Introduction

The impact of climate change on marine communities will depend on how species' differences in physiological tolerances and performance alter biological interactions within a changing resource landscape1,2,3. Rising atmospheric CO2 over the last century has already led to a reduction in ocean pH (ocean acidification) and an increase in ocean temperature4. Dissolved CO2 levels in the ocean are rising in line with atmospheric CO25 and are projected to exceed 900 μatm from the current day 390 μatm by 21006, while temperatures of the coastal ocean may increase by up to 3°C4. Coral reefs are particularly susceptible to these changes as they have evolved under relatively stable environmental conditions and many organisms may live near their maximum physiological tolerance limits for temperature and CO27,8. How reef communities are affected will be dependent on the ability of species to tolerate or adapt to the new conditions and how the altered characteristics of species influence the outcomes of key processes, such as competition and predation9.

Elevated levels of CO2 affect the acid-base balance of marine organisms10 and alters the processing of sensory information for organisms as diverse as snails, crabs and fishes11,12,13. For fishes this effect occurs through modification to the functioning of neuroreceptors14, which changes the way individuals interact with their habitats and with predators, leading to increased levels of mortality, particularly at critical life history bottlenecks such as settlement15,16. While research demonstrates major changes in behaviour and performance for isolated individuals, it is unknown how these conditions are likely to influence the biological interactions that form the foundation of community dynamics.

Ocean acidification and elevated water temperature also affect the live coral that is an important source of shelter and food for many coral reef fishes. High CO2 leads to a reduction in seawater pH that decreases the availability of carbonate ions required for coral skeleton growth17. In addition, ocean warming causes corals to lose the symbiotic zooxanthellae that produce much of their required food18. This coral bleaching has been responsible for wide scale loss of coral and combined with the effect of storms has resulted in widespread degradation of live coral and the fish communities that associate with them19,20,21. While shifts in fish community composition have been associated with coral degradation, it is unclear how degraded reef habitat will affect interactions among fishes that live in these environments in a high CO2 world.

Marine fishes display inter- and intra-specific variability in the extent to which they are affected by CO2-driven modifications to their environment. While many species require live bushy coral as a nursery habitat19,22, many have the flexibility to satisfy their resource requirements from a broader resource base and are not confined to one habitat (often classed as habitat generalists). Moreover, some fish species are more susceptible than others to the effects of elevated CO216,23 and even within species there can be high variability in the degree to which sensory systems are compromised13. This variability in the way species are affected modifies the selection profiles by predators on juvenile fishes under moderately elevated CO2 (~800–900 μatm CO2)24,25,26. One of the fundamental processes that affect the susceptibility of individuals to predation is how the prey within a multispecies assemblage interact to modify individual risk. Interference competition, is particularly important in affecting the small scale distribution of growth, body size, body condition and distribution in relation to predation risk27,28. While there are some studies demonstrating differences in sensitivity to CO2 among damselfish species, it is unclear how these differences may affect behavioural interactions under natural conditions of predation risk.

These effects will not occur in isolation, but rather as the coral habitats degrade fishes that inhabit these habitats will themselves be battling to adapt physiologically to the new resources and balance of senses available to them for risk assessment29. Our study examined how CO2 levels projected to occur in the next 70–100 y may affect the interactions between two damselfish species known to compete for space at settlement28 and how these interactions were modified in coral habitat degraded through bleaching. First, we reared wild-caught damselfish larvae of the two species (Pomacentrus amboinensis and P. moluccensis) in elevated CO2 (945 μatm) or control water for 4 days, which we have previously shown to be sufficiently long to yield the lasting behavioural effects associated with CO2-induced neuroreceptor modification14,15. Secondly, we transplanted treated fish to an array of natural reefs made of unbleached or bleached hard coral in a 2 × 2 design to test if exposure to elevated CO2 and coral state interacted to alter the behaviour of fishes and their risk of mortality during recruitment to adult habitat. Because CO2 impairs the senses used to detect and respond to habitat characteristics and assess predation15,30, we predicted that the effects of CO2 might interact with habitat state to affect behaviour and ultimately survival. Our results show that elevated CO2 reverses the competitive outcome between the two species with mortal consequences and that this reversal is accentuated in degraded habitats.

Results

There was a significant difference in mortality of P. moluccensis among the four treatments (X2 = 20.05, df = 3, p = 0.0002). Fish from elevated CO2 on unbleached coral had approximately four times the survival of fish from the other three treatments (Fig. 1a), indicating that there was an interaction between CO2 treatment and coral state. Mortality curves for P. amboinensis also differed from one another (X2 = 36.7728, df = 3, p < 0.0001), with fish that had been exposed to elevated CO2 having lower survival than those from current day CO2 conditions (Fig. 1b). Those exposed to elevated CO2 on bleached coral died within 48 h, while over 50% of fish from present day CO2 conditions survived 72 h (Fig. 1b).

Figure 1
figure 1

Influence of competition on survival.

Effects of interspecific competition on the proportional survival of (a) Pomacentrus moluccensis and (b) P. amboinensis on unbleached or bleached coral habitats when exposed to elevated or current day CO2 levels. Kapalin-Meier survival trajectories illustrate the different survival trajectories on patches of live unbleached or live-bleached Pocillopora damicornis hard coral. Individual fish were placed in size-matched pairs on the habitats (n = 25–29). Photographs: Mark McCormick.

Full size image

The maximum distance fishes ventured away from the coral shelter was affected by coral state and differed between species, but this species-specific effect was largely reversed with elevated CO2 (Species × State × CO2 treatment: F1,208 = 5.10, p = 0.025; Fig. 2a). On unbleached coral at ambient CO2, both species stayed a similar distance from the coral. However, on bleached coral, P. moluccensis was over twice as far from shelter than P. amboinensis. The trends between species were largely reversed by elevated CO2 conditions (Fig. 2b). On healthy coral P. moluccensis stayed closer to the coral, at a distance that did not differ from ambient CO2 condition. In contrast, P. amboinensis was located almost 4 times as far from unbleached coral than when exposed to ambient CO2 conditions and twice as far from the unbleached coral than P. moluccensis. These trends were accentuated when the patch reef comprised of bleached coral (Fig. 2b). The height of fish on patch reefs differed between species, but the magnitude of the effect changed with CO2 conditions experienced (Species × CO2 treatment: F1,208 = 12.31, p < 0.0006). P. amboinensis occupied the lower part of the patch regardless of CO2 treatment or coral state, while P. moluccensis occupied the highest part of the reef, but was significantly closer to the bottom when fish had been exposed to elevated CO2 (Tukey's test, p < 0.05; Fig. 3).

Figure 2
figure 2

Influence of exposure to elevated CO2 and habitat type on shelter use.

The mean maximum distance that fishes ventured from the shelter of unbleached or bleached hard coral within 3 min behavioural observations for Pomacentrus moluccensis (white) and P. amboinensis (grey) that had been reared under ambient or elevated CO2 conditions. Letters above the bars represent Tukey's HSD groups. Error bars are standard errors (n = 26–29).

Full size image
Figure 3
figure 3

Influence of exposure to elevated CO2 and fish species on space use.

Mean proportional height above the reef base within 3 min behavioural observations for Pomacentrus moluccensis (white) and P. amboinensis (grey) that had been reared under ambient or elevated CO2 conditions. Letters above the bars represent Tukey's HSD groups. Error bars are standard errors (n = 53–55).

Full size image

The levels of aggression exhibited by fish to their size-matched pair was strongly affected by coral state and differed between species, but this species-specific effect was reversed with elevated CO2 (Species × State × CO2 treatment: F1,208 = 11.757, p = 0.0007; Fig. 4). Under ambient CO2 conditions, P. amboinensis was much more aggressive than P. moluccensis with the dominant P. amboinensis being more aggressive on bleached coral (Fig. 4a). In contrast, under elevated CO2 conditions the pattern was reversed by species, with P. moluccensis being most aggressive on bleached coral compared to unbleached coral, but always more aggressive than P. amboinensis (Fig. 4b).

Figure 4
figure 4

Influence of exposure to elevated CO2 and habitat type on aggression.

Level of aggression displayed by juvenile Pomacentrus moluccensis (white) and P. amboinensis (grey) 45–60 min after release in size matched pairs onto a patch of habitat (healthy unbleached or bleached live Pocillopora damicornis). Fishes were either reared under ambient or elevated CO2 conditions. Aggression Index is calculated as Displays + (3× Bites) – Avoidances. Letters above the bars represent Tukey's HSD groups. Errors are standard errors. n = 26–29.

Full size image

Discussion

Elevated CO2 affected the outcome of competitive interactions between species, with a complete reversal in the species that was the most aggressive and dominant competitor. This reversal in the outcome of interference competition was most dramatic on degraded, bleached coral. The mechanism underlying this change was the differential effect of CO2 on the behaviour of the two species. A previous laboratory study has shown that P. moluccensis is less susceptible to exposure to elevated CO2 conditions than its congener P. amboinensis16, who is the competitive dominant under present day conditions28. For fish exposed to either current day or future CO2 conditions, the competitive interactions indirectly had mortal consequences because the aggressive interactions pushed the subordinate into areas of higher risk away from shelter; a pattern accentuated on degraded coral. The behavioural changes found in our study suggest that elevated CO2 may result in changes in the way species interact with each other and this will alter the processes that shape communities and ecosystems.

Understanding how altered species interactions affect communities and ecosystems is fundamental to predicting the impacts of climate change and ocean acidification1,31. Previous studies on benthic communities have observed enhanced competitive dominance of algae over corals under elevated CO23,9. In contrast, our results show a reversal of competitive dominance, indicating that ocean acidification may lead to far more complex outcomes than would be predicted with just enhanced competitive effects. Competitive reversals between species should be commonplace under natural circumstances32 for organisms with strong competitive asymmetry to coexist, however they are seldom conclusively demonstrated. Steneck et al. (1991) found that depth determined which of two species of crustose coralline algae was competitively dominant, but dominance was reversed when grazing limpets reduce fitness of one species33. Under present day CO2 conditions, McCormick and Weaver (2012) found that who won the contest between P. amboinensis and P. moluccensis was strongly asymmetrical and dependent on size, over-and-above other ecological factors such as prior residency28. The importance of size in dictating the outcome of competitive contests is a common finding for many organisms34. P. moluccensis juveniles only started to win contests when they were 1.2 mm larger than P. amboinensis. Because the winner of competitive interactions is based on body size, the stochasticity of recruitment at small spatial scales35 and its pulsed nature within a tropical recruitment season36 will mean that neither species will always win or lose. This may lead to a destabilizing influence on the competitive dominance of any one species, making competitive exclusion less likely.

It holds therefore that changes in the competitive outcome will also occur when the species change in their ability to defend a limited resource. Studies that have examined the effect of CO2 on the behaviour of solitary damselfish juveniles demonstrate that they display a willingness to take more risk, they fail to learn predators and as a result die faster than control fish15,24,37. However, the ability to cope with elevated CO2 varies among closely related species. Ferrari et al. (2011a) exposed 5 different species of damselfish from the genus Pomacentrus to damage-released chemical alarm cues from conspecifics16. Fish held at current day CO2 levels all responded with a fright response to the alarm cues, while those acclimated to higher CO2 levels exhibited different species-specific responses from no response to the alarm cues, through to a fright response that was little different from the controls. Further, in a mesocosm study, Ferrari et al. (2011b) found that the relative mortality through predation of 5 species of damselfishes changed when both predator and prey had been exposed to elevated CO224. However, the lack of behavioural information meant that the mechanism underlying the result was unclear. Our study shows that P. amboinensis exposed to elevated CO2 also becomes less aggressive and avoids similar sized competitors that they would normally have dominated. This lowering of aggression, which underlies interference competition, is critically important for the fish to maintain its status in the plankton feeding community and will have repercussions for such ecologically important activities as obtaining food of the highest possible quality and the acquisition and defence of spawning sites.

Species are likely to acclimate to some extent to the modified chemical conditions found in future CO2 conditions. However, previous studies have shown that the behavioural responses of juvenile damselfish are the same in individuals kept for 4 days at elevated CO2 and those reared from hatching at the same CO2 levels15,30. Therefore, there seems to be limited prospects for within-generation acclimation of behavioural changes caused by exposure to high CO2, at least during the critical early life history stages investigated here. In one of the only transgenerational studies to date, Miller et al. (2012) found that the routine metabolic rate, growth and survival of F1 offspring of the clownfish, Amphiprion melanopus, exposed to 1032 μatm returned to levels found in control fish as a result of non-genetic parental effects38. The generality of this result is unknown but likely to vary among species. Interestingly, a recent study has demonstrated that damselfish may have limited ability to compensate for CO2-induced behavioural changes through parental effects over one generation39. The ability of the species used in the current study to acclimate over multiple generations by non-genetic parental effects is currently unknown, however, it is likely that competing species who have been differentially affected by CO2 will be a realistic situation in the future.

CO2 appears to affect early life history traits such as respiration and growth in a different way to behaviour. Currently it is unknown for any species whether behaviour is likely to adapt over generations to elevated CO2. Recent studies of marine fishes and invertebrates have shown that at least part of the behavioural effect of elevated CO2 appears to occur through a change in the gradient of acid-base relevant ions across the neuron membrane40. This leads to a change in the operation of neuroreceptors (GABA-A receptors), switching them from an inhibitory to excitatory role. Research is required to determine whether this neural mechanism can adapt over multiple generations before the full ramifications of the present study can be determined.

Behavioural effects of high CO2 will not occur in isolation; ocean warming that affects the health of corals that fish use as shelter20 also has direct effects on the performance of interacting fish species. A number of recent studies have found that juvenile fish that use live coral as a nursery habitat move further away from the shelter when the coral is bleached and this exposes them to higher predation pressure in the laboratory41 and field28,42,43. The movement away from bleached corals was also found in the current study, with the subordinate species being furthest away from shelter. This was in part driven by an increase in the aggression by the dominant species, the identity of which changed in response to elevated CO2. Ocean warming may be expected to accentuate this effect by elevating routine metabolic rates40, increasing activity44,45 and the daily food requirements46, though some of this effect may be reduced through developmental and parental acclimation47,48.

Our results represent a first step towards an understanding of how community processes may change in response to elevated CO2. Future studies will also need to incorporate CO2 effects on predators, who themselves could be variously affected49. As demonstrated in the present study, competition can affect survival by altering an individual's exposure to predators, either directly through aggression pushing subordinates into riskier habitats, or indirectly by elevating hunger in subordinates and in so doing influencing their balance between foraging and vigilance50. Because both intra- and inter-specific competition is critically important in influencing prey vulnerability to predators investigation of the three-way interaction between dominant, subordinate and predator under elevated CO2 will be the next crucial step in understanding the regulation of communities in a high CO2 ocean.

Methods

Newly settled ambon damselfish, Pomacentrus amboinensis (Pomacentridae) and lemon damselfish, P. moluccensis compete for shelter in the same live coral living space when they first settle from the plankton51. Both are planktivores and show similar preferences for live, healthy coral at settlement28. Interaction with the other species at the reef edge reduces the growth of both species compared to locations where the other species is absent28. Within 2 mo on shallow reefs the two species display a disjunct distribution patterns with P. amboinensis associated with the base of the reef and P. moluccensis with live coral at the top of the reef28.

Settlement-stage larvae of P. amboinensis and P. moluccensis were collected overnight using light traps (for design see Meekan et al. 200152) moored in open water around Lizard Island (14′40° S, 145′28° E), in the northern Great Barrier Reef, Australia. Fish were sorted to species and transferred to 35 L aquaria supplied with a continuous flow of either control (present day CO2; 410 μatm) or elevated-CO2 seawater (945 μatm; see Supplementary file) for 4 days. This period of time has been found to be sufficient to elicit the full extent of behavioural effects of high CO215. Fish were fed 4 times daily ad libitum with newly hatched Artemia sp.

On the day of the field trial, fishes from the control and elevated-CO2treatments were placed into individually labelled plastic bags of seawater and measured with digital callipers. For each CO2 treatment, interspecific pairs of P. amboinensis and P. moluccensis were created such that the size difference between fishes was no more than 0.3 mm SL. Fish length was standardised as it has been found to be important in determining the outcome of interaction between newly settled fishes and their survival53,54. There was no significant difference in size of the fish between species used in the experiment (mean ± se: P. amboinensis 11.4 ± 0.03; P. moluccensis 11.5 ± 0.03; F1,214 = 3.3, p = 0.07). To reduce transport and handling stress, fish in bags were transported to the field site in a 60 L holding tank of seawater (to reduce temperature fluctuations) under subdued light conditions.

Patch reefs used in the field experiment were composed of one of two states of the bushy hard coral, Pocillopora damicornis: live unbleached coral or bleached coral. Bleached coral was produced by collecting healthy coral and placing it in freshwater for 4 minutes55. Over the next 48 h stressed coral lost their zooxanthellae and became pale in colouration but still had their tissue intact. An array of patch reefs (~18 × 15 × 18 cm) was constructed on a sandflat, arranged 4–5 m apart and 3–6 m away from continuous reef.

Both damselfish species naturally settle on patch reefs near continuous reef. In this habitat juveniles are exposed to a diverse range of predators that use a variety of feeding modes from ambush (lizardfish Synodus dermatogenys and small cods Cephalopholis microprion) to pursuit (dottybacks Pseudochromis fuscus and wrasse Thalassoma lunare). These fishes can be observed preying on juveniles that venture too far from shelter.

Experimental protocol

Length-matched pairs of P. amboinensis and P. moluccensis were placed onto unbleached or bleached coral patches. Patches were cleared of any other fishes or invertebrates using a hand net prior to release. Experimental fishes were released on natural reefs away from the study area. A small wire cage (~30 × 30 × 30 cm, 12 mm mesh size) was placed over the patch to allow fish to acclimatise to their new surroundings while being protected from predators. Cages were removed 40 to 60 min after release of the fish between 10:00 and 12:00 h. Fish presence was monitored 2 to 3 times per day (i.e. after the initial acclimatisation period, the evening after release and the following morning) for ~72 h. The field experiment and monitoring of survival were conducted over a continuous 11 d period during a recruitment pulse in late October 2012. Water temperature during the census periods averaged 25.5°C. Our previous studies with tagged fish have shown that settlement stage damselfish do not migrate from these isolated reefs and that any loss is most parsimoniously attributed to predation (e.g.42,44). This research was conducted under animal ethics approval from JCU (Approval A1737) and a research permit from the Great Barrier Reef Marine Park Authority (G12-34811.1).

Behavioural assessment

The behaviour of each fish placed on the patch reefs was monitored 45 to 60 min after placement on the reef following the protocol of McCormick (2009)42. Briefly, behaviour of the focal fish was assessed over a 3 min period by a scuba diver positioned 1.0 m away from the patch with the aid of a magnifying glass. Six aspects of activity and behaviour were assessed: a) maximum distance ventured from the habitat patch; b) height above substratum (categorized as % of the time spent within the bottom, middle or upper third of the patch); c) number of fin displays; d) the number of chases or bites; e) number of avoidance episodes in response to a conspecific approach. Two additional composite variables were created to summarize several of the variables recorded and enhance ecological interpretation. Relative height on the patch was summarized as a cumulative proportion of the time spent at varying heights over the 3 min observation period, with the top of the patch taken as height of 1, mid-patch a height of 0.5 and bottom a height of 0. For instance, when a fish spends 50% of its time at the top of the patch, 20% at the middle and 30% at the base the relative height value would be: (1 × 0.5) + (0.5 × 0.2) + (0 × 0.3) = 0.6. An aggression index was also created by adding the number of displays to the product of three times the number of chases/bites and then subtracting the number of avoidance events28,42. Our other studies have shown that a 3 min observation period is sufficient to quantify behaviour because it is consistent in the short term (10 min to 5 days44). The variable measured had been previously shown to be sensitive to competitive interactions between these two fish species28,43.

Statistical analyses

Survival (up to 72 h) of fish among the CO2 treatment (ambient and elevated) by habitat (unbleached, bleached) combinations were compared using multiple-sample Survival Analysis, which uses a Cox's proportional hazard model (Statistica 12.056). This regression model is not based on any assumptions concerning the nature or shape of the underlying survival distribution. Survival curves of each CO2 treatment and coral state were calculated and plotted using the Kaplan–Meier product–limit method. The Kaplan–Meier method is a non-parametric estimator of survival that incorporates incomplete (censored) observations, such as those cases where censuses had to be terminated on trials prior to their completion due to time limitations of a field trip.

Three behavioural variables, maximum distance ventured, relative height and the aggression index, were compared between species, coral habitat and CO2 treatment combinations with three-factor ANOVA (Type III SS) followed by Tukey's HSD tests. Residual analysis was used to examine assumptions of ANOVA.