Effects of ocean acidification on visual risk assessment in coral reef fishes


  • Maud C. O. Ferrari,

    Corresponding author
    1. Department of Biomedical Sciences, WCVM, University of Saskatchewan, Saskatoon, Saskatchewan S7N 5B4, Canada
      Correspondence author. E-mail: maud.ferrari@usask.ca
    Search for more papers by this author
  • Mark I. McCormick,

    1. ARC Centre of Excellence for Coral Reef Studies, and School of Marine and Tropical Biology, James Cook University, Townsville, Queensland 4811, Australia
    Search for more papers by this author
  • Philip L. Munday,

    1. ARC Centre of Excellence for Coral Reef Studies, and School of Marine and Tropical Biology, James Cook University, Townsville, Queensland 4811, Australia
    Search for more papers by this author
  • Mark G. Meekan,

    1. Australian Institute of Marine Science, UWA Ocean Sciences Centre (MO96), 35 Stirling Highway, Crawley, Western Australia 6009, Australia
    Search for more papers by this author
  • Danielle L. Dixson,

    1. ARC Centre of Excellence for Coral Reef Studies, and School of Marine and Tropical Biology, James Cook University, Townsville, Queensland 4811, Australia
    Search for more papers by this author
  • Oona Lönnstedt,

    1. ARC Centre of Excellence for Coral Reef Studies, and School of Marine and Tropical Biology, James Cook University, Townsville, Queensland 4811, Australia
    Search for more papers by this author
  • Douglas P. Chivers

    1. Department of Biology, University of Saskatchewan, Saskatoon, Saskatchewan S7N 5E2, Canada
    Search for more papers by this author

Correspondence author. E-mail: maud.ferrari@usask.ca


1. With the global increase in CO2 emissions, there is a pressing need for studies aimed at understanding the effects of ocean acidification on marine ecosystems. Several studies have reported that exposure to CO2 impairs chemosensory responses of juvenile coral reef fishes to predators. Moreover, one recent study pointed to impaired responses of reef fish to auditory cues that indicate risky locations. These studies suggest that altered behaviour following exposure to elevated CO2 is caused by a systemic effect at the neural level.

2. The goal of our experiment was to test whether juvenile damselfish Pomacentrus amboinensis exposed to different levels of CO2 would respond differently to a potential threat, the sight of a large novel coral reef fish, a spiny chromis, Acanthochromis polyancanthus, placed in a watertight bag.

3. Juvenile damselfish exposed to 440 (current day control), 550 or 700 μatm CO2 did not differ in their response to the chromis. However, fish exposed to 850 μatm showed reduced antipredator responses; they failed to show the same reduction in foraging, activity and area use in response to the chromis. Moreover, they moved closer to the chromis and lacked any bobbing behaviour typically displayed by juvenile damselfishes in threatening situations.

4. Our results are the first to suggest that response to visual cues of risk may be impaired by CO2 and provide strong evidence that the multi-sensory effects of CO2 may stem from systematic effects at the neural level.


The rapid increase in the concentration of atmospheric greenhouse gases during the past century, and the concomitant change in climate patterns, has led the scientific community to invest considerable effort towards understanding their effects on terrestrial and aquatic ecosystems (Orr et al. 2005; The Royal Society 2005; Fabry et al. 2008; Doney et al. 2009; Kerr 2010). The atmosphere and shallow oceans are in approximate gas equilibrium. Therefore, as CO2 concentrations increase in the atmosphere, more CO2 is absorbed by the ocean. Additional CO2 reacts with water to generate carbonic acid and hydrogen ions, which increases the acidity of the water. Increasing hydrogen ions bond with carbonate ions to form more bicarbonate, leading to a reduction in carbonate ion concentration (Orr et al. 2005; Fabry et al. 2008). The current emissions trajectory indicates that atmospheric CO2 will exceed 500 μatm by mid-century and could reach 850 μatm by the end of this century (Meehl et al. 2007; Raupach et al. 2007). The biological consequences of this reduction in carbonate ion availability have led most of the research on ‘ocean acidification’ to focus on calcifying organisms such as corals, echinoderms and crustaceans (Kleypas et al. 2006; Hofmann et al. 2010). Much less is known about how changes in ocean chemistry may affect other species such as fishes (Ishimatsu, Hayashi & Kikkawa 2008). A recent meta-analysis (Kroeker et al. 2010) shows that only 25% of the 198 tests reporting ocean acidification effects were performed on non-calcifiers, with only 2% of the studies being conducted on fishes. Despite the apparent similarities between CO2-induced ocean acidification and acid-rain-induced freshwater acidification, these two phenomena are markedly different. The ocean acidification issues are believed to stem from the carbonate imbalance, rather than from a pH decrease per se. For this reason, we will abstain from bringing freshwater literature into this study.

Most previous research on the effects of elevated CO2 on fish has been conducted at CO2 levels much higher than predicted to occur over the next 50–100 years because of anthropogenic greenhouse gas emissions and has focused on detecting physiological tolerance limits (Ishimatsu, Hayashi & Kikkawa 2008). However, a few recent studies have examined the effects of environmentally relevant sublethal concentrations of CO2 on the ecology of coral reef fishes. Munday et al. (2009) first demonstrated that upon exposure to c. 1000 p.p.m. CO2, clownfish Amphiprion percula, tested in a two-channel flume chamber, displayed impaired ability to distinguish between the odours of different habitats, becoming attracted to the chemical cues they normally avoided. Moreover, they also became unable to distinguish between the odour of kin and non-kin. This olfactory impairment was also demonstrated in a predation context, whereby clownfish exposed to c. 1000 p.p.m. CO2 preferred swimming on the side of the flume that contained predator odour, while control clownfish preferred staying on the seawater control side, away from the predator odour (Dixson, Munday & Jones 2010). Both Munday et al. (2010) and Ferrari et al. (2011a) showed similar olfactory impairments in juvenile damselfish and demonstrated a dramatic increase in mortality rate under natural conditions for fish exposed to 850 p.p.m. CO2. Chemosensory tests performed at different CO2 levels indicate that impairment occurs at c. 700 p.p.m. (Dixson, Munday & Jones 2010). Until recently, the behavioural effects of CO2 have been associated with chemosensory impairment. However, Simpson et al. (2011) just documented that damselfish exposed to c. 600 p.p.m. CO2 failed to avoid reef noise during the day in the same manner as control fish not exposed to CO2. It is unknown whether ocean acidification may impact other senses used in risk assessment.

Fishes use a variety of information sources to avoid being captured by predators. In coral reefs, fishes are known to rely on auditory and chemosensory cues when making decisions on where to settle (Simpson et al. 2005; Vail & McCormick 2011). However, once settled on the reef, the fishes rely heavily on visual and chemical sources of information. Chemical cues, both predator odours and alarm cues from conspecifics, are particularly important at night and in highly structured habitats, but visual cues have the advantage in that they provide the most accurate information about risk in both space and time (Ferrari, Wisenden & Chivers 2010).

The goal of our experiment was to test whether CO2 effects would manifest themselves in predation contexts that provided primarily visual information for risk assessment. Our experiment purposefully excluded chemosensory information that could be used by the fish to assess risk. If CO2 effects on risk assessment are manifest in multiple sensory systems, then the problem likely does not rest with impaired sensory perception but rather with a more generalized alteration in cognitive processing. Here, we exposed juvenile damselfish Pomacentrus amboinensis to a range of CO2 treatments (440 μatm – present-day level control, 550, 700 and 850 μatm CO2), following methodologies similar to Ferrari et al. (2011a), and exposed them to the sight of an adult spiny chromis Acanthochromis polyacanthus in a sealed plastic bag. The chromis is a larger coral reef fish to which the juvenile damselfish are naive and hence may represent a significant predation threat to the damselfish.

Materials and methods

Fish Collection and CO2 Treatment

The experiment was conducted at the Lizard Island research station (14°40′S, 145°28′E), on the Great Barrier Reef, Australia, in November 2010. Juvenile P. amboinensis (16–21 days old) were caught overnight using light traps (Meekan et al. 2001) moored c. 100 m off the fringing reef. These traps collect fish at the end of their pelagic phase, immediately prior to their settlement to the reef (Meekan, Milicich & Doherty 1993). Fishes caught in the traps were brought back to the station just after dawn and sorted by species, and small groups of P. amboinensis were transferred into 35-L aquaria maintained at one of four CO2 concentrations. Previous experiments have demonstrated that the behavioural effects of elevated CO2 are evident within 4 days of exposure to relevant CO2 treatments and that longer exposures result in identical behavioural impairments compared to larvae raised under the same CO2 levels from birth (Munday et al. 2010). Therefore, larvae were maintained in the CO2 treatments for four consecutive days, and experiments were completed within 48 h of the final day of exposure to CO2 treatments. The fish were fed freshly hatched Artemia nauplii three times a day.

CO2 treatments were maintained by CO2 dosing to a set pHNBS following standard techniques for ocean acidification research, as set out in the Best Practices Guides for Ocean Acidification Research (Gattuso et al. 2010). Seawater was pumped from the ocean into 4 × 60 L sumps, where it was diffused with ambient air (control) or CO2 to achieve a pH of c. 8·15 (control), 8·06, 7·97 or 7·89. The reduced pH values were selected to achieve the approximate CO2 conditions required, based on preliminary observations of total alkalinity, salinity and temperature of seawater at Lizard Island. A pH controller (Tunze Aquarientechnik, Penzberg, Germany) was attached to each of the CO2-treated sumps to maintain pH at the desired level. A solenoid injected a slow stream of CO2 into a powerhead at the bottom of the sump whenever the pH of the seawater rose above the set point. The powerhead rapidly dissolved CO2 into the seawater and also served as a vigorous stirrer. Equilibrated seawater from each sump was supplied at a rate of c. 500 mL s−1 to four replicate 35-L aquariums, each housing a group of larval fishes. To maintain oxygen levels and the required pCO2 levels, aquariums were individually aerated with air (control c. 390 p.p.m.) or CO2-enriched air (c. 550, 700 or 850 μatm). The concentration of CO2-enriched air was controlled by a scientific-grade pressure regulator and precision needle valve and measured continuously with an infrared CO2 probe (Vaisala GM70, Vaisala, Helsinki, Finland). Temperature and pHNBS of each aquarium were measured each morning and afternoon, using an HQ40d pH meter (Hach, Loveland, CO, USA) calibrated with fresh buffers. Total alkalinity of seawater was estimated by Gran titration from water samples taken twice weekly from each CO2 treatment. Alkalinity standardisations performed before processing each batch achieved accuracy within 1% of certified reference material from Dr A. Dickson (Scripps Oceanographic Institute). Average seawater pCO2 was calculated using these parameters in the program CO2SYS and using the constants of Mehrbach et al. (1973) refit by Dickson & Millero (1987). Estimated seawater parameters are shown in Table 1.

Table 1.   Mean (±SD) seawater parameters in the experimental system. Temperature, pH salinity and total alkalinity (TA) were measured directly. pCO2 was estimated from these parameters using CO2SYS
pHNBSTemp °CSalinity pptTA (μmol kg−1 SW)pCO2
8·15 (0·04)27·66 (0·98)352269·66 (15·01)440·53 (44·46)
8·06 (0·05)27·37 (0·93)352265·04 (27·00)554·04 (81·69)
7·97 (0·06)27·59 (0·97)352259·87 (11·55)718·37 (110·82)
7·89 (0·06)27·74 (0·99)352261·23 (14·92)879·95 (140·64)


Following CO2 conditioning, fish were transferred individually into 20-L clear plastic flow-through tanks (32 × 16 × 16 cm), containing a sand substrate, a coral object (shelter) and an airstone, placed on the opposite side of the coral, to which was attached a piece of tubing used as an injection hose to introduce food. Having the hose attached to the air stone allows a rapid diffusion of the Artemia throughout the tank. Damselfish juveniles treated with elevated CO2 retain their impaired behavioural responses for at least 48 h after being transferred back into control water (Munday et al. 2010); hence, the tanks were filled with freshly pumped ocean water (mean temperature: 28 °C). Each tank was covered on three sides with black plastic to ensure visual isolation from neighbouring tanks. A 4 × 4 cm grid was drawn in the front of the tank to allow the observer to record behavioural parameters. The fish were fed ad libitum and left undisturbed over night. Food was provided again in the morning, 1 h prior to testing.

Behavioural observations followed established protocols (Ferrari, Wisenden & Chivers 2010; Holmes & McCormick 2010b) and consisted of a 4-min pre-stimulus presentation period followed by a 4-min post-stimulus presentation period. The two observation periods were separated by a 45-sec stimulus introduction period, during which an adult A. polyacanthus (12·4 ± 1·1 cm SD fork length), placed in a watertight clear plastic bag (20 × 10 cm) containing oxygenated water, was gently introduced at the end of the tank on the opposite side of the coral object. The bag was oriented such that the side of A. polyancanthus was facing the fish (the long side of the bag was parallel to the short side of the tank). The bag also contained a thin layer of gravel to ensure it would settle on the bottom of the tank.

To stimulate activity, we introduced small quantities of Artemia into the tank on the opposite side of the coral, via the injection hose. Turbulence created by the airstone allowed the food to spread throughout the tank within a few seconds. The fish were fed 2·5 mL of food (solution containing c. 250 freshly hatched Artemia per mL) 5 min prior to the start of the trial and immediately prior to the two observation phases. During each observation period, we collected data on: (i) foraging, measured as the total number of feeding strikes in the 4-min period, regardless of success; (ii) activity level, measured as the total number of lines crossed; a line was crossed when the entire body of the fish crossed the line; and (iii) area use, measured as the total number of 4 × 4 cm squares visited. During the post-stimulus presentation only, we also measured (iv) minimum approach distance, as the smallest distance (in cm, assessed using the grid) between the fish and the bag containing the A. polyacanthus; and (v) the occurrence of bobbing behaviour (fish exposed to threats will often display rapid, small-amplitude vertical movements). Decreases in foraging, activity and area use and an increase in frequency of bobbing are common antipredator behaviours displayed by animals (Ferrari, Wisenden & Chivers 2010). We tested 22–25 fish in each treatment (mean ± SD standard length: 1·4 ± 0·1 cm). The same number of fish from each treatment was tested on a given day, the order of testing was randomized, and the observer was blind with respect to the CO2 treatment groups.

Statistical Analyses

Pre- and post-stimulus data for foraging, activity and area use were computed into a per cent change from the pre-stimulus baseline ((post-pre)/pre). Owing to the interdependency of the three behaviours, we analysed the three variables together using a one-way manova. Subsequent Tukey post hoc tests were performed to assess the differences in behavioural responses between the different CO2 levels. Foraging data did not follow parametric assumptions; hence, we used rank-transformed foraging data in the analyses. The data for minimum distance to the chromis were analysed using a one-way anova, followed by post hoc Tukey tests. The occurrence of bobbing behaviour was analysed using a chi-square test, followed by post hoc comparisons.


There was no effect of CO2 treatment on the behaviour of juvenile P. amboinensis measured during the pre-stimulus period (manova: Pillai’s trace: F9,276 = 0·84, P = 0·58). However, there was a statistically significant effect of CO2 concentration on the responses of fish to the presentation of the chromis (Pillai’s trace: F9,276 = 3·31, P = 0·001, Fig. 1). Post hoc tests revealed the same patterns for the three behavioural measures: fish exposed to 440, 550 or 700 μatm CO2 did not differ in their response to the chromis. However, fish exposed to 850 μatm CO2 showed a weaker antipredator response. That is, they displayed higher foraging rate, higher activity levels and greater area use than fish exposed to lower CO2 concentrations.

Figure 1.

 Mean proportion change (±SE) in feeding strikes (top left panel), area use (top right panel), line crosses (bottom left panel) and mean (±SE) minimum approach distance (bottom right panel) for juvenile damselfish treated with different CO2 concentration and exposed to the sight of a spiny chromis. Different letters refer to statistical differences at a 0·05 α level.

The one-way anova also revealed a statistically significant effect of CO2 on the minimum approach distance (F3,92 = 5·31, P = 0·002, Fig. 1). Post hoc tests revealed that fish exposed to increasing concentrations of CO2 decreased their minimum approach distance.

The chi-square test performed on the 2 × 4 contingency table revealed a significant interaction between CO2 concentration and the occurrence of bobbing (inline image = 9·7, P = 0·021; bobbing frequencies of 63%, 62·5%, 56% and 0% for 440, 550, 700 and 850 μatm CO2 fish, respectively). Post hoc comparisons revealed no difference in the occurrence of bobbing between fish exposed to 440 and 550 (two-tailed Fisher’s exact test, P > 0·99), or 550 and 700 μatm CO2 (Fisher’s Exact Test, P = 0·77). However, fish exposed to 850 μatm CO2 showed a significant decrease in the occurrence of this behaviour as compared to fish exposed to 700 μatm (Bonferroni-corrected α: 0·05/3 = 0·017; Fisher’s exact test, P = 0·003, Fig. 1).


Our results clearly show differential threat responses of fish treated with different concentrations of CO2. Juvenile fish exposed to 440, 550 and 700 μatm CO2 did not differ in their behavioural response to the adult chromis. They all showed strong decreases in foraging, activity and area use, typical responses observed by individuals responding to a potentially threatening situation (Ferrari, Wisenden & Chivers 2010). However, fish treated at 850 μatm CO2 showed a decrease in the intensity of response compared to fish with lower CO2 exposure; they did not reduce feeding, activity and area use as much, indicating that they perceived the chromis as less threatening than other conspecifics. All fish showed an increased vigilance upon introduction of the chromis bag in the tank, which likely provided mechanical and visual disturbances. However, the fish exposed to 850 μatm seem to overcome this disturbance and very quickly ignored or became attracted to it. In addition, the 850 μatm juveniles did not display bobbing behaviour as typical observed in risky situations but rather approached the chromis.

The decrease in response reported here supports previous results showing a decrease in antipredator response by CO2-treated fish. However, because of the absence of chemosensory cues, we can infer that effect documented in the present study is not mediated by an alteration of the olfactory system. We suspect that vision was the main sense used by the juvenile fish to assess risk, although we cannot completely rule out other senses such as auditory cues, given that we do not have information about whether A. polyacanthus produce sounds that could be used in risk assessment by the damselfish. The occurrence of bobbing behaviour is typically a visually mediated behaviour in our test species. Juvenile damselfishes recruiting to the reef have no experience with the diversity of predators that they will encounter on the reef; hence, the recruits often need to learn which large fish pose a threat to them (Mitchell et al. 2011). Spiny chromis are not typically considered predators of juvenile damselfish. However, we chose the spiny chromis because of its docile behaviour during stimulus presentation. Other piscivores such as dottybacks, wrasses or lizardfishes, which are common predators of recruit fishes (Holmes & McCormick 2010a), were not as compliant in the holding bag, striking the bag and trying repeatedly to get out, hence confounding the source of information available to the damselfish. Spiny chromis were rarely in contact with the bag.

Another interesting result is the threshold at which CO2 effects became apparent. In previous studies of chemosensory assessment, marked behavioural effects appeared after exposure to 700 μatm CO2 (Dixson, Munday & Jones 2010; Munday et al. 2010). In the one study on auditory responses, Simpson et al. (2011) documented significant effects at 600 μatm CO2. However, in the present study, fish exposed to 850 μatm CO2 were affected, while those exposed to lower concentrations were not. Moreover, the pattern of negative effects was consistent for all behavioural measures. Given that we are using the same CO2 system and the same test species in the same location as other studies reporting those effects (Munday et al. 2010; Ferrari et al. 2011a), it is reasonable to compare the CO2 thresholds. The relatively high threshold effect found in the present study may indicate that CO2 affects the visual systems independently of the chemosensory and auditory systems, i.e. via different pathways. Alternatively, the same physiological response may be responsible for changes in the chemosensory, auditory and visual responses, but the visual pathway is less sensitive to changes in CO2 concentrations.

The ability of marine species to persist through the challenges presented by elevated dissolved CO2 will depend on the adaptability of those species to their novel environmental conditions and the speed at which the environment will be changing. Regardless of the exact reason for the observed effects of dissolved CO2, different sensory inputs may be sensitive to different thresholds of CO2, and this may prove important for the survival of those species. If one sensory system is impaired at a given level of CO2, could another sense compensate for this loss? If so, this may slow down the functional effects of CO2 on individuals. Munday et al. (2010) and Ferrari et al. (2011a) both showed a graded effect of CO2 on the mortality of juvenile damselfishes, with an increase in mortality occurring at 700 p.p.m. (c. 700 μatm) and heavier mortality occurring at 850 p.p.m. (c. 850 μatm) compared to control fish. This implies that the loss of chemosensory assessment of risk at 700 p.p.m. CO2 and the impaired responses to auditory responses at 600 p.p.m. cannot be fully compensated by normal responses to visual stimuli at that same level.

Most coral reef fishes have a bipartite life history made of an initial pelagic stage whereby larvae reside in the plankton for a period of weeks to months (Leis 2007), followed by a benthic phase, for which juvenile fish must locate suitable habitat and in doing so face a new and abundant array of predatory reef fishes. It is during this transitioning period that predation pressure is intense; predators may remove at least 60% of newly settling fish in a single night (Almany & Webster 2006), creating population bottlenecks. Although these newly settled fish have juvenile form and coloration, they are largely naïve to the suite of predators that await them on the reef. Their ability to detect predators visually, chemically and mechanically is hence crucial for their survival, and our results suggest that CO2 has the ability to dramatically alter the dynamics of predator–prey interactions taking place in coral reefs. Indeed, Ferrari et al. (2011b) recently conducted a mesocosm experiment showing that the pattern of predator selectivity for different species of damselfishes was influenced by exposure to elevated CO2. An increase in the consumption of juvenile fishes by predators during their transition to benthic life will impact the replenishment of reefs and could have far-reaching implications on the biodiversity of coral reefs (2011b).

Our experiment was short term and ignored any potential for adaptation to dissolved CO2. While laboratory experiments showed that maintaining juveniles in elevated CO2 from birth does not further the effects of CO2 (Munday et al. 2010), it is possible that species may adapt through selection over several generations. Longer-term experiments will be needed to answer questions about adaptation and selection, even if laboratory conditions cannot fully replicate the selective environment exerted by predators on different CO2-sensitive phenotypes in the wild. We currently do not know the exact cause of behavioural alterations mediated by CO2; however, the accumulating experimental evidence shows that altered behaviour following exposure to elevated CO2 is caused by a systemic effect at the neural level. Behavioural lateralization (tendency of individuals to turn left or right) is a non-sensory-related behaviour and an expression of brain functional asymmetries. Recent evidence suggests that such lateralization in coral reef fishes is also affected by elevated exposures to CO2 (Domenici et al. 2011), which strengthens the idea that elevated CO2 affects brain function in larval fishes. It seems likely that such effects are related to ionic changes associated with acid–base regulation, but additional studies are required to pinpoint the exact mechanisms involved (Munday et al. 2010; Simpson et al. 2011).


Funding to MF and DC from the Natural Sciences and Engineering Council of Canada, to MF from the Yulgilbar Foundation, to MIM, PM, MF and DC from the Australian Research Council and the ARC Centre of Excellence for Coral Reef Studies and to MGM from the Australian Institute of Marine Science is acknowledged. All work reported here followed animal ethics guidelines at James Cook University.