Review paper 9 pages

L E T T E R
Ocean acidification reduces coral recruitment by disrupting

intimate larval-algal settlement interactions

Christopher Doropoulos,1,2*

Selina Ward,1 Guillermo

Diaz-Pulido,2,3 Ove

Hoegh-Guldberg2,4 and

Peter J. Mumby1,2

Abstract
Successful recruitment in shallow reef ecosystems often involves specific cues that connect planktonic

invertebrate larvae with particular crustose coralline algae (CCA) during settlement. While ocean acidification

(OA) can reduce larval settlement and the abundance of CCA, the impact of OA on the interactions between

planktonic larvae and their preferred settlement substrate are unknown. Here, we demonstrate that CO2
concentrations (800 and 1300 latm) predicted to occur by the end of this century significantly reduce coral
(Acropora millepora) settlement and CCA cover by ‡ 45%. The CCA important for inducing coral settlement
(Titanoderma spp., Hydrolithon spp.) were the most deleteriously affected by OA. Surprisingly, the only preferred

settlement substrate (Titanoderma) in the experimental controls was avoided by coral larvae as pCO2 increased,

and other substrata selected. Our results suggest OA may reduce coral population recovery by reducing coral

settlement rates, disrupting larval settlement behaviour, and reducing the availability of the most desirable

coralline algal species for successful coral recruitment.

Keywords
Acropora, coral, crustose coralline algae, electivity, Hydrolithon, ocean acidification, recruitment, settlement,

Titanoderma.

Ecology Letters (2012) 15: 338–346

INTRODUCTION

The effects of ocean acidification (OA) have raised concerns about

coral reef ecosystem function by reducing the calcification rates of

benthic organisms important to maintaining habitat structure and

biodiversity (Hoegh-Guldberg et al. 2007; Kroeker et al. 2010).

Anthropogenic emissions of carbon dioxide (CO2) have increased

atmospheric CO2 from approximately 280 ppm prior to the year 1750

to > 380 ppm in 2005 (Jansen et al. 2007), and these are continuing to

rise (Le Quere et al. 2009). The absorption of this atmospheric CO2 by

the oceans has reduced global pH by 0.1 units and carbonate

saturation state by 20% since 1800 (Orr et al. 2005). Numerous

laboratory studies have demonstrated that corals (Schneider & Erez

2006; Anthony et al. 2008), calcifying algae (Anthony et al. 2008;

Kuffner et al. 2008), and coral reef communities (Langdon et al. 2000;

Andersson et al. 2009) have reduced calcification in seawater with

lower pH due to depleted carbonate saturation.

Ecological processes pivotal to coral reef resilience, including coral

recruitment, herbivory, trophic integrity, and connectivity (Knowlton

2001; Mumby et al. 2007), under high CO2 levels have hardly been

investigated (Doney et al. 2009). Yet, growing evidence suggests that

interactions between species are altered as CO2 increases. Under

conditions of OA, corals in contact with fleshy macroalgae had higher

mortality (Diaz-Pulido et al. 2011), and fish mortality increased as OA

reduced the ability of juvenile fish to detect their predators (Munday

et al. 2010). Furthermore, it has been suggested that turf algae can

decrease the recruitment of crustose coralline algae (CCA) (Kuffner

et al. 2008; Russell et al. 2009) and kelp (Connell & Russell 2010)

because of greater space occupation at elevated pCO2. While these

examples illustrate that ecological interactions can be altered as CO2
increases, potential interactions of OA on coral recruitment have not

been addressed.

Recruitment is critical to community recovery as it represents a

crucial process in the development of populations in the post-

disturbance period. A key ecological process in the formation of

coral reefs is the settlement of coral larvae from the plankton to

the reef substrata. Many larvae test benthic substrates for

microhabitat suitability prior to settlement (i.e. attachment and

metamorphosis), with the selection of optimal microhabitats critical

in the post-settlement survival of benthic invertebrates (Raimondi

& Keough 1990; Harrington et al. 2004). Different benthic algae

offer both inductive and inhibitive settlement cues for planktonic

invertebrate larvae (Rodriguez et al. 1993; Kuffner et al. 2006; Diaz-

Pulido et al. 2010), and larvae often search for appropriate substrata

associated with specific CCA and microbial communities for

successful settlement (Morse et al. 1988; Johnson & Sutton 1994;

Heyward & Negri 1999; Negri et al. 2001; Webster et al. 2004).

While recent evidence demonstrates the settlement of coral larvae is

reduced as pCO2 increases (Albright et al. 2010; Albright &

Langdon 2011; Nakamura et al. 2011), the interactions between

planktonic larvae and the CCA community under elevated CO2
levels are unknown.

1School of Biological Sciences, University of Queensland, St Lucia, Qld 4072,

Australia
2Australian Research Council Centre of Excellence for Coral Reef Studies,

University of Queensland, St Lucia, Qld 4072, Australia

3Griffith School of Environment and Australian Rivers Institute, Nathan Campus,

Griffith University, Nathan, QLD 4111, Australia
4Global Change Institute, University of Queensland, St Lucia, Qld 4072, Australia

*Correspondence: E-mail: [email protected]

Ecology Letters, (2012) 15: 338–346 doi: 10.1111/j.1461-0248.2012.01743.x

� 2012 Blackwell Publishing Ltd/CNRS

Here, we test the hypothesis that elevated pCO2 (400 control, 800

and 1300 latm) alters the recruitment of a spawning coral (Acropora
millepora) by affecting the benthic algal community structure, and the

interactions between the substrata and larvae during settlement. We

used a mechanistic approach with three complementary experiments

to investigate how OA reduces larval settlement. First, to investigate

whether OA caused a shift in the community structure of the

settlement substrata to alter coral settlement, we preconditioned

settlement tiles in treatment seawater for 60 days prior to conducting

6 day settlement assays on those tiles in ambient seawater (expt. 1).

Second, we conducted the reciprocal experiment by isolating the

exposure of elevated pCO2 seawater to the coral larvae and settlement

substrata during the 6 days settlement phase only (expt. 2). Finally, we

explored whether there was a combined effect on coral settlement

when the settlement substrata and coral larvae were both exposed to

elevated pCO2 for 60 and 6 days respectively (expt. 3). From this

series of experiments, we show that OA decreases coral settlement

rates by reducing the availability of specific CCA preferred for larval

settlement, as well as interfering with the interaction between larvae

and CCA by altering the settlement behaviour of the coral larvae, such

that previously avoided substrata are preferentially selected as pCO2
increases in all three conditions.

MATERIAL AND METHODS

CO2 treatments and general protocol

Coral settlement experiments were conducted from October to

December 2009, at Heron Island Research Station, southern Great

Barrier Reef (GBR). Settlement substrata and coral larvae were

exposed to three treatments, which represented control (pH 8.04,

401 latm), and two elevated (pH 7.79, 807 latm; pH 7.60,
1299 latm) levels of future CO2 concentrations (Table 1). Treatments
were based on the worst-case stabilisation levels V (pCO2 700–

850 latm) and VI (pCO2 > 900 latm) predicted by the Intergovern-
mental Panel on Climate Change (IPCC) (Meehl et al. 2007). These

were chosen for the experiment as current CO2 emissions are tracking

the most carbon intensive levels (A1FI) predicted by the IPCC (Le

Quere et al. 2009).

As pH is reduced in a predictable manner by elevated pCO2, the

CO2 levels of the experimental seawater were controlled by adjusting

the pH of the seawater in 200 L sumps (Table 1) (see Diaz-Pulido

et al. 2011 for system details). Briefly, the total pH of the seawater was

continuously measured with temperature compensated pH electrodes

(InPro4501VP; Mettler-Toledo, Melbourne, Victoria, Australia),

which maintained the targeted pH levels with a control unit

(Aquatronica, AEB technologies, Italy) that opened solenoid valves

that injected CO2 into the seawater when pH exceeded the desired

threshold. The calibration of the pH probes was checked daily, and

recalibrated with Mettler-Toledo calibration buffers to 0.01 pH units

when necessary. Alkalinity was measured on seawater samples taken

every 6 h over a spring tidal cycle (2.4 m range) at the end of the study

period to capture the largest variation in the seawater alkalinity and

consolidate the pH treatments to the CO2 levels. Alkalinity replicates

within a sample were analysed until a maximum 2% error was met,

using a Metrohm auto-titrator at Edith Cowan University, WA. The

carbonate chemistry of the control and experimental seawater was

calculated with CO2SYS (Lewis & Wallace 2006) using pH, total

alkalinity, salinity (35.4 ppt ± 0.2 SEM; n = 8), and temperature as

the inputs, with the constants from Mehrbach et al. (1973) refitted by

Dickson & Millero (1987).

The settlement tile CO2 conditioning and settlement assays were

conducted on tiles in replicate tanks in the outdoor flow-through

aquarium system (details of the CO2 exposure times, tank and tile

replication are described below in the protocols under each

experiment and in supplementary Fig. S1). The three treatments were

fed from the 200 L sumps into replicate 12 L tanks at a mean flow

rate of 2.4 (± 0.2 SEM) L min
)1

, and each tank had a small

powerhead for extra seawater circulation. This flow rate and water

movement maintained the target pH levels, which were verified

regularly with a portable SG2 SevenGo
TM

pH meter. Replicate tanks

were randomised on the aquarium table under shade-cloth to account

for the heterogeneity in light, which averaged 406 (± 18 SEM)

lmol m)2 s)1 between 6 AM to 6 PM.

Settlement tile preparation

Unglazed terracotta settlement tiles (� 5 · 5 · 0.5 cm) were initially
preconditioned on the Heron Island reef flat (23� 26¢ 42.2¢¢ S, 151�
54¢ 47.0¢¢ E) for 5 months to develop a microbial and encrusting
community important to coral settlement (Heyward & Negri 1999).

Tiles were collected and carefully cleaned of fouling organisms using a

toothbrush, tweezers, and a plastic scraper. The tiles were then

randomly placed in replicate 12 L aquaria and conditioned in the

control and elevated CO2 treatments for 60 days prior to the

settlement assays. During this time the walls of the aquaria were

cleaned regularly to minimize any algal growth. Settlement tiles were

orientated horizontally at the bottom of the tanks and were stacked in

tile pairs with a 0.5 cm spacer, maximising the amount of cryptic

surfaces available for settlement, as coral larvae generally settle in

cryptic areas in shallow habitats (Wallace 1985).

Coral larvae collection

Gravid adult colonies of Acropora millepora were located on the Heron

Island reef flat around the time of the predicted spawning (2nd Dec

2009). Acropora millepora was chosen as a model organism as it is

Table 1 Summary of the physical and chemical seawater values for CO2 treatment levels

Treatment

Temp*

pH*

TA* pCO2� HCO3
)� CO3

2)�

XAragonite��C lmol kg)1 latm lmol kg)1 lmol kg)1

Control 26.0 (± 0.6) 8.04 (± 0.01) 2355 (± 14) 401 (± 11) 1800 (± 20) 227 (± 4) 3.6 (± 0.07)

Stabilisation level V 26.0 (± 0.6) 7.79 (± 0.01) 2365 (± 20) 807 (± 14) 2019 (± 23) 142 (± 3) 2.3 (± 0.06)

Stabilisation level VI 26.0 (± 0.6) 7.60 (± 0.01) 2363 (± 19) 1299 (± 21) 2125 (± 19) 97 (± 3) 1.6 (± 0.06)

*Temperature, pH, and total alkalinity are means (± SEM) of five replicates.

�pCO2, bicarbonate, carbonate and aragonite saturation state (X) were calculated using CO2 SYS (Lewis & Wallace 2006).

Letter Elevated CO2 alters CCA-larval interactions 339

� 2012 Blackwell Publishing Ltd/CNRS

commonly found on GBR and Indo-Pacific shallow reef flats. Five

colonies were collected and transported to outdoor aquarium facilities

where they were housed in 60 L flow-through aquaria until they

released their egg-sperm bundles. The bundles were broken apart by

gently stirring and agitating the water, and gametes from the different

colonies were collected and cross-fertilized. Fertilization took place

for 2 h, after which the embryos were collected and reared in a

laboratory at 25 �C in ambient seawater, using 200 L sumps with
aeration. At least half the seawater was changed every few hours for

the first 24 h and every 6–12 h thereafter. This removed dead larvae

and unfertilized gametes, to minimise contamination of the rearing

sumps. The larvae developed cilia and began swimming 3 days after

spawning, after which they were used for the settlement assays.

Swimming A. millepora larvae were randomly removed from the

rearing sumps, added to the experimental aquaria, and allowed 6 days

to settle (i.e. attach and metamorphose). The number of larvae added

to each tank during the settlement assays was standardised to 150

(± 10) per tile. After this time, the tiles were removed from the tanks

at random and inspected for settlement with a dissecting microscope.

Experiment 1

To isolate whether changes to the benthic community altered coral

settlement, settlement tiles were conditioned at 400, 800, and

1300 latm pCO2 for 60 days. Following the conditioning period,
coral settlement assays were conducted for 6 days on those tiles with

control seawater only. Three replicate tanks per treatment, with 8 tiles

and 1200 (± 80) larvae per tank, were used for the experiment (Fig. S1).

Experiment 2

A reciprocal experiment was conducted to determine whether

settlement was altered by elevated pCO2 exposure of the coral larvae

and the benthic community during the settlement assays only.

Settlement assays were conducted for 6 days using the three CO2
seawater treatments described above with settlement tiles that were

conditioned with control seawater only. Two replicate tanks per

treatment, with 6 tiles and 900 (± 60) larvae per tank, were used for

the experiment (Fig. S1).

Experiment 3

Finally, to investigate the combined effect of prolonged exposure of

elevated pCO2 on the benthic community and the settling larvae,

settlement tiles were conditioned in the three CO2 treatments for

60 days prior to conducting 6 day settlement assays on those tiles in

the treatment seawater described above. Three replicate tanks per

treatment, with 10 tiles and 1500 (± 150) larvae per tank, were used

for the experiment (Fig. S1).

Response variables and data analyses

We analysed total larval settlement, benthic community and CCA

community cover of the settlement tiles, and coral settlement

substrate preferences for each of the three experiments. The number

of settled (i.e. attached and metamorphosed) coral larvae was initially

quantified for all orientations of each tile. However, we only analysed

the undersides of each tile (for this and all other variables) as the

number of corals settled in this orientation accounted for ‡ 95% of
the total settlement.

The benthic community of the settlement tiles was quantified by

placing a grid on a tile, and evaluating the dominant substrate in a

square (7.5 mm
2
) using a dissecting microscope, with 224–377 squares

per tile. The substrata were characterised into eight major benthic

groups which were: bare tile, CCA, dead crustose coralline algae

(DCCA), endolithic algae found in dead crustose coralline algae

(EDCCA), turf algae found on dead crustose coralline algae

(TDCCA), turf algae (Turf), encrusting fleshy algae (EFA), and other

organisms which included biofilm, bryozoans, foraminifera, and other

encrusting organisms (Other). CCA specimens were identified to the

finest taxonomic resolution where possible and included nine CCA

taxa (see Appendix S1 in Supporting Information for details on CCA

identification). When CCA specimens could not be identified to genus

or species, they were placed in to an Unknown CCA group, which

represented � 6% of the total CCA community. See supplementary
Fig. S2 for images of the dominant benthic groups and CCA taxa.

The substrate settled on by each individual was quantified to

investigate larval settlement behaviour using Vanderploeg and Scavia�s
electivity index (E*). This index is analogous to Ivlev�s E, but
incorporates a selectivity coefficient and the number of substrata

available for settlement (Lechowicz 1982). Therefore: E* = [Wa )
(1 ⁄ n)] ⁄ [Wa + (1 ⁄ n)], where n is the total number of substrate types
available on each tile and W is the selectivity coefficient for substrate

�a� determined by: Wa = [ra ⁄ pa] ⁄
P

(ra ⁄ pa),(rb ⁄ pb)…(rz ⁄ pz), r is the
proportion of coral larvae settled on substrata a to z on each tile, and

p is the proportion of substrata a to z available for settlement on each

tile. A substrate was selected at random for larval settlement when E*

was � 0, preferably settled on when E* was > 0, and avoided for
settlement when E* was < 0. The number of coral larvae settled per tile was analysed with a generalised linear mixed effects model using Poisson distribution. We tested the effects of elevated pCO2 on counts of coral settlement amongst CO2 treatment (3 levels, fixed) with replicate tanks as a random effect and nested in CO2 treatment. The effect of elevated pCO2 on the percent cover of the broad benthic community and CCA community composition were tested using a mixed effects permuta- tional MANOVA (PERMANOVA), with the same fixed and random factors described for the previous model. When significant differences were detected (P < 0.05), pair-wise comparisons were performed to investigate treatment effects. In multivariate analyses, SIMPER analysis was used to determine the variables that characterised the dissimilarity amongst treatments. Univariate ANOVA was conducted within CCA cover to determine any significant treatment effects. All percentage cover data were sin )1 �x transformed to meet require- ments of homogeneity (permDISP) prior to analysis. Finally, we tested the effect of CO2 treatment (3 levels, fixed) on coral settlement behaviour with tanks as replicates using PERMANOVA. In all PERMANOVA main effect and pair-wise tests, we used the P-values generated by 99 999 permutations when the number of unique permutations were large, and the Monte Carlo asymptotic P-value otherwise (Anderson 2005). RESULTS We report the results of each of the three settlement experiments in turn, describing the impacts of OA on overall coral settlement density, 340 C. Doropoulos et al. Letter � 2012 Blackwell Publishing Ltd/CNRS the structure of benthic substrata, and settlement behaviour of the larvae. Results are summarised in Table 2. Experiment 1 A reduction in the cover of CCA and shift in the CCA community from elevated CO2 decreased coral settlement in the OA treatments. The reduction in settlement decreased significantly from an average of 11.0 individuals per 25 cm 2 in the control, to 1.6 and 5.5 individuals at 800 and 1300 latm, respectively (Table 2; Fig. 1a). The cover of CCAs changed dramatically in the elevated CO2 treatments with a significant decline of � 50% (ANOVA: F2,6 = 13.283; P = 0.014; Table 2; supplementary Fig. S3a). The CCA community structure also changed as pCO2 increased, with three out of ten coralline algal taxa declining with increasing CO2 concentrations (MANOVA: F2,6 = 3.286; P = 0.017; Table 2). Titanoderma spp., Hydrolithon boreale, and H. farinosum were the species that characterised the loss of CCA cover in both the elevated CO2 treatments (supplementary Fig. S3b). The settlement behaviour of the larvae, as measured by their substrate selectivity, was significantly affected by the exposure of settlement tiles to elevated pCO2 prior to the settlement assays (MANOVA: F2,6 = 4.291; P = 0.004; Table 2). Titanoderma spp. was the only preferred settlement substrate in the control treatment (E* = 0.8) and there were lower rates of settlement on all other substrata than would be expected by chance (supplementary Fig. S4a). At 800 latm, larvae did not show any clear settlement preferences and most substrata were avoided (supplementary Fig. S4b), while the larvae showed a weak preference for H. farinosum (E* = 0.2) at 1300 latm (supplementary Fig. S4c). Experiment 2 As expected, there were no differences between the broad community composition, the CCA percent cover, or the CCA community amongst the settlement tiles that were allocated for use in these settlement assays (Table 2). Yet, exposure of coral larvae and the settlement tiles to elevated pCO2 during the 6 day settlement assays significantly reduced coral settlement, as it declined from an average of 11.9 individuals per 25 cm 2 in the control, to 4.7 and 2.8 individuals at 800 and 1300 latm, respectively (Table 2; Fig. 1b). A similar disruption to larval settlement behaviour occurred to that found when only the tiles were pre-exposed to elevated pCO2 for a prolonged period of time (exp. 1). Again, coral larvae preferred to settle on Titanoderma spp. (E* = 0.75) in controls (supplementary Fig. S5a), most substrata were avoided at 800 latm (supplementary Fig. S5b), and a weak preference for H. farinosum (E* = 0.2) was found at 1300 latm (supplementary Fig. S5c). Experiment 3 Again, settlement was reduced when the tiles were conditioned in the CO2 treatments for 60 days, and 6 day settlement assays were conducted on those tiles under elevated pCO2. The magnitude of the effect was similar to whether the tiles were conditioned in the CO2 treatments for 60 days prior to the 6 day settlement assays with control seawater only (exp.1), or whether the larvae and tiles were exposed to the CO2 treatments during the 6 day settlement assays only (exp. 2) (Table 2). Increased CO2 reduced the settlement of A. millepora from an average of 9.7 individuals per 25 cm 2 in the control, to 5.2 and 4.2 individuals at 800 and 1300 latm, respectively (Fig. 1c). The reduction in settlement was significant between the control and highest CO2 treatment (P = 0.046) and marginally significant between the control and intermediate treatment (P = 0.060). The changes in tile community structure were similar to those in experiment 1, but the effects of OA appeared to be less variable in this experiment. As a result, the wider benthic community structure on the tile undersides was found to differ significantly amongst the CO2 treatments (MANOVA: F2,6 = 2.612; P = 0.003; Table 2; Fig. 2a). The loss of coralline algae was partly replaced by an increase of 8% in the cover of �bare tile� (Fig. 2a). As in experiment 1, OA led to a significant reduction in the cover of CCAs on the tiles (ANOVA: F2,6 = 40.538; P = 0.002; Table 2), characterised by declines in Titanoderma spp., H. boreale, and H. farinosum (Fig. 2b). Coral settlement behaviour was again altered significantly by elevated pCO2 (MANOVA: F2,6 = 4.224; P = 0.004; Table 2; Fig. 3). Of the 19 substrata available, Titanoderma spp. was again the only preferred settlement substrate in the control (E* = 0.6), while all other substrata were avoided (Fig. 3a). At 800 latm, Hydrolithon reinboldii was the only preferred coral settlement substrate (E* = 0.3), and all other settlement substrata were either randomly settled on or avoided (Fig 3b). No substrate was preferred for settlement at 1300 latm, with random settlement on bare tile (E* = )0.05), and all other substrata were avoided (Fig 3c). DISCUSSION In our study, the settlement density of coral larvae decreased by ‡ 45% as pCO2 increased from 400 to 800 and 1300 latm in all three Table 2 Changes to the response variables in Experiments 1, 2, and 3, comparing elevated CO2 treatments (800 and 1300 latm) to the controls (400 latm) Response variable Experiment 1 Experiment 2 Experiment 3 800 latm 1300 latm 800 latm 1300 latm 800 latm 1300 latm 1. Total settlement fl 82%*** fl 45% fl 58%*** fl 75%*** fl 50% fl 60%* 2. Benthic community structure NS NS NS NS CCA CCA** 3. CCA cover fl 47%* fl 52%* NS NS fl 42%* fl 63%*** 4. CCA community structure Titanoderma** Titanoderma* NS NS NS Titanoderma** 5a). Overall settlement behaviour Titanoderma* Titanoderma* NS NS Sporolithon* Titanoderma** 5b). Selectivity from Titanoderma fl 72% fl 74% fl 69% fl 65% fl 35% fl60% SIMPER analysis determined the variable that characterised the difference between the control and elevated CO2 treatments in multivariate analyses. Coral behaviour (5) is divided into the change in settlement preferences of the larvae (5a) among the substrate community, and (5b) from Titanoderma spp., the only preferred settlement substrate in the controls. Significance values are indicated by: NS = non-significant, * = < 0.05, ** = < 0.01, *** = < 0.001. Letter Elevated CO2 alters CCA-larval interactions 341 � 2012 Blackwell Publishing Ltd/CNRS experiments. The reduction in settlement was accompanied by a profound decline in the cover of CCA when the settlement substrata were conditioned in elevated CO2 treatments for 60 days prior to the settlement assays (expt. 1 & 3). While recent studies have also found inverse relationships between elevated pCO2 and rates of coral settlement (Albright et al. 2010; Albright & Langdon 2011; Nakamura et al. 2011), and overall CCA cover (Hall-Spencer et al. 2008; Kuffner et al. 2008; Russell et al. 2009; Fabricius et al. 2011), our study is the first to directly link benthic community cover with coral settlement and it provides three important novel insights. First, we identified the most susceptible CCA to OA and found that they are the most important taxa for coral settlement, particularly Titanoderma. Secondly, we discovered that OA reduced the affinity between the settling larvae and Titanoderma, their preferred settlement substrate. Third, we found that similar changes in settlement behaviour occurred under all three (a) (b) (c) Figure 1 Coral (Acropora millepora) settlement rates on experimental tiles (25 cm 2 ) in response to increasing pCO2. Assays occurred on (a) settlement tiles conditioned in treatment seawater for 60 days prior to 6 day larval settlement assays on those tiles with control seawater (n = 3); (b) settlement tiles and larvae exposed to treatment seawater for 6 days during the settlement assays on tiles conditioned in control seawater only (n = 2); and (c) settlement tiles and larvae exposed to treatment seawater for 60 and 6 days, respectively (n = 3). Data are means ± SEM. Significance values comparing elevated CO2 treatments to the control are indicated by: * = < 0.05, ** = < 0.01, *** = < 0.001. (a) (b) Figure 2 Percent cover of (a) the broad benthic community and (b) the crustose coralline algae community in response to increasing pCO2. Settlement tiles were exposed to the treatments for 66 days, which involved a 60 day pre-exposure period prior to the 6 day settlement assays (expt. 3). CCA = crustose coralline algae. DCCA = dead crustose coralline algae. EDCCA = endolithic algae in dead crustose coralline algae. TDCCA = turf on dead crustose coralline algae. Turf = filamentous algal turf. EFA = encrusting fleshy algae. Other = biofilm, carbonate, bryozoans, encrusting foraminifera, and unidentified. H. boreale = Hy- drolithon boreale. H. farinosum = Hydrolithon farinosum. H. reinboldii = Hydrolithon reinboldii. Data are means ± SEM; n = 3. 342 C. Doropoulos et al. Letter � 2012 Blackwell Publishing Ltd/CNRS experimental conditions. As we explain below, this surprising result implies that coral settlement behaviour is mediated by cues associated with coralline algae that appear to be highly sensitive to elevated pCO2. We designed our experiments to distinguish the effects of OA on the settling organisms (corals) from the settlement surfaces (the benthic community on the tiles). In experiment 1, we subjected tiles to a 60 day exposure to elevated pCO2 that resulted in profound changes to the coralline algal assemblage. When these tiles were then placed in control (ambient) conditions with coral planulae, the settlement behaviour of the larvae was disrupted. Because the larvae never experienced OA conditions in this experiment, the result implies that prolonged exposure of substrates to OA may alter the cues associated with CCA that are used by larvae to settle preferentially on Titanoderma. To examine the influence of OA on the settling larvae themselves (expt. 2), we exposed them to OA treatments during the 6 day settlement assays. In this case, all benthic substrata were precondi- tioned in control seawater prior to the experiment and were exposed to the treatment seawater for the 6 day period during the assays. Again, we found the same qualitative disruption to larval settlement behaviour, suggesting that the 6 day exposure of the benthic community to OA disrupted the signalling from the CCA, and potentially that the larvae may also be directly affected by elevated pCO2. In the third experiment, we found a similar qualitative result when larvae were subjected to OA and offered settlement substrates that had also been exposed …