable measure of pelagicpredator visitation. Given our observations and the findings of Ford and Swearer (2012) in the same system, where pelagic predators caused risk enhancement only to large groups of prey, it is possible that predator aggregation on larger shoals resulted in the absence of a group size effect. The role of patchy habitat on predator and prey behavior The use of isolated patch reefs in this study likely contributed to the overwhelming effect of density on prey mortality (Sandin and Pacala 2005, White et al. 2010). Although the densities of T. caudimaculatus stocked to patch reefs were well within those observed on natural reefs (Hunt et al. 2011; J. R. Ford, unpublished data), these comparisons do not take into the potential of prey to share refuges with nearby aggregations on continuous reefs (White et al. 2010). Therefore, spot densities of prey on continuous reefs may not adequately represent the same prey to refuge ratio as on patch reefs. Prey are confined to a single patch and limited by the amount of shelter resource, leading to much more intense competition for refuges on patch reefs at the same density. The patch vs. continuous reef effect is further compounded by an increase in scramble competition for shoaling species on reef edges, where an aggregation can only access the refuges on the reef edge closest to them (White et al. 2010). Competition for the limited number of refuges on a reef edge will be enhanced when the shoal is larger, regardless of the size the reef or the number of available refuges in a wider area. In effect, the doubling of prey number in the large-shoal treatment could have doubled the effect of scramble competition for the refuges on that reef edge. This may explain the marginally nonsignificant trend of higher mortality in larger shoal sizes at the pelagic predator-dominated site. The dynamics we observe on artificial patch reefs are likely to be very similar to those observed on naturally patchy reefs, or continuous reefs where suitable habitat is patchy. However, the strength of direct density dependence may instead be reduced on natural continuous reefs with abundant refuge, and increased benefits of group size. The effect of food limitation, like shelter competition, may also be magnified on patch reefs. As an almost continuously feeding shoaling zooplanktivore, feeding opportunity is important in maintaining longer term natural populations of T. caudimaculatus (Hunt et al. 2011). The strategy of shoaling, while providing feeding opportunity, is also energetically costly, and reduced nutrition could reduce an individual s effectiveness in predator evasion (White and Warner 2007a), explaining the lagged effect of per capita food availability on mortality in the second two weeks of the experiment. Given the dynamic foraging ranges of T. caudimaculatus on natural continuous reefs, it is likely that shoals migrate to areas with high zooplankton abundance, as observed in its sister species T. taeniatus (Gregson and Booth 2005). There is no opportunity to forage in a similar manner on patch reefs that does not increase predation risk by leaving the safety of the reef edge. Therefore, the competition for food is also likely to be strengthened on patch reefs, with food per prey ratios much lower than those experienced on continuous reefs. ation was a minor but significant factor in both periods, where increased sedimentation rates resulted in an increase in hulafish mortality. Turbidity can affect both feeding success and subsequently condition (Wenger et al. 2012), and visual cues to initiate predator evasion, particularly from pelagics when shoaling (Miner and Stein 1996, Vogel and Beauchamp 1999). However,

8 1076 JOHN R. FORD AND STEPHEN E. SWEARER Ecology, Vol. 94, No. 5 turbid conditions may provide an advantage for zooplanktivores in avoiding piscivorous predators, as feeding by piscivores is more negatively affected by turbidity than plantivores (De Robertis et al. 2003). Furthermore, the relationship between turbidity and sedimentation rates measured by sediment traps has been questioned in recent studies (Storlazzi et al. 2011, Brown et al. 2012), particularly with regards to the under-recording of resuspended fine particles that cause turbidity. Regardless, it is clear that sedimentation rates, either directly or as a proxy for another factor such as turbidity, were important in explaining hulafish mortality. Hunt et al. (2011) found a similar effect of sedimentation on the recruitment of T. caudimaculatus on natural reefs, where low recruitment was attributed to reduced feeding opportunity and capacity to detect pelagic predators in turbid environments. The mechanism of sedimentation affecting mortality may also have shifted over the course of the experiment, where predator detection was hampered by reduced visibility in the first two weeks and the ability to forage in the second. Conclusion Density-dependent processes strongly influence the population dynamics of T. caudimaculatus on small-scale patch reefs, and the nature of the limiting resource changed from refuge to food over a short two-week time period. The strongest density-dependent effects in reef fish often occur immediately postsettlement when densities are at their highest, and our study supports the prediction that refuge competition is the primary driver during this period. However, we demonstrate that food competition can also occur over short periods and results in direct density-dependent mortality. Both mechanisms of direct density dependence (refuge and food limitation) are likely strengthened on patch reefs compared to continuous reefs, where shoals can respond to patchiness in resource abundance and hence reduce competition. Lastly, this study evaluated the dynamics of fish populations on small patch reefs over a short period of time. Selecting appropriate spatial and temporal scales of observation will determine the applicability of the outcomes of any study (Ray and Hastings 1996). It is well established that a scaling-up of results will require similar mechanisms to be driving the systems at all scales (Steele and Forrester 2005). Having established the mechanisms of density-dependent mortality in T. caudimaculatus, the next step is to determine if shelter and food limitation regulate populations at larger spatial and temporal scales. ACKNOWLEDGMENTS The authors thank invaluable field hands: Christian Jung, Dean Chamberlain, Evan Hallein, Matt Le Feuvre, John Ahern and Seann Chia. Michael Sams and Mick Keough provided constructive statistical advice. We thank J. Wilson-White and one anonymous reviewer for their constructive comments on the manuscript. Operational costs were covered by grants from the Holsworth Wildlife Research Endowment to J. R. Ford and the Australia and Pacific Science Foundation to S. E. Swearer. Animal ethics approval was granted under Melbourne University AEC The authors declare that they have no conflict of interest. LITERATURE CITED Almany, G. R., and M. S. Webster The predation gauntlet: early post-settlement mortality in reef fishes. Coral Reefs 25: Anderson, T. W Predator responses, prey refuges, and density-dependent mortality of a marine fish. Ecology 82: Beauchamp, G What is the magnitude of the group-size effect on vigilance? Behavioral Ecology 19:1361. Booth, D. J Juvenile groups in a coral-reef damselfish: density-dependent effects on individual fitness and population demography. Ecology 76: Brown, N. K., S. G. Smithers, C. T. Perry, and P. V. Ridd A field-based technique for measuring sediment flux on coral reefs: application to turbid reefs on the Great Barrier Reef. Journal of Coastal Research 28: Carr, M. H., and M. A. Hixon Predation effects on early post-settlement survivorship of coral reef fishes. Marine Ecology Progress Series 124: Chidgey, S., and M. Edmunds Standing crop and nutrient content of macrophytes in Port Phillip Bay. Technical Report Number 32: Port Phillip Bay Environmental Study. Commonwealth Scientific and Industrial Research Organisation (CSIRO), Clayton South, Victoria, Australia. Connell, S. D Is there safety-in-numbers for prey? Oikos 88: De Robertis, A., C. H. Ryer, A. Veloza, and R. D. Brodeur Differential effects of turbidity on prey consumption of piscivorous and planktivorous fish. Canadian Journal of Fisheries and Aquatic Sciences 60: Dhondt, A. A., B. Kempenaers, and F. Adriaensen Density-dependent clutch size caused by habitat heterogeneity. Journal of Animal Ecology 61: Ford, J. R., and S. E. Swearer Shoaling behaviour enhances risk of predation from multiple predator guilds in a marine fish. Oecologia. Forrester, G. E Factors influencing the juvenile demography of a coral reef fish. Ecology 71: Forrester, G. E., and M. A. Steele Predators, prey refuges, and the spatial scaling of density-dependent prey mortality. Ecology 85: Gregson, M., and D. Booth Zooplankton patchiness and the associated shoaling response of the temperate reef fish Trachinops taeniatus. Marine Ecology Progress Series 299: Grömping, U Estimators of relative importance in linear regression based on variance decomposition. American Statistician 61:139. He, F., and R. P. Duncan Density-dependent effects on tree survival in an old-growth Douglas fir forest. Journal of Ecology 88: Hixon, M. A., and M. H. Carr Synergistic predation, density dependence, and population regulation in marine fish. Science 277: Hixon, M. A., and G. P. Jones Competition, predation, and density-dependent mortality in demersal marine fishes. Ecology 86: Hixon, M. A., S. W. Pacala, and S. A. Sandin Population regulation: historical context and contemporary challenges of open vs. closed systems. Ecology 83: Holbrook, S. J., and R. J. Schmitt Competition for shelter space causes density-dependent predation mortality in damselfishes. Ecology 83: Holmes, T. H., and M. I. McCormick Location influences size-selective predation on newly settled reef fish. Marine Ecology Progress Series 317:

9 May 2013 TWO S COMPANY, THREE S A CROWD 1077 Hunt, T. L., J. R. Ford, and S. E. Swearer Ecological determinants of recruitment to populations of a temperate reef fish, Trachinops caudimaculatus (Plesiopidae). Marine and Freshwater Research 62: Ioannou, C. C., and J. Krause Searching for prey: the effects of group size and number. Animal Behaviour 75: Johnson, D. W. 2006a. Predation, habitat complexity, and variation in density-dependent mortality of temperate reef fishes. Ecology 87: Johnson, D. W. 2006b. Density dependence in marine fish populations revealed at small and large spatial scales. Ecology 87: Johnson, D. W. 2007a. Habitat complexity modifies postsettlement mortality and recruitment dynamics of a marine fish. Ecology 88: Johnson, D. W. 2007b. Combined effects of condition and density on post-settlement survival and growth of a marine fish. Oecologia 155: Johnson, J. B., and K. S. Omland Model selection in ecology and evolution. Trends in Ecology and Evolution 19: Lima, S. L Back to the basics of anti-predatory vigilance: the group-size effect. Animal Behaviour 49: Miner, J. G., and R. A. Stein Detection of predators and habitat choice by small bluegills: effects of turbidity and alternative prey. Transactions of the American Fisheries Society 125: Murdoch, W. W Population regulation in theory and practice. Ecology 75: Osenberg, C. W., C. M. St. Mary, R. J. Schmitt, S. J. Holbrook, P. Chesson, and B. Byrne Rethinking ecological inference: density dependence in reef fishes. Ecology Letters 5: Ostfeld, R., C. Canham, and S. Pugh Intrinsic densitydependent regulation of vole populations. Nature 366: Overholtzer-McLeod, K. L Consequences of patch reef spacing for density-dependent mortality of coral-reef fishes. Ecology 87: Peters, T. M., and P. Barbosa Influence of population density on size, fecundity, and developmental rate of insects in culture. Annual Review of Entomology 22: Pitcher, T. J., A. E. Magurran, and I. J. Winfield Fish in larger shoals find food faster. Behavioral Ecology and Sociobiology 10: Pitcher, T. J., and J. K. Parrish Functions of shoaling behaviour in teleosts. Pages in T. J. Pitcher, editor. Behaviour of teleost fishes. Second edition. Chapman and Hall, London, UK. Ray, C., and A. Hastings Density dependence: are we searching at the wrong spatial scale? Journal of Animal Ecology 65: Roberts, G Why individual vigilance declines as group size increases. Animal Behaviour 51: Sale, P. F., and N. Tolimieri Density dependence at some time and place? Oecologia 124: Samhouri, J. F., R. R. Vance, G. E. Forrester, and M. A. Steele Musical chairs mortality functions: densitydependent deaths caused by competition for unguarded refuges. Oecologia 160: Sandin, S. A., and S. W. Pacala Fish aggregation results in inversely density-dependent predation on continuous coral reefs. Ecology 86: Sedinger, J. S., M. S. Lindberg, B. T. Person, M. W. Eichholz, M. P. Herzog, and P. L. Flint Density-dependent effects on growth, body size, and clutch size in Black Brant. The Auk 115: Shima, J. S Regulation of local populations of a coral reef fish via joint effects of density- and number-dependent mortality. Oecologia 126: Shima, J. S Mechanisms of density- and numberdependent population regulation of a coral-reef fish. Marine and Freshwater Research 53: Shima, J. S., and C. W. Osenberg Cryptic density dependence: effects of covariation between density and site quality in reef fish. Ecology 84: Skogland, T The effects of density-dependent resource limitations on the demography of wild reindeer. Journal of Animal Ecology 54: Steele, M. A., and G. E. Forrester Small-scale field experiments accurately scale up to predict density dependence in reef fish populations at large scales. Proceedings of the National Academy of Sciences USA 102: Stier, A. C., S. W. Geange, and B. M. Bolker Predator density and competition modify the benefits of group formation in a shoaling reef fish. Oikos 122: Storlazzi, C. D., M. Field, and M. H. Bothner The use (and misuse) of sediment traps in coral reef environments: theory, observations, and suggested protocols. Coral Reefs 30: SYSTAT Software SYSTAT version 13. SYSTAT Software, Chicago, Illinois, USA. Tupper, M., and R. Boutilier Effects of conspecific density on settlement, growth and post-settlement survival of a temperate reef fish. Journal of Experimental Marine Biology and Ecology 191: Turchin, P Population regulation: a synthetic view. Oikos 84:153. Vogel, J. L., and D. A. Beauchamp Effects of light, prey size, and turbidity on reaction distances of lake trout (Salvelinus namaycush) to salmonid prey. Canadian Journal of Fisheries and Aquatic Sciences 56: Webster, M Role of predators in the early postsettlement demography of coral-reef fishes. Oecologia 131: Webster, M. S Density dependence via intercohort competition in a coral-reef fish. Ecology 85: Wenger, A. S., J. L. Johansen, and G. P. Jones Increasing suspended sediment reduces foraging, growth and condition of a planktivorous damselfish. Journal of Experimental Marine Biology and Ecology 428: White, J. W Spatially correlated recruitment of a marine predator and its prey shapes the large-scale pattern of density-dependent prey mortality. Ecology Letters 10: White, J. W., J. F. Samhouri, A. C. Stier, C. L. Wormald, S. L. Hamilton, and S. A. Sandin Synthesizing mechanisms of density dependence in reef fishes: behavior, habitat configuration, and observational scale. Ecology 91: White, J. W., and R. R. Warner. 2007a. Behavioral and energetic costs of group membership in a coral reef fish. Oecologia 154: White, J. W., and R. R. Warner. 2007b. Safety in numbers and the spatial scaling of density-dependent mortality in a coral reef fish. Ecology 88: SUPPLEMENTAL MATERIAL Appendix Detailed description of field methods (Ecological Archives E A1).