for Coral Reef Studies, The University of Queensland, St Lucia, Queensland 4072 Australia

Transcription

1 Reports Ecology, 98(2), 216, pp by the Ecological Society of America Sensitivity of coral recruitment to subtle shifts in early community succession Christopher Doropoulos, 1,2,3 George Roff, 2 Mart-Simone Visser 2 and Peter J. Mumby 2 1CSIRO Oceans and Atmosphere, Dutton Park, Queensland 412 Australia 2Marine Spatial Ecology Laboratory, School of Biological Sciences, Australian Research Council Centre of Excellence for Coral Reef Studies, The University of Queensland, St Lucia, Queensland 472 Australia Abstract. Community succession following disturbance depends on positive and negative interactions, the strength of which change along environmental gradients. To investigate how early succession affects coral reef recovery, we conducted an 18- month experiment in Palau, using recruitment tiles and herbivore exclusion cages. One set of reefs has higher wave exposure and had previously undergone a phase shift to macroalgae following a major typhoon, whereas the other set of reefs have lower wave exposure and did not undergo a macroalgal phase shift. Similar successional trajectories were observed at all sites when herbivores were excluded: turf algae dominated early succession, followed by shifts to foliose macroalgae and heterotrophic invertebrates. However, trajectories differed in the presence of herbivores. At low wave exposure reefs, herbivores promoted coralline algae and limited turf and encrusting fleshy algae in crevice microhabitats, facilitating optimal coral recruitment. Under medium wave exposure, relatively higher but still low coverage of turf and encrusting fleshy algae (15 25%) found in crevice microhabitats inhibited coral recruitment, persisting throughout multiple recruitment events. Our results indicate that altered interaction strength in different wave environments following disturbance can drive subtle changes in early succession that cascade to alter secondary succession to coral recruitment and system recovery. Key words: cascade; facilitation; herbivory; inhibition; recovery; recruitment; resilience. Introduction In their classic paper of community succession following ecological disturbance, Connell and Slatyer (1977) proposed two theoretical models of succession: facilitation and inhibition. Facilitating organisms are earlysuccessional species that modify their environment, making it more suitable for later- successional species; whereas inhibiting organisms are early colonists that secure space and prevent invasion by subsequent colonists. The authors emphasized the importance of disturbances on modulating successional trajectories and highlighted the lack of experimental data to test the theoretical models. A surge of experiments then tested positive and negative species interactions in ecological communities, which were synthesized in a series of reviews (including Bertness and Callaway 1994, Stachowicz 21, Bruno et al. 23). Manuscript received 3 October 216; accepted 16 November 216. Corresponding Editor: Richard B. Aronson. 3 christopher.doropoulos@csiro.au Species interactions comprise both negative (competition, predation) and positive (facilitation) components, but interaction strengths often differ among environmental gradients (Paine 198). For example, Choler et al. (21) found a switch from strong competitive effects in low altitude sheltered sites to strong facilitative responses in high altitude exposed sites in alpine plant communities; and Bertness et al. (1999) found that algal canopies have strong facilitative effects in the high rocky intertidal, whereas predation was a strong determinant of community structure in the lower intertidal. In marine systems, wave exposure influences bottom- up forcing by increasing primary productivity (Leigh et al. 1987) and altering community interactions (Dayton 1975), favoring earlier successional phases as wave exposure increases. In the rocky intertidal, predators release space for secondary succession by perennial algae at wave protected sites, whereas predation is lower at higher wave exposure sites and mussels dominate a stable phase of primary succession (Lubchenco and Menge 1978). In subtidal temperate zones, herbivorous fish control habitat forming kelp in sheltered and semi- exposed sites but release from 34

2 February 216 SUCCESSION AND CORAL RECRUITMENT 35 top- down control occurs at high exposure sites and kelp proliferates (Taylor and Schiel 21). Natural ecosystems are often composed of ecosystem engineers that dictate the patterns of distribution and abundance of the entire community (Dayton 1975). Coral reefs are also an ideal model system to study early successional trajectories because they represent a classic model of disturbance- driven ecosystems, prone to the loss and recovery of the ecosystem engineer (Connell et al. 1997). Importantly, corals are secondary colonizers, so primary succession in disturbed areas of reef has major implications to subsequent coral recruitment and habitat recovery. Studies of algal succession on coral reefs have found that turf algae are typically early space occupiers of bare substrate. Subsequent succession is determined by top- down control and roving herbivorous fish deflect late- successional fleshy algal taxa by feeding on foliose and encrusting macroalgae, thereby facilitating midsuccessional crustose coralline algae (CCA; Carpenter 1986, Lewis 1986, Hixon and Brostoff 1996, Smith et al. 21). These differential trajectories then influence coral recruitment because fleshy algae inhibit coral recruitment whereas some CCA facilitate coral recruitment (reviewed in Birrell et al. 28). Thus, herbivores promote the reestablishment of corals by driving patterns of macroalgal succession, enhancing some facilitators and depressing inhibitors (Hughes et al., Arnold et al. 21, Smith et al. 21, Steneck et al. 214). Bottom- up and top- down controls are key drivers of algal succession in marine ecosystems, and it is apparent that top- down forcing generally plays a greater role (Burkepile and Hay 26). To date, most studies in coral reef systems that have explored species interactions and successional trajectories of the coral reef benthos have looked at a single environment and excluded herbivory entirely (Carpenter 1986, Lewis 1986, Hixon and Brostoff 1996, Hughes et al., Burkepile and Hay 29). Few studies have explored the effects of herbivory on successional trajectories in naturally disturbed environments, or the cascading effects to coral recruitment (but see Smith et al. 21 for an exception). Moreover, the relative importance of bottom- up vs. top- down control on early successional trajectories and subsequent impact on coral recruitment has not been assessed in areas where environmental regimes differ. Understanding the interactive effects of acute and chronic stressors on early succession is necessary because a relaxation of herbivory can occur following the loss of structural complexity, which can interact with competition and bottom- up drivers to influence system recovery (Steneck 1988, Mumby and Steneck 28, Scheffer et al. 28). Here, we take advantage of a natural disturbance event that occurred on coral reefs following a typhoon in Palau (Micronesia) to explore (1) how changes in species interactions postdisturbance affect successional trajectories of benthic communities in different environmental gradients and (2) how successional changes impact upon coral recruitment and post- disturbance recovery. Methods Site description and recent history This study was conducted in Palau, located in the western Pacific. Palau was impacted by subsequent supertyphoons in December 212 (Bopha) and November 213 (Haiyan) after no typhoon disturbances for >7 yr. The typhoons severely reduced coral cover along the eastern barrier reef and significantly reduced juvenile coral abundances (Gouezo et al. 215). Immediately following the first typhoon, a bloom of the red foliose macroalga Liagora lasted for 9 months at reefs with moderate to high wave exposure (Roff et al. 215a). The Liagora bloom immediately inhibited coral settlement (Doropoulos et al. 214) and facilitated a phase shift to the encrusting brown macroalga, Lobophora, which has persisted for 18 months (Roff et al. 215b) and throughout this study. Reefs located in lower wave exposure regimes, impacted or unimpacted by the typhoons, did not undergo any macroalgal phase shift (Roff et al. 215a). Experimental design Six sites were established within two different wave environments on the east coast of Palau (Fig. 1a, b): three nested at reefs that have lower wave exposure (<1 J/ m 3 ), and three nested within reefs that have higher wave exposure (<1 15 J/m 3 ). All three sites (LM, LN, NS) located at reefs of medium wave exposure underwent phase shifts to macroalgae following the loss of live coral from typhoon- Bopha. At the low wave exposure reefs, two sites (ES, SDO) were unimpacted by Bopha, while the third site (BL) suffered substantial loss of coral cover but did not undergo a shift to macroalgae (Roff et al. 215a, b). Sites are separated by 1 2 km within and 5 8 km between wave exposure environments. Experiments using settlement tiles and herbivore exclusion were established at 7 m depth in April 213 and surveyed until October 214. Censuses took place at, 6, 1, 145,,, and d following deployment. Herbivore access to settlement tiles was manipulated using exclusion cages and cage controls (Fig. 1c, d). Settlement tiles were custom built to investigate the influence of microhabitat complexity (exposed and crevice microhabitats) on community succession. Tiles were cast using a mix of carbonate beach sand and cement at a ratio of 2:1. Each tile measured 1 1 cm and had 24 equally spaced crowns and crevices on one side. Each crown measured cm, resulting in a total area of cm 2 for exposed microhabitats and cm 2 for crevices (incorporating crevice bottoms and vertical edges). Analyses on coral abundances incorporate unequal surface area, while analyses on percent cover were quantified from a planar view of microhabitats and are thus equal. Tiles were individually located 5 cm above the benthos in an upward facing orientation, with 15 tiles at each site separated by 1 m. Each tile was randomly

3 36 CHRISTOPHER DOROPOULOS ET AL. Ecology, Vol. 98, No ' E 134 2' E 134 3' E 134 4' E 8 ' N a b 7 5' N 7 4' N 7 3' N 7 2' N Beluu Lukes Ngederrak South 7 1' N East Sheltered Lighthouse North Lighthouse Mid 7 ' N Short Drop Off 6 5' N Reports c) medium wave exposure reefs (LM, LN, NS) d) low wave exposure reefs (BL, ES, SDO) 1 Time after deployment (d)

4 February 216 SUCCESSION AND CORAL RECRUITMENT 37 Fig. 1. (a) Map of Palau showing the (b) six study sites located in the two different environments. The three sites on the western side of the study area (LM, LN, NS) are found under medium levels of chronic wave exposure, and were impacted by a typhoon and underwent a phase shift from coral to macroalgae prior to the study. The three sites on the eastern side of the study area (BL, ES, SDO) are found under low levels of chronic wave exposure, and did not undergo a phase shift following the typhoon even though one of the sites (BL) lost significant live coral cover. Image panels are representative of community succession found on the upward facing sides of settlement tiles in caged and open fish exclusion treatments at the (c) medium and (d) low wave exposure reefs. allocated to a herbivore exclusion treatment: caged, partially caged, and open. Cages were constructed using PVCcoated galvanized wire that measured cm with a mesh size of cm. Partial cages had part of the roof and two sides removed, successfully allowing fish access to the settlement tiles (C. Doropoulos, Personal observation). In total, there were five tile replicates per herbivore exclusion treatment per site. Grazing potential An index of grazing potential was quantified using the biomass of herbivorous fish and the availability of grazeable substrate at each site in 212, 213, and 214. At each site, roving herbivorous fish (including Acanthuridae, Scaridae, and Siganidae) were surveyed using ten replicate 3 4 m transects by a single observer (P. J. Mumby). The identity and total body length (to the nearest cm) were recorded for each individual, and converted to biomass based on allometric scaling relationships. Concurrent lineintercept transects were conducted at each time point, using three replicate 3- m transects, to quantify the percent cover of benthic substrata available to grazers. Grazeable substrate includes coralline algae, encrusting and foliose fleshy algae, turf algae, and the epilithic algal matrix, and excludes live corals and other benthic invertebrates. The grazing potential was then calculated by dividing the average biomass of herbivorous fish by the average cover of grazeable substrate at each site (i.e., [g herbivorous fish]/ [m 2 grazeable substrate]). Settlement tile response variables During each census, planar images of the tiles were taken in situ to quantify the benthic community on the settlement tiles. Only the upward facing side of the tile was censused. For tiles in caged treatments, lids were opened, the census occurred, and lids were then resealed. Percent cover was quantified using Coral Point Count (Kohler and Gill 26) by randomly placing 1 points on each image and classifying the underlying functional groups: bare tile, sand, dense turf algae, cyanobacteria, crustose coralline algae, epilithic algal matrix (incorporating a sparse mix of crustose coralline and turf algae), encrusting fleshy algae (predominantly Lobopohora), foliose macroalgae (predominantly Dictyota > Padina > Caulerpa > Halimeda), heterotrophic invertebrates (including ascidians, bryozoans, sponges), and scleractinian corals. The location of each data point was classified according to occurrence in an exposed or crevice microhabitat. Given the cryptic nature and minute size of coral recruits, coral recruit abundance and microhabitat location were quantified in situ because it is not possible to observe new recruits from images. At three of the census dates (i.e.,,, and d following tile deployment), the same observer (C. Doropoulos) meticulously scanned the grid of exposed and crevice microhabitats on the upward- facing side of each tile, mapping the location and size of any recruit identified. The average (± SD) maximum diameter of the recruits was 9.3 ± 12.3 mm, the smallest of which were 1. mm. Data analysis Considering the recent disturbance history of the study area (see Site description and recent history), an initial investigation of the data was conducted to elucidate whether the sites should be grouped according to typhoon impact (impacted LM, LN, NS, BL; unimpacted ES, SDO) or wave exposure (medium LM, LN, NS; low BL, ES, SDO) because both can influence the potential impact of top- down and bottom- up processes on early community succession. The following approach was conducted. First, the grazing potential was examined at all sites before and after the typhoon disturbance. Second, comparisons of the settlement tile community were made (1) between BL and the three sites that were also impacted but have medium wave exposure, and (2) between BL and the two sites that were unimpacted but similarly have low wave exposure. Third, to test for caging effects, the partial cage on the settlement tile community was tested against the caged and open herbivore exclusion treatments. The outcome of the initial data investigation are described in Results: Environmental drivers, but briefly, the final model grouped BL according to wave exposure rather than typhoon impact and partial cages were not significantly different than open plots so were subsequently pooled. Multivariate community cover was tested using a mixed effects model with wave exposure (two levels: medium and low), herbivore exclusion (two levels: caged and open), and microhabitat (two levels: crevice and exposed) as orthogonal fixed effects, time as a continuous covariate, with sites (six levels) random and nested in wave exposure. The data was not homogeneous before or after transformation, therefore a conservative α value of.1 was used on the raw data (Underwood 1997). The multivariate analysis was conducted on a Bray- Curtis similarity matrix, using 999 permutations to generate P values. Step- wise model simplification occurred by pooling highly nonsignificant terms (P >.25). A principal coordinate analyses (PCO) was conducted to visualize the trajectories of community succession according to wave exposure and microhabitat, in caged and open herbivore exclusion

5 38 CHRISTOPHER DOROPOULOS ET AL. Ecology, Vol. 98, No. 2 6 a) tiles Medium wave exposure crevice microhabitat Medium wave exposure exposed microhabitat 4 Low wave exposure crevice microhabitat Low wave exposure exposed microhabitat PCO2 (24.4% of total variation) PCO1 (37.9% of total variation) b) tiles PCO2 (24.4% of total variation) PCO1 (37.9% of total variation) Fig. 2. Principal coordinates analysis of community succession in (a) caged and (b) open fish exclusion treatments. Trajectories are from to d in crevice (dark red filled diamonds) and exposed (light red open squares) microhabitats at medium wave exposure reefs, and crevice (light green open triangles) and exposed (dark green filled triangles) microhabitats at low wave exposure reefs. Numbers next to symbols represent number of days following deployment.

6 February 216 SUCCESSION AND CORAL RECRUITMENT 39 treatments. All multivariate routines were conducted in PERMANOVA (Anderson et al. 28). The cover of the eight most abundant functional groups were analyzed in univariate space using generalized additive mixed models (GAMMs). Analyses used the same model design as for the multivariate analysis, but temporal trends were smoothed using natural splines. Model fits and their standard errors were visualized according to time on the x- axis with environment, herbivore exclusion, and microhabitat as categorical predictors. Analyses were conducted in R (R Development Core Team 215) using gamm4 (Wood and Scheipl 213). The abundance of new coral recruits was assessed within each point in time that the in situ coral abundance surveys were conducted: (November 213), (May 214), and (October 214) days. The abundance of total coral recruits (new recruits + survivors) was also assessed for May and October 214. Abundances were analysed using generalized linear mixed models (GLMM) among wave exposure, herbivore exclusion, and microhabitats, with sites nested in wave exposure. Negativebinomial variance structure was used to account for the overdispersion of recruit abundances and zero- inflation incorporated to account for the high frequency of zerocounts. The log of microhabitat surface area was included as an offset variable to incorporate any confounding effect caused by the greater surface area available in crevices compared to exposed microhabitats but still retaining abundance data as an integer. GLMMs were conducted in R using glmmadmb (Bolker et al. 212). Finally, the relationship between coral recruitment and biotic facilitators and inhibitors were directly investigated using multiple regression models. Five analyses were conducted for new coral recruitment and total coral recruitment for the three time periods (described in previous paragraph). All fixed and random effects were pooled to isolate the effects of only benthic substrata regardless of any other predictor. Analyses used the BEST routine to select the groups of substrata that most influenced coral abundances on a Euclidean distance matrix based on Akaike information criterion. Substrata that had a significant influence on coral abundances were then modelled and visualized in univariate space using R. Multiple regression models were conducted in PERMANOVA (Anderson et al. 28). Results Environmental drivers Initial inspection of the data revealed three patterns in environmental drivers that influence the understanding of the system and experimental design. First, there was a 7% reduction in the herbivorous fish grazing potential at the four sites that resulted from the loss of the majority of live coral cover following typhoon Bopha (LM, LN, NS, BL), decreasing from an average of 51.8 g/m 2 grazeable substrate in 212 to 14.7 g/m 2 in In comparison, at the two sites unimpacted by the typhoon (ES, SDO), there was no change in grazing potential between 212 and 213, averaging 32.8 and 29. g/m 2 respectively, but a 35% increase to 44.3 g/m 2 in 214. Second, phase shifts from coral to macroalgae occurred at the three sites impacted by the typhoon and located in medium wave exposure environments (LM, LN, NS), whereas sites located in low wave exposure environments did not undergo macroalgal phase shifts whether they were impacted by the typhoon (BL) or not (ES, SDO) (Roff et al. 215a, b). Planned comparisons in settlement tile community structure revealed that BL differed significantly from the three impacted sites (LM, LN, NS) under medium wave exposure (P =.1) but not the two unimpacted sites under similarly low wave exposure (P =.76). This effect was especially apparent in the open but not caged exclusion treatments (Appendix S1: Fig. S1). Thus, BL was pooled with two low wave exposure sites. Comparisons in settlement tile community structure among exclusion treatments showed highly significant differences between the partial and caged treatments (P =.2) and open and caged treatments (P =.2) across all sites, and no difference between the open and partially caged treatments (P =.26). Therefore, open and partial treatments were pooled into a single open treatment. Community succession and herbivory Multivariate community composition showed two significant three- way interactions: time exclusion site (exposure) (P =.1) and time exposure exclusion (P =.2). Community succession did not differ between low and medium wave exposure sites when herbivores were excluded (Fig. 2a; P =.188), but distinct trajectories occurred in the presence of herbivores (Fig. 2b; P =.1). Succession also diverged between open and caged treatments at low wave exposure sites from to d (Appendix S1: Fig. S2a; P =.1). In contrast, trajectories were not distinct at medium wave exposure sites (Appendix S1: Fig. S2b), even though there was a significant difference on community structure between herbivore exclusion treatments (P =.16). Functional- group succession When the major functional groups were analysed in univariate space (Fig. 3), differences between microhabitats became prominent as well as the effect of herbivore exclusion under medium and low wave exposure. There were no differences in turf algal cover between wave exposures or microhabitats when herbivores were excluded, but in the presence of herbivores turf cover was highest in crevice microhabitats at medium wave exposure sites and lowest on exposed microhabitats at low wave exposure sites (Fig. 3a; exposure exclusion, P <.1; exposure microhabitat, P =.1; exclusion microhabitat, P =.14). Sand was more than five times higher in crevices than exposed microhabitats but this quickly declined from to 15 d (Fig. 3b; Microhabitat, P <.1). Encrusting fleshy algae (EFA) was two times more

7 31 CHRISTOPHER DOROPOULOS ET AL. Ecology, Vol. 98, No. 2 Cover (%) 4 a) Turf algae b) Sand Cover (%) c) Encrusting fleshy algae d) Epilithic algal matrix Cover (%) e) Crustose coralline algae f) Foliose macroalgae g) Invertebrates (ascidians, bryozoans, sponges) h) Coral 2 2 Cover (%) Coral abundance (cm -2 ) i) New coral recruitment Time (d) j) Total coral recruitment Time (d) Medium wave exposure crevice microhabitat Medium wave exposure exposed microhabitat Low wave exposure crevice microhabitat Low wave exposure exposed microhabitat

8 February 216 SUCCESSION AND CORAL RECRUITMENT 311 Fig. 3. Temporal trends (days) of generalized additive mixed models (GAMM) means (±SE) for (a h) percent cover of benthic functional groups and (i, j) coral recruit abundance per cm 2 according to fish exclusion treatments (caged, open), wave exposure (medium = red, low = green), and microhabitat (exposed = light shades, crevice = dark shades). Sampling for percent cover analyses were conducted at, 6, 1, 145,,, and d following deployment. Arrows on x- axes represent in situ sampling times for coral recruits (,, and d following deployment). Note different scales on y-axes. abundant at medium wave exposure sites, where herbivore exclusion had a greater effect on EFA abundance than at low wave exposure sites (Fig. 3c; exposure exclusion, P <.1). Crustose coralline algae (CCA) showed the opposite pattern (Fig. 3e). It was twice as abundant at low than at medium wave exposure sites, herbivore exclusion reduced CCA cover by more than two times at low wave exposure sites, and the highest CCA cover was observed on exposed microhabitats (exposure exclusion, P <.1; exclusion microhabitat, P =.31). Macroalgae (Fig. 3f) and invertebrates (Fig. 3g) became highly abundant after 2 d and had much higher cover in caged than open treatments (exclusion, P <.1), with higher Invert cover at medium wave exposure sites (exposure exclusion, P =.6). Coral cover was very low relative to other groups, but became apparent after 5 d and highest in crevices of the open treatment under low wave exposure (Fig. 3h; exposure exclusion, P =.25; exclusion microhabitat, P =.17). Coral recruitment In total, 187 new recruits were quantified in the three censuses. The initial census in November 213, showed that coral recruitment was nine times higher in crevices than exposed microhabitats at low wave exposure sites regardless of fish exclusion (Fig. 3i; microhabitat, P <.1). In May 214, the addition of new recruits (Fig. 3i) and total recruitment (Fig. 3j) both had abundances highest in crevices but only at low wave exposure sites (exposure microhabitat, P <.1). There was no effect of fish exclusion at this time point. The final census in October 214 showed the same pattern for new (Fig. 3i; exposure microhabitat, P =.9; exclusion microhabitat, P =.4) and total recruitment (Fig. 3j; exposure microhabitat, P =.1; exclusion microhabitat, P <.1), and the effect of fish exclusion became prominent at low wave exposure sites. The abundance of new coral recruits was two times higher and total recruitment was more than four times higher in crevices of open treatments compared to those that excluded fish (Appendix S1: Fig. S3). Coral facilitation and inhibition Across all study sites, the only benthic algal group that facilitated coral recruitment was crustose coralline algae (CCA), whereas encrusting fleshy algae (EFA) and turf algae (TA) inhibited recruitment (Appendix S1: Figs. S4 S8). CCA had a positive linear effect (P <.1) on coral recruit abundances at every time point throughout succession. In contrast, EFA had a weak negative linear effect on new recruit abundances in the earlier phases of succession in November 213 and May 214, and TA had a nonlinear negative effect on new and total coral recruit abundances in the later phases of succession in May and October 214. Discussion While it has previously been shown that the strength of top- down interactions on community structure weakens as wave exposure increases in temperate systems (Dayton 1975, Lubchenco and Menge 1978, Taylor and Schiel 21), changes in interaction strengths across environmental gradients following disturbance are poorly documented. Consistent with previous studies of Pacific reefs (Smith et al. 21, Doropoulos et al. 216, Mumby et al. 216), we found that exclusion of herbivores resulted in a gradual shift toward an algal dominated system at both medium and low wave exposure sites. Macroalgal cover ranged between 1 2% after 25 d, similar to other studies in the Pacific but three times lower than studies from the Caribbean (Roff and Mumby 212). However, even though an experimental reduction of herbivory equally limited secondary succession to macroalgae and invertebrate dominance in both wave environments, a different pattern emerged when herbivores were free to forage; here, it was the dominance of early successional turf and encrusting fleshy algae and scarcity of CCA in crevice microhabitats at medium wave exposure reefs that inhibited coral recruitment. In the presence of herbivores at low wave exposure sites however, higher abundances of CCA, coupled with lower abundances of turf and encrusting fleshy algae, cascaded to facilitate high rates of coral recruitment. Thus, the altered interaction strength between biotic (herbivory) and abiotic (wave exposure) forces caused subtle changes to benthic community trajectories at the microhabitat scale that cascade to limit coral recruitment and system recovery. In the absence of herbivores at medium and low wave exposure sites, while the cover of foliose macroalgae and turf algae remained similar, the cover of encrusting fleshy algae (mainly Lobophora spp.) was two times higher in caged tiles at the medium than low wave exposure sites. As Lobophora has a limited dispersal range and recruits locally at the scale of meters (van Steveninck and Breeman 1987), it seems likely that the increase in encrusting fleshy algae at the medium wave exposure sites resulted from local recruitment from the adjacent reef that had undergone a phase shift to Lobophora following the typhoon disturbance (Roff et al. 215b). In the presence of herbivores, the interaction strength of herbivory on early community succession decreased from the low to medium wave exposure sites. Under low wave exposure,

9 312 CHRISTOPHER DOROPOULOS ET AL. Ecology, Vol. 98, No. 2 top- down control influenced successional trajectories in both crevice and exposed microhabitats, with rapid transitions in early compared to later successional stages (compare 145 d with 145 d in the ordination). This resulted in an early successional community dominated by CCA after 25 d, significant to the facilitation of coral recruits. In contrast, successional trajectories of the uncaged treatment at medium wave exposure sites were similar to caged treatments, undergoing circularity between 145 and d. The early successional turfdominated community transitioned to a community dominated by encrusting fleshy algae, and the presence of these cryptic benthic algae inhibited coral recruitment. It appears that the dominance of turf and encrusting fleshy algae in crevice microhabitats of tiles open to herbivory at medium wave exposure sites resulted from the altered interaction strength between top- down and bottom- up forcing. The immediate loss of coral cover (~7% cover to zero) at four of the study sites following typhoon Bopha (Doropoulos et al. 214) resulted in a substantial relaxation of grazing intensity due to the increase in substrata available to herbivory (Mumby and Steneck 28). Coupled with the higher wave exposure at three of the four disturbed sites, recruitment pulses of macroalgae from adjacent habitats that had undergone phase shifts to benthic macroalgae (Roff et al. 215b) were able to saturate the area and escape grazing (Steneck 1988, Doropoulos et al. 213). In contrast, the impacted site under low wave exposure did not undergo a shift to benthic macroalgae following the loss of live coral cover even though grazing intensity had relaxed, because wave exposure was insufficient to drive a net build- up of macroalgae (Roff et al. 215a, b). Ultimately, our study also found that the successional community was more strongly influenced by wave exposure rather than the loss of coral cover and relaxation of grazing intensity in the presence of herbivores. Previous work using settlement tiles with microhabitat complexity has confirmed that crevice microhabitats provide the optimal coral recruitment space (Doropoulos et al. 216). In our current study, differences in the abundances of known coral settlement facilitators and inhibitors in crevice microhabitats between the two areas of wave exposure when herbivores were present are most apparent. The cover of CCA was five times higher in crevices at low wave exposure sites, whereas turf and encrusting fleshy algae were two times higher in crevices at medium wave exposure sites. CCA has repeatedly been shown to facilitate coral settlement in lab experiments (e.g., Harrington et al. 24, Birrell et al. 28, Doropoulos et al. 216) and was a significant positive coral recruitment facilitator in this present study. As an inhibitor, turf algae have varied effects on coral recruitment depending on turf length and density (Birrell et al. 25, Arnold et al. 21), exemplified by the non- linear negative effects observed in this study. Thus, the regulation by herbivores of these algal groups in cryptic microhabitats appears to have significant cascading effects on the facilitation of coral recruitment. Recent work has also highlighted the interaction among turf, sediments and herbivores as sensitive indicators of reef degradation (Goatley et al. 216). As large scale disturbances that were once rare become more frequent and severe, understanding the early responses of community succession in different environments is key for the recovery of ecosystem engineers and the ecosystem services they support. While most previous studies have demonstrated that profound increases in turf and macroalgal cover (>5%) driven by biotic and abiotic interactions can inhibit the recruitment of corals (e.g., Hughes et al., Arnold et al. 21, Dixson et al. 214, Steneck et al. 214), our study shows that relatively minor changes occurring at the microscale can have cascading effects. Coral recruits were detectable after 25 d at low wave exposure sites, whereas at medium wave exposure sites, even relatively low levels of encrusting fleshy macroalgae had negative effects on coral recruitment. Macroalgal blooms on reefs can affect coral settlement and post- settlement survival through indirectly mediated chemical inhibition (Dixson et al. 214, Doropoulos et al. 214), and it has recently been suggested that Pacific corals are particularly sensitive to macroalgae compared to Caribbean corals (Mumby et al. 216). Interestingly, despite foliose macroalgae (Dictyota, Padina, Caulerpa) and heterotrophs (ascidians, sponges, byrozoans) reaching higher levels of cover in herbivore exclusions, neither functional- group had a strong effect on coral recruitment, further implicating Lobophora as a key inhibitor of recruitment in phase shifts in the Pacific (Mumby et al. 216). Collectively, our results show that altered interaction strengths along environmental gradients can drive changes in early succession of the benthos that cascade to inhibit system recovery following disturbance. In our study, most notable was the weakening strength of herbivory between areas of low to moderate wave exposure on coral recruitment facilitators and inhibitors, which cascaded to high and low coral recruitment. On coral reefs where herbivore biomass remains high after disturbance, grazing intensity may not be weakened sufficiently to be overcome by higher wave exposure, and herbivory may promote coral recruitment and trajectories of recovery. Our results show that even subtle changes in the early stages of benthic community succession can cascade to profound effects on coral recruitment. Hence, wave exposure appears to be driving both ultimate and proximate effects in this system: ultimate in terms of the historical impact, which led to a phase shift (Roff et al. 215b), and proximate in terms of altering interaction strengths and subsequent reef recovery. This is likely to dampen reef recovery rates as chronic wave exposure increases, a consideration affecting natural resource management following disturbance (Gouezo et al. 215). Acknowledgments We thank Yimnang Golbuu and staff at PICRC for support throughout this project. Clare Gallagher, Alyssa Marshell,

10 February 216 SUCCESSION AND CORAL RECRUITMENT 313 María Catalina Reyes- Nivia, Alberto Rodríguez- Ramírez, Alice Rogers, Katharine Sampson, and Johanna Werminghausen provided valuable field assistance. Funding was received through an Australian Endeavour Award Postdoctoral Fellowship to C. Doropoulos, and an Australian Research Council Laureate Fellowship to P. J. Mumby. We are grateful to the Handling Editor and two anonymous reviewers for comments that substantially improved the manuscript. Literature Cited Anderson, M. J., R. N. Gorley, and K. R. Clarke. 28. PERMANOVA+ for PRIMER: guide to software and statistical methods. PRIMER-E, Plymouth, UK. Arnold, S. N., R. S. Steneck, and P. J. Mumby. 21. Running the gauntlet: inhibitory effects of algal turfs on the processes of coral recruitment. Marine Ecology Progress Series 414: Bertness, M. D., and R. Callaway Positive interactions in communities. Trends in Ecology and Evolution 9: Bertness, M. D., G. H. Leonard, J. M. Levine, P. R. Schmidt, and A. O. Ingraham Testing the relative contribution of positive and negative interactions in rocky intertidal communities. Ecology 8: Birrell, C. L., L. J. McCook, and B. L. Willis. 25. Effects of algal turfs and sediment on coral settlement. Marine Pollution Bulletin 51: Birrell, C. L., L. J. McCook, B. L. Willis, and G. A. Diaz-Pulido. 28. Effects of benthic algae on the replenishment of corals and the implications for the resilience of coral reefs. Oceanography and Marine Biology 46: Bolker, B., H. Skaug, A. Magnusson, and A. Nielsen Getting started with the glmmadmb package. glmmadmb. r-forge. r-project. org/glmmadmb Bruno, J. F., J. J. Stachowicz, and M. D. Bertness. 23. Inclusion of facilitation into ecological theory. Trends in Ecology and Evolution 18: Burkepile, D. E., and M. E. Hay. 26. Herbivore vs. nutrient control of marine primary producers: context- dependent effects. Ecology 87: Burkepile, D. E., and M. E. Hay. 29. Nutrient versus herbivore control of macroalgal community development and coral growth on a Caribbean reef. Marine Ecology Progress Series 389: Carpenter, R. C Partitioning herbivory and its effects on coral reef algal communities. Ecological Monographs 56: Choler, P., R. Michalet, and R. M. Callaway. 21. Facilitation and competition on gradients in alpine plant communities. Ecology 82: Connell, J. H., and R. O. Slatyer Mechanisms of succession in natural communities and their role in community stability and organization. American Naturalist 111: Connell, J. H., T. P. Hughes, and C. C. Wallace A 3- year study of coral abundance, recruitment, and disturbance at several scales in space and time. Ecological Monographs 67: Dayton, P. K Experimental evaluation of ecological dominance in a rocky intertidal algal community. Ecological Monographs 45: Dixson, D. L., D. Abrego, and M. E. Hay Chemically mediated behavior of recruiting corals and fishes: A tipping point that may limit reef recovery. Science 345: Doropoulos, C., G. A. Hyndes, D. Abecasis, and A. Vergés Herbivores strongly influences algal recruitment in both coral and algal dominated coral reef habitats. Marine Ecology Progress Series 486: Doropoulos, C., G. Roff, M. Zupan, V. Nestor, A. L. Isechal, and P. J. Mumby Reef- scale failure of coral settlement following typhoon disturbance and macroalgal bloom in Palau, Western Pacific. Coral Reefs 33: Doropoulos, C., G. Roff, Y.-M. Bozec, M. Zupan, J. Werminghausen, and P. J. Mumby Characterizing the ecological trade- offs throughout the early ontogeny of coral recruitment. Ecological Monographs 86:2 44. Goatley, C. H. R., R. M. Bonaldo, R. J. Fox, and D. R. Bellwood Sediments and herbivory as sensitive indicators of coral reef degradation. Ecology and Society 21:29. Gouezo, M., Y. Golbuu, R. van Woesik, L. Rehm, S. Koshiba, and C. Doropoulos Impact of two sequential super typhoons on coral reef communities in Palau. Marine Ecology Progress Series 54: Harrington, L., K. Fabricius, G. De Ath, and A. Negri. 24. Recognition and selection of settlement substrata determine post- settlement survival in corals. Ecology 85: Hixon, M. A., and W. N. Brostoff Succession and herbivory: Effects of differential fish grazing on Hawaiian coralreef algae. Ecological Monographs 66:67 9. Hughes, T. P., M. J. Rodrigues, D. R. Bellwood, D. Ceccarelli, O. Hoegh-Guldberg, L. McCook, N. Moltschaniwskyj, M. S. Pratchett, R. S. Steneck, and B. Willis.. Phase shifts, herbivory, and the resilience of coral reefs to climate change. Current Biology 17: Kohler, K. E., and S. M. Gill. 26. Coral Point Count with Excel extensions (CPCe): a visual basic program for the determination of coral and substrate coverage using random point count methodology. Computers and Geosciences 32: Leigh, E. G., R. T. Paine, J. F. Quinn, and T. H. Suchanek Wave energy and intertidal productivity. Proceedings of the National Academy of Sciences USA 84: Lewis, S. M The role of herbivorous fishes in the organization of a Caribbean reef community. Ecological Monographs 56: Lubchenco, J., and B. A. Menge Community development and persistence in a low rocky intertidal zone. Ecological Monographs 48: Mumby, P. J., and R. S. Steneck. 28. Coral reef management and conservation in light of rapidly evolving ecological paradigms. Trends in Ecology and Evolution 23: Mumby, P. J., R. S. Steneck, M. Adjeroud, and S. N. Arnold High resilience masks underlying sensitivity to algal phase shifts of Pacific coral reefs. Oikos 125: Paine, R. T Food webs: linkage, interaction strength and community infrastructure. Journal of Animal Ecology 49: R Development Core Team R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Roff, G., and P. J. Mumby Global disparity in the resilience of coral reefs. Trends in Ecology and Evolution : Roff, G., I. Chollett, C. Doropoulos, Y. Golbuu, R. S. Steneck, and P. J. Mumby. 215a. Exposure- driven phase shift following catastrophic disturbance on coral reefs. Coral Reefs 34: Roff, G., C. Doropoulos, M. Zupan, A. Rogers, R. Steneck, Y. Golbuu, and P. Mumby. 215b. Phase shift facilitation following cyclone disturbance on coral reefs. Oecologia 178:

11 314 CHRISTOPHER DOROPOULOS ET AL. Ecology, Vol. 98, No. 2 Scheffer, M., E. H. Van Nes, M. Holmgren, and T. Hughes. 28. Pulse- driven loss of top- down control: the critical- rate hypothesis. Ecosystems 11: Smith, J. E., C. L. Hunter, and C. M. Smith. 21. The effects of top- down versus bottom- up control on benthic coral reef community structure. Oecologia 163: Stachowicz, J. J. 21. Mutualism, facilitation, and the structure of ecological communities. BioScience 51: Steneck, R. S Herbivory on coral reefs: a synthesis. Proceedings of the 6th International Coral Reef Symposium 1: Steneck, R. S., S. N. Arnold, and P. J. Mumby Experiment mimics fishing on parrotfish: insights on coral reef recovery and alternative attractors. Marine Ecology Progress Series 56: Taylor, D. I., and D. R. Schiel. 21. Algal populations controlled by fish herbivory across a wave exposure gradient on southern temperate shores. Ecology 91: Underwood, A. J Experiments in ecology: their logistical design and interpretation using analysis of variance. Cambridge University Press, Cambridge, UK. van Steveninck, E., and A. Breeman Deep water vegetations of Lobophora variegata (Phaeophyceae) in the coral reef of Curacao population dynamics in relation to mass mortality of the sea urchin Diadema antillarum. Marine Ecology Progress Series 36:81 9. Wood, S., and F. Scheipl gamm4: Generalized additive mixed models using mgcv and lme4. R package version Supporting Information Additional supporting information may be found in the online version of this article at doi/1.12/ecy.1663/suppinfo